CN114971062A - Photovoltaic power prediction method and device - Google Patents
Photovoltaic power prediction method and device Download PDFInfo
- Publication number
- CN114971062A CN114971062A CN202210665580.0A CN202210665580A CN114971062A CN 114971062 A CN114971062 A CN 114971062A CN 202210665580 A CN202210665580 A CN 202210665580A CN 114971062 A CN114971062 A CN 114971062A
- Authority
- CN
- China
- Prior art keywords
- processed
- data
- hidden state
- traversal
- state corresponding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000005070 sampling Methods 0.000 claims abstract description 192
- 230000002159 abnormal effect Effects 0.000 claims abstract description 53
- 230000010354 integration Effects 0.000 claims abstract description 12
- 230000000737 periodic effect Effects 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims description 41
- 239000011159 matrix material Substances 0.000 claims description 33
- 239000013598 vector Substances 0.000 claims description 21
- 238000007781 pre-processing Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000008030 elimination Effects 0.000 claims description 4
- 238000003379 elimination reaction Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 description 19
- 238000004891 communication Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000012549 training Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 238000012731 temporal analysis Methods 0.000 description 3
- 238000000700 time series analysis Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000001351 cycling effect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2474—Sequence data queries, e.g. querying versioned data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
- G06F17/13—Differential equations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Human Resources & Organizations (AREA)
- General Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Operations Research (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Marketing (AREA)
- Fuzzy Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Probability & Statistics with Applications (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Computational Linguistics (AREA)
- Algebra (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Photovoltaic Devices (AREA)
Abstract
The application provides a photovoltaic power prediction method and a photovoltaic power prediction device, which comprise the following steps: the method comprises the steps of obtaining processed meteorological data and processed operation data at a plurality of sampling moments, setting and predicting the processed meteorological data and the processed operation data at the plurality of sampling moments to obtain a target hidden state corresponding to a real predicted value of the photovoltaic power, and performing time integration on the target hidden state to obtain a predicted value of the photovoltaic power within a set time. The method comprises the steps that abnormal meteorological data are removed and normalized from initial meteorological data obtained by periodic sampling, so that the processed meteorological data at multiple sampling moments are sporadic data and do not have periodicity.
Description
Technical Field
The application relates to the field of photovoltaic power generation, in particular to a photovoltaic power prediction method and device.
Background
The electric power company needs to predict the photovoltaic power generation power in order to perform the power generation scheduling operation. Accurate photovoltaic power prediction is an important, cost-effective energy management element that helps photovoltaic plants and aggregate systems to efficiently and directly participate in the electricity market and to increase revenue by optimizing supply plans.
At present, photovoltaic power prediction is basically a time series analysis-based method, and the time series analysis-based method refers to: and predicting the photovoltaic power through data acquired at fixed time intervals. However, under a real working condition, the data acquired at the fixed time interval may have data pollution, data loss and the like, so that the data acquired at the fixed time interval becomes sporadic data, and the photovoltaic power predicted by the time series analysis-based method is inaccurate.
Disclosure of Invention
In view of this, the present application provides a method and an apparatus for photovoltaic power prediction, which are used for performing photovoltaic power prediction on null-satellite data, and the technical scheme is as follows:
a photovoltaic power prediction method, comprising:
acquiring processed meteorological data and processed operation data at a plurality of sampling moments, wherein the processed meteorological data are obtained by performing abnormal meteorological data elimination and normalization on initial meteorological data obtained by periodic sampling, the processed operation data are obtained by performing pretreatment on the initial operation data, the pretreatment is used for marking abnormal operation data in the initial operation data and normalizing unmarked operation data, and the initial operation data are photovoltaic system operation data acquired at the sampling moments corresponding to the initial meteorological data from which the abnormal meteorological data are eliminated;
setting and predicting the processed meteorological data and the processed operating data at a plurality of sampling moments to obtain a target hidden state corresponding to a real predicted value of the photovoltaic power;
and performing time integration on the target hidden state to obtain a photovoltaic power predicted value within a set time.
Optionally, the setting and predicting the processed meteorological data and the processed operating data at a plurality of sampling moments to obtain the target hidden state corresponding to the real predicted value of the photovoltaic power includes:
traversing the processed meteorological data and the processed operation data at each sampling moment according to the sequence of the sampling moments:
for the processed meteorological data and the processed operation data at the current traversed sampling time, determining a first hidden state corresponding to the current traversal according to the processed meteorological data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, wherein the first hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a theoretical predicted value of photovoltaic power obtained based on the processed meteorological data under the current traversal and each forward traversal, and if the current traversal is the first traversal, the second hidden state corresponding to the previous traversal is a preset initial hidden state;
updating a first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling moment to obtain a second hidden state corresponding to the current traversal, wherein the second hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a real predicted value of the photovoltaic power obtained based on the current traversal and the updating of the processed operation data at each previous traversal;
and taking the second hidden state corresponding to the last traversal as a target hidden state.
Optionally, determining a first hidden state corresponding to the current traversal according to the processed meteorological data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, where the determining includes:
determining an updating gate control and an updating vector corresponding to the current traversal according to the processed meteorological data at the current traversed sampling moment and a second hidden state corresponding to the previous traversal;
and obtaining a first hidden state corresponding to the current traversal based on the second hidden state corresponding to the previous traversal, the update gating and the update vector corresponding to the current traversal.
Optionally, based on the processed running data at the currently traversed sampling time, updating the first hidden state corresponding to the current traversal to obtain a second hidden state corresponding to the current traversal, where the updating includes:
acquiring an identification matrix corresponding to the processed running data at the currently traversed sampling time, wherein each element of the identification matrix is respectively used for indicating whether a component contained in the processed running data at the currently traversed sampling time is marked abnormal running data;
for each component contained in the processed operation data at the currently traversed sampling time, determining a sparse component corresponding to the component according to the component, an element of an identification matrix corresponding to the component and a first hidden state corresponding to the current traversal, so as to obtain sparse operation data corresponding to the processed operation data at the currently traversed sampling time;
and obtaining a second hidden state corresponding to the current traversal according to the sparse running data corresponding to the processed running data at the current traversed sampling moment and the first hidden state corresponding to the current traversal.
Optionally, the processed meteorological data includes one or more of the following meteorological data: solar irradiance, wind direction, wind speed and ambient temperature in the plane;
the processed operational data includes one or more of the following operational data: maximum power point current, voltage and power, output value and module temperature of the output end of the photovoltaic array, inverter alternating side output value, and solar position parameters.
Optionally, the processed meteorological data and the processed operation data at the multiple sampling moments include: processed meteorological data and processed operational data at low irradiance, and processed meteorological data and processed operational data at high irradiance;
the method for setting and predicting the processed meteorological data and the processed operation data at a plurality of sampling moments comprises the following steps:
and respectively carrying out setting prediction processing on the processed meteorological data and the processed operation data under the low irradiance, and the processed meteorological data and the processed operation data under the high irradiance.
A photovoltaic power prediction apparatus, comprising: the device comprises a data acquisition module, a setting prediction processing module and a photovoltaic power prediction value determination module;
the data acquisition module is used for acquiring processed meteorological data and processed operation data at a plurality of sampling moments, wherein the processed meteorological data are obtained by carrying out abnormal meteorological data elimination and normalization on initial meteorological data obtained by periodic sampling, the processed operation data are obtained by preprocessing the initial operation data, the preprocessing is used for marking abnormal operation data in the initial operation data and normalizing unmarked operation data, and the initial operation data are photovoltaic system operation data acquired at the sampling moments corresponding to the initial meteorological data from which the abnormal meteorological data are eliminated;
the setting and predicting module is used for performing setting and predicting processing on the processed meteorological data and the processed operation data at a plurality of sampling moments to obtain a target hidden state corresponding to a real predicted value of the photovoltaic power;
and the photovoltaic power predicted value determining module is used for performing time integration on the target hidden state to obtain a photovoltaic power predicted value within set time.
Optionally, the setting prediction processing module includes: the device comprises a traversing module and a target hidden state determining module;
and the traversing module is used for traversing the processed meteorological data and the processed operation data at each sampling moment according to the sequence of the sampling moments: for the processed meteorological data and the processed operation data at the current traversed sampling time, determining a first hidden state corresponding to the current traversal according to the processed meteorological data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, wherein the first hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a theoretical predicted value of photovoltaic power obtained based on the processed meteorological data under the current traversal and each forward traversal, and if the current traversal is the first traversal, the second hidden state corresponding to the previous traversal is a preset initial hidden state; updating a first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling moment to obtain a second hidden state corresponding to the current traversal, wherein the second hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a real predicted value of the photovoltaic power obtained based on the current traversal and the updating of the processed operation data at each previous traversal;
and the target hidden state determining module is used for taking the second hidden state corresponding to the last traversal as the target hidden state.
Optionally, when the traversal module determines the first hidden state corresponding to the current traversal according to the processed meteorological data at the currently traversed sampling time and the second hidden state corresponding to the previous traversal, the method includes: the system comprises a first GRU network computing module and an ordinary differential network computing module;
the first GRU network computing module is used for determining an updating gate control and an updating vector corresponding to the current traversal according to the processed meteorological data at the current traversed sampling moment and a second hidden state corresponding to the previous traversal;
and the ordinary differential network computing module is used for obtaining a first hidden state corresponding to the current traversal based on the second hidden state corresponding to the previous traversal and the updating gating and updating vector corresponding to the current traversal.
Optionally, the updating, by the traversal module, the first hidden state corresponding to the current traversal based on the processed running data at the currently traversed sampling time to obtain the second hidden state corresponding to the current traversal includes: the device comprises an identification matrix acquisition module, a sparse processing module and a second GRU network calculation module;
the identification matrix acquisition module is used for acquiring an identification matrix corresponding to the processed running data at the currently traversed sampling time, wherein each element of the identification matrix is respectively used for indicating whether a component contained in the processed running data at the currently traversed sampling time is marked abnormal running data or not;
the sparse processing module is used for determining a sparse component corresponding to each component contained in the processed running data at the current traversed sampling moment according to the component, the element of the identification matrix corresponding to the component and the first hidden state corresponding to the current traversal, so as to obtain sparse running data corresponding to the processed running data at the current traversed sampling moment;
and the second GRU network computing module is used for obtaining a second hidden state corresponding to the current traversal according to the sparse running data corresponding to the processed running data at the current traversed sampling time and the first hidden state corresponding to the current traversal.
According to the technical scheme, the photovoltaic power prediction method comprises the steps of firstly obtaining processed meteorological data and processed operation data at a plurality of sampling moments, then conducting set prediction processing on the processed meteorological data and the processed operation data at the plurality of sampling moments to obtain a target hidden state corresponding to a real predicted value of photovoltaic power, and finally conducting time integration on the target hidden state to obtain a photovoltaic power predicted value within set time. The method comprises the steps that the processed meteorological data at the multiple sampling moments are obtained by removing abnormal meteorological data from initial meteorological data obtained by periodic sampling and normalizing the initial meteorological data, the processed meteorological data at the multiple sampling moments are sporadic data and no longer have periodicity, the processed meteorological data and the processed running data at the sporadic multiple sampling moments can be set and predicted to obtain a target hidden state, and the target hidden state is a hidden state corresponding to a real predicted value of photovoltaic power, so that the target hidden state is subjected to time integration, and a photovoltaic power predicted value with higher accuracy can be obtained.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic flowchart of a photovoltaic power prediction method according to an embodiment of the present disclosure;
FIG. 2 is a schematic diagram of two-stage hidden state hopping of a photovoltaic power prediction model;
fig. 3 is a schematic structural diagram of a photovoltaic power prediction apparatus provided in an embodiment of the present application;
fig. 4 is a block diagram of a hardware structure of a photovoltaic power prediction apparatus according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
In view of the problems in the prior art, the inventor of the present invention has made extensive research and thought that a photovoltaic power prediction model may be constructed based on a time sequence under the condition of irregular and sporadic observation of data sampling time intervals, and photovoltaic power prediction is realized by adopting a data-driven mode. When the photovoltaic power prediction model is trained, the sporadic data corresponding to historical moments of 2 years or more are divided into a training set and a test set, wherein 10 different data sets can be extracted from an original time sequence (the sporadic data corresponding to the historical moments of 1 year), the data sets are sequentially or randomly divided by taking 10%, 30%, 50% and 70% as dividing points, any one of the divided parts is selected as the training set, the photovoltaic power prediction model is trained based on the training set, the trained photovoltaic power prediction model is tested based on the test set, the trained photovoltaic power prediction model is obtained, then the sporadic data to be predicted are input to the trained photovoltaic power prediction model, and the photovoltaic power prediction value in set time can be obtained.
Based on the above concept, the present application finally provides a photovoltaic power prediction method, which can be applied to the photovoltaic power prediction model (optionally, the photovoltaic power prediction model is a model obtained by modeling based on a GRU-ordinary differential-bayes update network, where an english language of a GRU is collectively referred to as a Gate recovery Unit, that is, a Gate control cycle Unit), and the following embodiments are used to describe the photovoltaic power prediction method provided by the present application in detail.
Referring to fig. 1, a schematic flow chart of a photovoltaic power prediction method provided in an embodiment of the present application is shown, where the photovoltaic power prediction method may include:
step S101, acquiring processed meteorological data and processed operation data at a plurality of sampling moments.
In this step, if the photovoltaic power prediction is performed based on the photovoltaic power prediction model, it is necessary to first obtain the processed meteorological data and the processed operating data at a plurality of sampling times.
The processed meteorological data are obtained by removing abnormal meteorological data from the initial meteorological data obtained by periodic sampling and normalizing the abnormal meteorological data.
Specifically, the initial meteorological data obtained by periodic sampling includes meteorological data acquired by a plurality of sampling periods, for example, a set of meteorological data is acquired every 30 minutes, 48 sets of meteorological data are acquired in one day, and the 48 sets of meteorological data are the initial meteorological data in this step (note that the initial meteorological data are data types of time series, and therefore the correlation degree of the meteorological data in adjacent time periods before and after is very high), it can be understood that the initial meteorological data may include abnormal meteorological data, for example, the 48 sets of meteorological data may include 8 sets of abnormal meteorological data, that is, the abnormal meteorological data are acquired at 6 sampling times, the abnormal meteorological data need to be removed from the initial meteorological data, and then the remaining meteorological data are normalized, so that the processed meteorological data at the plurality of sampling times can be obtained, for example, assuming that 48 groups of meteorological data correspond to sampling times from t1 to t48 and 8 groups of abnormal meteorological data correspond to sampling times from t5, t14 to t16, t26, t30, t41 and t42, the processed meteorological data at the sampling times obtained in the step are processed meteorological data at sampling times from t1 to t4, t6 to t13, t17 to t25, t26 to t29, t31 to t40 and t43 to t 48.
Optionally, the process of determining abnormal weather data from the initial weather data may include: firstly, primary screening is carried out on initial meteorological data to screen normal working condition data, and optionally, the primary screening refers to data cleaning on the initial meteorological data to clean invalid data; and then continuously filtering and mining the screened normal working condition data to ensure the usability of the data, namely, thoroughly checking the screened normal working condition data to find error values, abnormal values, blank values, repetition values and the like, specifically, setting a threshold range according to the normal meteorological data, detecting characteristics which are obviously different from the normal meteorological data from the primarily screened normal working condition data based on the set threshold range, and detecting missing data and repeated data from the primarily screened normal working condition data by searching for a null value (NA), wherein all the detected error and missing data are the abnormal meteorological data.
In this step, the processed operation data is obtained by preprocessing the initial operation data, the preprocessing is used for marking abnormal operation data in the initial operation data, and normalizing the unmarked operation data, and the initial operation data is the photovoltaic system operation data collected at the sampling moment corresponding to the initial meteorological data from which the abnormal meteorological data are removed.
Specifically, taking the time in the above example as an example, the sampling times corresponding to the initial meteorological data after the abnormal meteorological data is removed, that is, the sampling times t1 to t4, t6 to t13, t17 to t25, t26 to t29, t31 to t40, and t43 to t48, in this step, the photovoltaic system operation data acquired at these sampling times can be regarded as the initial operation data. It can be understood that the initial operating data includes multiple types of data, and the multiple types of data collected at each sampling time may be abnormal, so that the abnormal operating data in the initial operating data needs to be marked, so that prediction can be performed based on the unmarked operating data only when photovoltaic power is predicted in the subsequent process. When the abnormal allowable data is marked, the unmarked running data can be normalized to obtain processed running data.
Optionally, the process of determining the abnormal operation data from the initial operation data in this step is the same as the process of determining the abnormal weather data from the initial weather data, which can be described in detail with reference to the foregoing embodiment and is not described herein again.
In this step, the initial meteorological data and the unmarked operation data from which the abnormal meteorological data are removed can be normalized by the following formula:
in the formula, x norm Is normalized data, x is initial meteorological data after abnormal meteorological data are removed, or x is unmarked running data min Is the minimum of x, x max Is the maximum value in x.
It should be noted that the normalization in this step is performed to eliminate the influence of different dimensions/logarithm value ranges between indexes (i.e., the initial meteorological data after the abnormal meteorological data are removed, and the unmarked operating data), and to solve the comparability between data indexes by performing the normalization process. After the raw data are subjected to data standardization processing, all indexes are in the same order of magnitude.
Optionally, in this step, the processed meteorological data includes one or more of the following meteorological data: in-plane solar irradiance (i.e., in-plane solar altitude G) 1 ) Wind direction W a Wind speed W s And the ambient temperature T amb The processed operational data includes one or more of the following operational data: maximum power point current I mp Voltage V mp And power P mp Output value and module temperature of the output of the photovoltaic array, output value on the ac side of the inverter, and sun position parameter (i.e. sun azimuth angle)And elevation angle alpha), of course, the processed meteorological data and the processed operational data may be other, and the application is not limited thereto.
In an optional embodiment, considering that the influence of in-plane solar irradiance on photovoltaic power is large, and the solar irradiance difference at different sampling moments is large, in order to improve the accuracy of a predicted value of the photovoltaic power, an irradiance filter may be applied to classify processed meteorological data at a plurality of sampling moments to obtain processed meteorological data at a low irradiance and processed meteorological data at a high irradiance, and correspondingly, processed running data at a plurality of sampling moments are also correspondingly divided into processed running data at a low irradiance and processed running data at a high irradiance. For example, the irradiance filter includes a low-pass filter for filtering out data with irradiance greater than 600 watts per square meter, i.e., maintaining low and high irradiance conditions, and a high-pass filter for filtering out data with irradiance less than or equal to 600 watts per square meter, i.e., maintaining high irradiance conditions. Based on this, the processed meteorological data and the processed operation data at the plurality of sampling moments obtained in this step include: processed meteorological data and processed operating data at low irradiance, and processed meteorological data and processed operating data at high irradiance.
It should be noted that the "multiple sampling times" in this step generally refers to recent sampling times, for example, sampling times corresponding to the past day or days, the past month, and the like, and based on the processed meteorological data and the processed operation data at the recent sampling times, the photovoltaic power in the future set time is predicted, and the accuracy of the prediction result is relatively higher.
And S102, performing set prediction processing on the processed meteorological data and the processed operation data at a plurality of sampling moments to obtain a target hidden state corresponding to the real predicted value of the photovoltaic power.
Optionally, in this step, the processed meteorological data and the processed operating data at multiple sampling moments are subjected to setting prediction processing through a GRU-ordinary differential-bayesian updating network, so as to obtain a target hidden state capable of representing a real predicted value of the photovoltaic power. Based on the above, the target hidden state includes information of the processed meteorological data and the processed operation data at a plurality of sampling moments.
Optionally, the processed meteorological data and the processed operation data at the multiple sampling moments obtained in the previous step include: the step can respectively set and predict the processed meteorological data and the processed running data under the low irradiance, and the processed meteorological data and the processed running data under the high irradiance during the processing of the meteorological data and the processed running data under the low irradiance and the processed running data under the high irradiance.
And S103, performing time integration on the target hidden state to obtain a photovoltaic power predicted value within a set time.
Specifically, the predicted value of the photovoltaic power within the set time can be obtained by predicting the time integral of the target hidden state. The target hidden state is a hidden state corresponding to the real predicted value of the photovoltaic power, so that the predicted value of the photovoltaic power obtained based on the target hidden state is more accurate.
According to the photovoltaic power prediction method, firstly, processed meteorological data and processed operating data at a plurality of sampling moments are obtained, then, the processed meteorological data and the processed operating data at the plurality of sampling moments are set and predicted to obtain a target hidden state corresponding to a real predicted value of photovoltaic power, and finally, time integration is carried out on the target hidden state to obtain a predicted value of the photovoltaic power within set time. The method comprises the steps that the processed meteorological data at the multiple sampling moments are obtained by removing abnormal meteorological data from initial meteorological data obtained by periodic sampling and normalizing the initial meteorological data, so that the processed meteorological data at the multiple sampling moments are sporadic data and do not have periodicity any more.
In an embodiment of the present application, a process of "step S102, performing a setting prediction process on the processed meteorological data and the processed operation data at multiple sampling times to obtain a target hidden state corresponding to a true predicted value of the photovoltaic power" in the above embodiment is described.
Optionally, the process of "step S102, performing the setting and prediction processing on the processed meteorological data and the processed operating data at the multiple sampling moments to obtain the target hidden state corresponding to the real predicted value of the photovoltaic power" may include:
s1, traversing the processed meteorological data and the processed operation data at each sampling moment according to the sequence of the sampling moments:
determining a first hidden state corresponding to the current traversal according to the processed meteorological data and the processed operation data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, wherein the first hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a theoretical prediction value of photovoltaic power obtained based on the processed meteorological data under the current traversal and the previous traversal, and if the current traversal is the first traversal, the second hidden state corresponding to the previous traversal is a preset initial hidden state;
updating a first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling moment to obtain a second hidden state corresponding to the current traversal, wherein the second hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a real predicted value of the photovoltaic power obtained based on the current traversal and the processed operation data updated in each previous traversal.
For example, the plurality of sampling times include t 1-t 4, t 6-t 13, t 17-t 25, t 26-t 29, t 31-t 40, and t 43-t 48, then the step first traverses the processed meteorological data and the processed operating data at the sampling time of t1, determines a first hidden state corresponding to the current traversal (i.e., the first hidden state corresponding to the sampling time of t 1) according to the processed meteorological data and the preset initial hidden state at the sampling time of t1, and updates the first hidden state corresponding to the sampling time of t1 according to the processed operating data at the sampling time of t1 to obtain a second hidden state corresponding to the current traversal (i.e., the second hidden state corresponding to the sampling time of t 1); then, traversing the processed meteorological data and the processed running data at the sampling time of t2, determining a first hidden state corresponding to the current traversal (namely, a first hidden state corresponding to the sampling time of t 2) according to the processed meteorological data at the sampling time of t2 and a second hidden state corresponding to the sampling time of t1, and updating the first hidden state corresponding to the sampling time of t2 according to the processed running data at the sampling time of t2 to obtain a second hidden state corresponding to the current traversal (namely, a second hidden state corresponding to the sampling time of t 2); then, traversing processed meteorological data and processed operation data … at the sampling time of t3 until the processed meteorological data and the processed operation data at the sampling time of t48 are traversed, and obtaining a second hidden state corresponding to the sampling time of t 48.
In the process, the first hidden state corresponding to the current traversal is used for representing the hidden state corresponding to the theoretical predicted value of the photovoltaic power obtained based on the current traversal and the processed meteorological data in each previous traversal, the second hidden state corresponding to the current traversal is used for representing the hidden state corresponding to the real predicted value of the photovoltaic power obtained after updating the processed operation data based on the current traversal and the previous traversals, that is, the processed meteorological data can be used for predicting the hidden state (namely, the first hidden state) corresponding to the theoretical predicted value of the photovoltaic power, the processed operation data is directly related to the photovoltaic power, so that the hidden state (namely, the first hidden state) corresponding to the theoretical predicted value of the photovoltaic power is updated based on the processed operation data, so that the second hidden state is closer to the hidden state corresponding to the real predicted value of the photovoltaic power. It can be understood that, in a certain range, as the number of traversal times increases, the first hidden state more approaches to the hidden state corresponding to the theoretical predicted value of the photovoltaic power, and the second hidden state more approaches to the hidden state corresponding to the real predicted value of the photovoltaic power.
In the embodiment of the application, the hidden variable has a special property according to the updating formula, namely the hidden variable always changes between [ -1,1], namely the gradient direction always points to 0, if the preset initial hidden state starts out of the [ -1,1] region, the negative feedback can rapidly push the subsequent first hidden state and the second hidden state into the [ -1,1] region, so that the system becomes more stable.
Optionally, for the processed meteorological data and the processed operating data at each traversed sampling time, the set prediction processing in the photovoltaic power prediction model provided by the present application may include two processing stages, where the first processing stage is a GRU-constant stage, that is, a stage of "determining a first hidden state corresponding to a current traversal according to the processed meteorological data at the current traversed sampling time and a second hidden state corresponding to a previous traversal", and the stage may propagate the hidden state of the system (i.e., the first hidden state) between the processed meteorological data in time; the second processing stage is a GRU-bayes stage, that is, a stage of updating the first hidden state corresponding to the current traversal based on the processed operation data at the sampling time of the current traversal to obtain the second hidden state corresponding to the current traversal, and the stage can update the first hidden state propagated by the first stage so as to enable the second hidden state to approach the hidden state corresponding to the real predicted value of the photovoltaic power.
Optionally, the process of the first processing stage may include:
and S11, determining an updating gating and an updating vector corresponding to the current traversal according to the processed meteorological data at the current traversed sampling moment and the second hidden state corresponding to the previous traversal.
In the step, the updating gating and the updating vector corresponding to the current traversal can be determined by adopting a classic GRU network, namely, the step can input the processed meteorological data at the current traversed sampling moment and the second hidden state corresponding to the previous traversal into the classic GRU network so as to determine the updating gating and the updating vector corresponding to the current traversal.
Here, the formulas for reset gating, update gating, and update vectors in the forward propagation formula of the classical GRU network are as follows:
r t =σ(W r x t +U r h t-1 +b r );
z t =σ(W z x t +U z h t-1 +b z );
g t =tanh(W h x t +U h (r t ⊙h t-1 )+b h );
in the formula, r t Mean reset gating, z t Means updating the gating, g t Refers to updating the vector.
The updating gating and updating vectors in the step are respectively based on the processed meteorological data x at the current traversed sampling moment t Second hidden state h corresponding to previous traversal t-1 Determined update gate z t And update vector g t 。
And S12, obtaining a first hidden state corresponding to the current traversal based on the second hidden state corresponding to the previous traversal and the update gating and update vector corresponding to the current traversal.
The hidden state formula (first hidden state) of the classical GRU network is:
h t =z t ⊙h t-1 +(1-z t )⊙g t 。
this step may employ an ordinary differential network to determine the first hidden state corresponding to the current traversal. The first hidden state can be expressed as h t =GRU(h t-1 ,x t ) Then in the predetermined ordinary differential network, h can be subtracted by the formula of the first hidden state t-1 And eliminate (1-z) t ) The following differential equation of adjacent intervals is obtained:
when the time step is infinitely reduced and the number of hidden layers is infinitely increased, the left side of the differential equation can be regarded as derivation on t, so that the following ordinary differential equation is obtained:
in the ordinary differential network, for continuous observation values x (t), they can naturally input x (t) into GRU-ordinary differential, if the collected data point is failed or the collected system communication is failed, if the continuous input signal cannot be provided (i.e. continuous observation values x (t)), nothing is taken as x (t) input, and g (t) and z (t) depend on h (t) only. And, given an initial value, the solution of the ordinary differential equation is a definite function h (t) which represents the sequence of the hidden state changing with time, if the numerical solution of the ordinary differential equation at the moment can be obtained, the forward propagation of the GRU network is completed, and the change of the first hidden state h (t) finally represents the forward propagation result. Thus, the neural network becomes an ordinary differential power system, and the training and prediction of the system can be reduced to the solving problem of the ordinary differential.
In this step, the second hidden state corresponding to the previous traversal, the update gate corresponding to the current traversal, and the update vector are input to the ordinary differential network, so that the first hidden state corresponding to the current traversal can be determined based on the ordinary differential networkHidden state h t . Since the calculation process of the ordinary differential network is prior art, it will not be described in detail here.
Optionally, the process of the second processing stage may include:
and S13, acquiring an identification matrix corresponding to the processed running data at the currently traversed sampling time, wherein each element of the identification matrix is respectively used for indicating whether a component contained in the processed running data at the currently traversed sampling time is marked abnormal running data.
Specifically, when the "preprocessing" mentioned in step S101 is used to mark abnormal operation data in the initial operation data, an identification matrix may be correspondingly generated, where each element of the identification matrix is used to indicate whether a component included in the processed operation data (which is a vector) at the currently traversed sampling time is the marked abnormal operation data, that is, the identification matrix is used to specify which variables have observed values at the currently traversed sampling time.
For example, the processed operation data at the currently traversed sampling time contains 8 components, and the 8 components respectively correspond to: maximum power point current I mp Voltage V mp And power P mp Output value and module temperature of the output of the photovoltaic array, output value on the ac side of the inverter, and solar azimuthAnd the elevation angle alpha, the identification matrix corresponding to the processed operation data is a matrix with 1 x 8 dimensions, namely the identification matrix contains 8 elements in total. It can be understood that, at the currently traversed sampling time, marked abnormal operation data may exist in the 8 components, and optionally, in the identification matrix, an element corresponding to the abnormal operation data is 0, and an element corresponding to the normal operation data is 1.
And S14, for each component contained in the processed operation data at the currently traversed sampling time, determining a sparse component corresponding to the component according to the component, the element of the identification matrix corresponding to the component and the first hidden state corresponding to the current traversal, so as to obtain sparse operation data corresponding to the processed operation data at the currently traversed sampling time.
Specifically, due to the characteristic that the processed running data is extremely sparse, the processed running data at each sampling time is not completely observed and vector sparse, that is, the component at each sampling time does not have a value, so that the processed running data at the currently traversed sampling time needs to be sparse processed by the step to obtain the sparse running data.
Optionally, the sparse processing may be performed on each component included in the processed running data at the currently traversed sampling time, that is, for each component included in the processed running data at the currently traversed sampling time, the sparse component corresponding to the component is determined according to the component, the element of the identification matrix corresponding to the component, and the first hidden state corresponding to the current traversal. And performing sparse processing on each component to obtain sparse operation data corresponding to the processed operation data at the current traversed sampling moment.
For convenience of the following description, the thinning process may be denoted as f prep Then, the step performs sparse processing on each component, which is denoted as f prep (y[k],m[k],h(t - ) Wherein y [ k ]]Refers to the kth component, m [ k ], contained in the processed operating data]Refers to the element of the identification matrix corresponding to the k-th component, h (t) - ) The current traversal corresponds to a first hidden state.
And S15, obtaining a second hidden state corresponding to the current traversal according to the sparse running data corresponding to the processed running data at the current traversed sampling time and the first hidden state corresponding to the current traversal.
In this step, the first hidden state corresponding to the current traversal may be updated according to sparse running data corresponding to the processed running data at the current traversed sampling time, so as to obtain a second hidden state corresponding to the current traversal. In the foregoing, it has been explained that the first hidden state corresponding to the current traversal is used to represent the hidden state corresponding to the theoretical predicted value of the photovoltaic power obtained based on the current traversal and the processed meteorological data in each previous traversal, and then the hidden state corresponding to the theoretical predicted value of the photovoltaic power may be corrected based on the sparse post-operation data corresponding to the processed post-operation data at the sampling time of the current traversal, so that the obtained second hidden state corresponding to the current traversal is closer to the hidden state corresponding to the true predicted value.
Optionally, in this step, another GRU network may be used to determine the second hidden state corresponding to the current traversal, and then this step may be denoted as h (t) + )=GRU(h(t - ),f prep (y[k],m[k],h(t - ) H (t))) in which h (t) + ) The current traversal corresponds to the second hidden state.
For example, referring to fig. 2, the first hidden state corresponding to the current traversal obtained based on the first processing stage (i.e., the GRU-constant stage) is Loss pre Namely the hidden state corresponding to the theoretical predicted value of the photovoltaic power, and the first hidden state Loss is processed based on the second processing stage (namely the GRU-Bayes stage) pre Updating to obtain a second hidden state Loss corresponding to the current traversal of the hidden state closer to the hidden state corresponding to the real predicted value post It can be seen that the alternation between GRU-ordinary differential and GRU-Bayes results in an ordinary differential with jumps that occur at the observation point (e.g., at the sampling time t [ k ])]、t[k+1]) When there is a new observation, the step discretely updates the first hidden state, so that the complex dynamics of the hiding process can be learned.
And S2, taking the second hidden state corresponding to the last traversal as a target hidden state.
For example, in the present step, the second hidden state corresponding to the t48 sampling time in the above-mentioned example of S1 may be taken as the target hidden state, and since the target hidden state is obtained by continuously cycling through a process of "predicting the first hidden state based on the processed meteorological data and the second hidden state corresponding to the previous traversal, and correcting the first hidden state to the second hidden state corresponding to the current traversal based on the processed operating data", the target hidden state may be regarded as the hidden state corresponding to the true predicted value of the photovoltaic power.
It should be noted that, in the above application, the specific implementation of photovoltaic power prediction of the photovoltaic power prediction model is described by taking the processed meteorological data and the processed operating data at a plurality of sampling times corresponding to one day as an example, and in the actual prediction process, the photovoltaic power prediction may be performed based on the processed meteorological data and the processed operating data at a plurality of sampling times corresponding to a plurality of days, a month, or even a plurality of months, which is not limited in this application.
The embodiment of the present application further provides a photovoltaic power prediction apparatus, which is described below, and the photovoltaic power prediction apparatus described below and the photovoltaic power prediction method described above may be referred to in a corresponding manner.
Referring to fig. 3, a schematic structural diagram of a photovoltaic power prediction apparatus provided in an embodiment of the present application is shown, and as shown in fig. 3, the photovoltaic power prediction apparatus may include: the device comprises a data acquisition module 301, a setting prediction processing module 302 and a photovoltaic power prediction value determination module 303.
The data acquisition module 301 is configured to acquire processed meteorological data and processed operational data at multiple sampling moments, where the processed meteorological data is obtained by performing abnormal meteorological data rejection and normalization on initial meteorological data obtained by periodic sampling, the processed operational data is obtained by preprocessing the initial operational data, the preprocessing is used to mark abnormal operational data in the initial operational data and normalize unmarked operational data, and the initial operational data is photovoltaic system operational data acquired at sampling moments corresponding to the initial meteorological data from which the abnormal meteorological data are rejected.
The setting prediction processing module 302 is configured to perform setting prediction processing on the processed meteorological data and the processed operating data at multiple sampling moments to obtain a target hidden state corresponding to a real predicted value of the photovoltaic power.
And the photovoltaic power predicted value determining module 303 is configured to perform time integration on the target hidden state to obtain a photovoltaic power predicted value within a set time.
According to the photovoltaic power prediction device, firstly, processed meteorological data and processed operation data at a plurality of sampling moments are obtained, then, the processed meteorological data and the processed operation data at the plurality of sampling moments are set and predicted, a target hidden state corresponding to a real predicted value of photovoltaic power is obtained, and finally, time integration is carried out on the target hidden state, so that a photovoltaic power predicted value within set time is obtained. The method comprises the steps that the processed meteorological data at the multiple sampling moments are obtained by removing abnormal meteorological data from initial meteorological data obtained by periodic sampling and normalizing the initial meteorological data, so that the processed meteorological data at the multiple sampling moments are sporadic data and do not have periodicity any more.
In a possible implementation manner, the setting prediction processing module 302 may include: the device comprises a traversing module and a target hidden state determining module.
The traversal module is used for traversing the processed meteorological data and the processed operation data at each sampling moment according to the sequence of the sampling moments: determining a first hidden state corresponding to the current traversal according to the processed meteorological data and the processed operation data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, wherein the first hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a theoretical prediction value of photovoltaic power obtained based on the processed meteorological data under the current traversal and the previous traversal, and if the current traversal is the first traversal, the second hidden state corresponding to the previous traversal is a preset initial hidden state; updating a first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling moment to obtain a second hidden state corresponding to the current traversal, wherein the second hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a real predicted value of the photovoltaic power obtained based on the current traversal and the processed operation data updated in each previous traversal.
And the target hidden state determining module is used for taking the second hidden state corresponding to the last traversal as the target hidden state.
In a possible implementation manner, when the traversing module determines the first hidden state corresponding to the current traversal according to the processed meteorological data at the current traversed sampling time and the second hidden state corresponding to the previous traversal, the traversing module may include: the system comprises a first GRU network computing module and an ordinary differential network computing module.
The first GRU network computing module is used for determining an update gate control and an update vector corresponding to the current traversal according to the processed meteorological data at the current traversed sampling moment and a second hidden state corresponding to the previous traversal.
And the ordinary differential network computing module is used for obtaining a first hidden state corresponding to the current traversal based on the second hidden state corresponding to the previous traversal and the updating gating and updating vector corresponding to the current traversal.
In a possible implementation manner, the foregoing traversal module, based on the processed operation data at the currently traversed sampling time, updates the first hidden state corresponding to the current traversal, and when obtaining the second hidden state corresponding to the current traversal, may include: the device comprises an identification matrix acquisition module, a sparse processing module and a second GRU network calculation module.
The identification matrix acquisition module is used for acquiring an identification matrix corresponding to the processed running data at the currently traversed sampling time, wherein each element of the identification matrix is respectively used for indicating whether a component contained in the processed running data at the currently traversed sampling time is marked abnormal running data.
And the sparse processing module is used for determining a sparse component corresponding to each component contained in the processed running data at the current traversed sampling moment according to the component, the element of the identification matrix corresponding to the component and the first hidden state corresponding to the current traversal, so as to obtain sparse running data corresponding to the processed running data at the current traversed sampling moment.
And the second GRU network computing module is used for obtaining a second hidden state corresponding to the current traversal according to the sparse running data corresponding to the processed running data at the current traversed sampling time and the first hidden state corresponding to the current traversal.
In one possible implementation, the processed weather data in the data acquiring module 301 may include one or more of the following weather data: solar irradiance, wind direction, wind speed, and ambient temperature in-plane.
The processed operation data in the data obtaining module 301 may include one or more of the following operation data: maximum power point current, voltage and power, output value and module temperature of the output end of the photovoltaic array, inverter alternating side output value, and solar position parameters.
In a possible implementation manner, the processed meteorological data and the processed operation data at multiple sampling times in the data obtaining module 301 may include: processed meteorological data and processed operational data at low irradiance, and processed meteorological data and processed operational data at high irradiance.
Based on this, the setting prediction processing module 302 may be specifically configured to perform the setting prediction processing on the processed meteorological data and the processed operating data under the low irradiance, and the processed meteorological data and the processed operating data under the high irradiance, respectively.
The embodiment of the application also provides photovoltaic power prediction equipment. Alternatively, fig. 4 shows a block diagram of a hardware structure of the photovoltaic power prediction device, and referring to fig. 4, the hardware structure of the photovoltaic power prediction device may include: at least one processor 401, at least one communication interface 402, at least one memory 403 and at least one communication bus 404;
in the embodiment of the present application, the number of the processor 401, the communication interface 402, the memory 403 and the communication bus 404 is at least one, and the processor 401, the communication interface 402 and the memory 403 complete communication with each other through the communication bus 404;
the memory 403 may include a high-speed RAM memory, and may further include a non-volatile memory (non-volatile memory) or the like, such as at least one disk memory;
wherein the memory 403 stores a program and the processor 401 may call the program stored in the memory 403 for:
acquiring processed meteorological data and processed operation data at a plurality of sampling moments, wherein the processed meteorological data are obtained by performing abnormal meteorological data elimination and normalization on initial meteorological data obtained by periodic sampling, the processed operation data are obtained by performing pretreatment on the initial operation data, the pretreatment is used for marking abnormal operation data in the initial operation data and normalizing unmarked operation data, and the initial operation data are photovoltaic system operation data acquired at the sampling moments corresponding to the initial meteorological data from which the abnormal meteorological data are eliminated;
setting and predicting the processed meteorological data and the processed operation data at a plurality of sampling moments to obtain a target hidden state corresponding to a real predicted value of the photovoltaic power;
and performing time integration on the target hidden state to obtain a photovoltaic power predicted value within set time.
Alternatively, the detailed function and the extended function of the program may be as described above.
Embodiments of the present application further provide a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method for predicting photovoltaic power as described above is implemented.
Alternatively, the detailed function and the extended function of the program may be as described above.
Finally, it is further noted that, herein, relational terms such as second and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for predicting photovoltaic power, comprising:
acquiring processed meteorological data and processed operation data at a plurality of sampling moments, wherein the processed meteorological data are obtained by removing abnormal meteorological data from initial meteorological data obtained by periodic sampling and normalizing the initial meteorological data, the processed operation data are obtained by preprocessing the initial operation data, the preprocessing is used for marking the abnormal operation data in the initial operation data and normalizing the unmarked operation data, and the initial operation data are photovoltaic system operation data acquired at the sampling moments corresponding to the initial meteorological data from which the abnormal meteorological data are removed;
setting and predicting the processed meteorological data and the processed operation data at the plurality of sampling moments to obtain a target hidden state corresponding to the real predicted value of the photovoltaic power;
and performing time integration on the target hidden state to obtain a photovoltaic power predicted value within a set time.
2. The photovoltaic power prediction method according to claim 1, wherein the step of performing the setting prediction processing on the processed meteorological data and the processed operation data at the plurality of sampling moments to obtain the target hidden state corresponding to the real predicted value of the photovoltaic power includes:
traversing the processed meteorological data and the processed operation data at each sampling moment according to the sequence of the sampling moments:
determining a first hidden state corresponding to the current traversal according to the processed meteorological data and the processed operation data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, wherein the first hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a theoretical prediction value of photovoltaic power obtained based on the processed meteorological data under the current traversal and the previous traversal, and if the current traversal is the first traversal, the second hidden state corresponding to the previous traversal is a preset initial hidden state;
updating a first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling time to obtain a second hidden state corresponding to the current traversal, wherein the second hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a real predicted value of the photovoltaic power obtained after the processed operation data under the current traversal and forward traversal are updated;
and taking the second hidden state corresponding to the last traversal as the target hidden state.
3. The method for predicting photovoltaic power according to claim 2, wherein the determining a first hidden state corresponding to a current traversal according to the processed meteorological data at the current traversed sampling time and a second hidden state corresponding to a previous traversal comprises:
determining an updating gate control and an updating vector corresponding to the current traversal according to the processed meteorological data at the current traversed sampling moment and a second hidden state corresponding to the previous traversal;
and obtaining a first hidden state corresponding to the current traversal based on the second hidden state corresponding to the previous traversal and the updating gating and updating vector corresponding to the current traversal.
4. The photovoltaic power prediction method according to claim 2, wherein the updating the first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling time to obtain the second hidden state corresponding to the current traversal comprises:
acquiring an identification matrix corresponding to the processed running data at the currently traversed sampling time, wherein each element of the identification matrix is respectively used for indicating whether a component contained in the processed running data at the currently traversed sampling time is the marked abnormal running data;
for each component contained in the processed running data at the currently traversed sampling time, determining a sparse component corresponding to the component according to the component, the element of the identification matrix corresponding to the component and a first hidden state corresponding to the current traversal, so as to obtain sparse running data corresponding to the processed running data at the currently traversed sampling time;
and obtaining a second hidden state corresponding to the current traversal according to sparse running data corresponding to the processed running data at the current traversed sampling time and the first hidden state corresponding to the current traversal.
5. The photovoltaic power prediction method of claim 1, wherein the processed meteorological data comprises one or more of the following meteorological data: solar irradiance, wind direction, wind speed and ambient temperature in the plane;
the processed operational data includes one or more of the following operational data: maximum power point current, voltage and power, output value and module temperature of the output end of the photovoltaic array, inverter alternating side output value, and solar position parameters.
6. The photovoltaic power prediction method of claim 5, wherein the processed meteorological data and processed operational data at the plurality of sampling instants comprises: processed meteorological data and processed operational data at low irradiance, and processed meteorological data and processed operational data at high irradiance;
the setting and predicting processing of the processed meteorological data and the processed operating data at the plurality of sampling moments comprises the following steps:
and respectively carrying out the setting prediction processing on the processed meteorological data and the processed operation data under the low irradiance, and the processed meteorological data and the processed operation data under the high irradiance.
7. A photovoltaic power prediction apparatus, comprising: the device comprises a data acquisition module, a setting prediction processing module and a photovoltaic power prediction value determination module;
the data acquisition module is used for acquiring processed meteorological data and processed operation data at a plurality of sampling moments, wherein the processed meteorological data are obtained by performing abnormal meteorological data elimination and normalization on initial meteorological data obtained by periodic sampling, the processed operation data are obtained by preprocessing the initial operation data, the preprocessing is used for marking abnormal operation data in the initial operation data and normalizing unmarked operation data, and the initial operation data are photovoltaic system operation data acquired at the sampling moments corresponding to the initial meteorological data from which the abnormal meteorological data are eliminated;
the setting prediction processing module is used for performing setting prediction processing on the processed meteorological data and the processed operation data at the plurality of sampling moments to obtain a target hidden state corresponding to a real predicted value of the photovoltaic power;
and the photovoltaic power predicted value determining module is used for performing time integration on the target hidden state to obtain a photovoltaic power predicted value within set time.
8. The photovoltaic power prediction device of claim 7, wherein the setting prediction processing module comprises: the device comprises a traversing module and a target hidden state determining module;
the traversal module is used for traversing the processed meteorological data and the processed operation data at each sampling moment according to the sequence of the sampling moments: determining a first hidden state corresponding to the current traversal according to the processed meteorological data and the processed operation data at the current traversed sampling time and a second hidden state corresponding to the previous traversal, wherein the first hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a theoretical prediction value of photovoltaic power obtained based on the processed meteorological data under the current traversal and the previous traversal, and if the current traversal is the first traversal, the second hidden state corresponding to the previous traversal is a preset initial hidden state; updating a first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling moment to obtain a second hidden state corresponding to the current traversal, wherein the second hidden state corresponding to the current traversal is used for representing a hidden state corresponding to a real predicted value of the photovoltaic power obtained based on the current traversal and the updating of the processed operation data at each previous traversal;
and the target hidden state determining module is used for taking a second hidden state corresponding to the last traversal as the target hidden state.
9. The photovoltaic power prediction device of claim 8, wherein the traversing module, when determining the first hidden state corresponding to the current traversal according to the processed meteorological data at the current traversed sampling time and the second hidden state corresponding to the previous traversal, comprises: the system comprises a first GRU network computing module and an ordinary differential network computing module;
the first GRU network computing module is used for determining an update gate control and an update vector corresponding to the current traversal according to the processed meteorological data at the current traversed sampling moment and a second hidden state corresponding to the previous traversal;
and the ordinary differential network computing module is used for obtaining a first hidden state corresponding to the current traversal based on the second hidden state corresponding to the previous traversal and the update gating and update vector corresponding to the current traversal.
10. The photovoltaic power prediction device according to claim 8, wherein the traversing module updates the first hidden state corresponding to the current traversal based on the processed operation data at the current traversed sampling time, and obtains the second hidden state corresponding to the current traversal, and the method includes: the device comprises an identification matrix acquisition module, a sparse processing module and a second GRU network calculation module;
the identification matrix obtaining module is configured to obtain an identification matrix corresponding to the processed running data at the currently traversed sampling time, where each element of the identification matrix is respectively used to indicate whether a component included in the processed running data at the currently traversed sampling time is the marked abnormal running data;
the sparse processing module is configured to determine, for each component included in the processed running data at the currently traversed sampling time, a sparse component corresponding to the component according to the component, the element of the identifier matrix corresponding to the component, and the first hidden state corresponding to the current traversal, so as to obtain sparse running data corresponding to the processed running data at the currently traversed sampling time;
and the second GRU network computing module is used for obtaining a second hidden state corresponding to the current traversal according to sparse running data corresponding to the processed running data at the current traversed sampling time and the first hidden state corresponding to the current traversal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210665580.0A CN114971062A (en) | 2022-06-14 | 2022-06-14 | Photovoltaic power prediction method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210665580.0A CN114971062A (en) | 2022-06-14 | 2022-06-14 | Photovoltaic power prediction method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114971062A true CN114971062A (en) | 2022-08-30 |
Family
ID=82960702
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210665580.0A Pending CN114971062A (en) | 2022-06-14 | 2022-06-14 | Photovoltaic power prediction method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114971062A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116957223A (en) * | 2023-06-14 | 2023-10-27 | 华能澜沧江水电股份有限公司 | Photovoltaic power prediction method and device based on illumination image |
-
2022
- 2022-06-14 CN CN202210665580.0A patent/CN114971062A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116957223A (en) * | 2023-06-14 | 2023-10-27 | 华能澜沧江水电股份有限公司 | Photovoltaic power prediction method and device based on illumination image |
CN116957223B (en) * | 2023-06-14 | 2024-03-08 | 华能澜沧江水电股份有限公司 | Photovoltaic power prediction method and device based on illumination image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110610280B (en) | Short-term prediction method, model, device and system for power load | |
CN109873610B (en) | Photovoltaic array fault diagnosis method based on IV characteristic and depth residual error network | |
CN109146162B (en) | A kind of probability wind speed forecasting method based on integrated Recognition with Recurrent Neural Network | |
Baptista et al. | Hardware implementation of an artificial neural network model to predict the energy production of a photovoltaic system | |
Al-Rousan et al. | Assessment of predicting hourly global solar radiation in Jordan based on Rules, Trees, Meta, Lazy and Function prediction methods | |
CN116707331B (en) | Inverter output voltage high-precision adjusting method and system based on model prediction | |
CN116739172B (en) | Method and device for ultra-short-term prediction of offshore wind power based on climbing identification | |
Girimurugan et al. | Application of deep learning to the prediction of solar irradiance through missing data | |
Lungu et al. | PREDICTION INTELLIGENT SYSTEM IN THE FIELD OF RENEWABLE ENERGIES THROUGH NEURAL NETWORKS. | |
Van Gompel et al. | Cost-effective fault diagnosis of nearby photovoltaic systems using graph neural networks | |
CN112200038A (en) | CNN-based rapid identification method for oscillation type of power system | |
Lan et al. | Remaining useful life estimation of turbofan engine using LSTM neural networks | |
CN116845868A (en) | Distributed photovoltaic power prediction method and device, electronic equipment and storage medium | |
CN114971062A (en) | Photovoltaic power prediction method and device | |
CN116663746A (en) | Power load prediction method and device, computer equipment and storage medium | |
CN110222840B (en) | Cluster resource prediction method and device based on attention mechanism | |
CN115564155A (en) | Distributed wind turbine generator power prediction method and related equipment | |
CN115759365A (en) | Photovoltaic power generation power prediction method and related equipment | |
Wibawa et al. | Long Short-Term Memory to Predict Unique Visitors of an Electronic Journal | |
CN117596191A (en) | Power Internet of things abnormality detection method, device and storage medium | |
CN108875960A (en) | A kind of learning method and system of the timing ambiguity Cognitive Map based on gradient decline | |
CN117669656A (en) | TCN-Semi PN-based direct-current micro-grid stability real-time monitoring method and device | |
Alam | Recurrent neural networks in electricity load forecasting | |
Sun et al. | Rotor Angle Stability Prediction using Temporal and Topological Embedding Deep Neural Network Based on Grid-Informed Adjacency Matrix | |
CN115796338A (en) | Photovoltaic power generation power prediction model construction and photovoltaic power generation power prediction method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |