CN111612648A - Training method and device of photovoltaic power generation prediction model and computer equipment - Google Patents
Training method and device of photovoltaic power generation prediction model and computer equipment Download PDFInfo
- Publication number
- CN111612648A CN111612648A CN202010425653.XA CN202010425653A CN111612648A CN 111612648 A CN111612648 A CN 111612648A CN 202010425653 A CN202010425653 A CN 202010425653A CN 111612648 A CN111612648 A CN 111612648A
- Authority
- CN
- China
- Prior art keywords
- photovoltaic
- power generation
- data
- generation prediction
- prediction model
- 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.)
- Granted
Links
- 238000010248 power generation Methods 0.000 title claims abstract description 194
- 238000012549 training Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000009499 grossing Methods 0.000 claims abstract description 68
- 238000005070 sampling Methods 0.000 claims abstract description 19
- 210000002569 neuron Anatomy 0.000 claims description 66
- 230000002068 genetic effect Effects 0.000 claims description 21
- 238000004590 computer program Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 description 11
- 238000013528 artificial neural network Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 230000035772 mutation Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 3
- 238000005457 optimization Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 210000000349 chromosome Anatomy 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000000126 substance Substances 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- 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
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- 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)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- General Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Physiology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Educational Administration (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a training method and device of a photovoltaic power generation prediction model, computer equipment and a storage medium. The method comprises the following steps: acquiring photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has corresponding sampling time points; smoothing the similar solar photovoltaic data, and extracting photovoltaic data trend characteristics of the similar days; training an initial photovoltaic power generation prediction model according to the trend characteristics of the photovoltaic data of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected. By adopting the method, the output power prediction accuracy of the photovoltaic array can be improved.
Description
Technical Field
The present application relates to the field of power technologies, and in particular, to a training method and apparatus for determining a model of photovoltaic power, a computer device, and a storage medium.
Background
Photovoltaic power generation is an important component in renewable energy, and in recent years, distributed photovoltaic power stations are developed, particularly photovoltaic flat price is gradually realized on the internet, and the photovoltaic power generation has stronger application characteristics.
After the distributed photovoltaic power station is connected to the grid in a large scale, the safety and the stability of the power grid are greatly influenced. In order to reduce the impact of the distributed photovoltaic power station on the stability of the power grid, the power grid resources need to be reasonably scheduled. Therefore, the accurate prediction of the output power of the photovoltaic array is realized, and the method has important effects on power dispatching and evaluation of the operation condition of the photovoltaic array.
In the prior art, one often predicts the photovoltaic array output power directly based on data provided by a weather forecast system. However, the data provided by the weather forecast system often has some interference and errors due to many uncertainty factors, which causes the prior art to have a large prediction error when predicting the output power of the photovoltaic array.
Therefore, the output power prediction method of the current photovoltaic array has the problem of low accuracy.
Disclosure of Invention
In view of the above, it is necessary to provide a training method, an apparatus, a computer device and a storage medium for a photovoltaic power generation prediction model, which can improve the output power prediction accuracy of a photovoltaic array.
A method of training a photovoltaic power generation prediction model, the method comprising:
acquiring photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has corresponding sampling time points;
smoothing the similar solar photovoltaic data, and extracting photovoltaic data trend characteristics of the similar days;
training an initial photovoltaic power generation prediction model according to the trend characteristics of the photovoltaic data of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected.
In one embodiment, the smoothing of the similar solar photovoltaic data to extract the photovoltaic data trend feature of the similar day includes:
acquiring photovoltaic data observation time sequences of the similar days according to the photovoltaic data of the similar days;
substituting the photovoltaic data observation time sequence into a preset weighted moving average model to determine a time sequence prediction equation of the similar day;
determining the photovoltaic data smoothing time sequence of the similar days according to the time sequence prediction equation; the photovoltaic data smooth time series are used for representing photovoltaic data trend characteristics of the similar days.
In one embodiment, the photovoltaic data includes meteorological data and power data, and the training of the initial photovoltaic power generation prediction model according to the photovoltaic data trend characteristics of the similar days to obtain the optimized photovoltaic power generation prediction model includes:
obtaining a model training sample; the model training sample comprises a meteorological data smoothing time sequence of the similar day and a power data smoothing time sequence of the similar day;
training the initial photovoltaic power generation prediction model based on the meteorological data smoothing time sequence and the power data smoothing time sequence;
and when the trained initial photovoltaic power generation prediction model meets the preset training condition, obtaining the optimized photovoltaic power generation prediction model.
In one embodiment, the photovoltaic power generation prediction model includes a plurality of neuron nodes, and the training of the initial photovoltaic power generation prediction model based on the meteorological data smoothing time series and the power data smoothing time series includes:
inputting the meteorological data smooth time sequence into the initial photovoltaic power generation prediction model, and determining a node prediction value corresponding to each neuron node;
based on an error reverse propagation algorithm, according to the power data smooth time sequence, determining a node expected value corresponding to each neuron node;
and optimizing the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node.
In one embodiment, the optimizing the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node includes:
acquiring a parameter population corresponding to the model parameters of the initial photovoltaic power generation prediction model based on a genetic algorithm; the parameter population comprises a plurality of parameter individuals; each individual parameter corresponds to a set of model parameters;
determining the fitness corresponding to each parameter individual according to the difference between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node;
performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain optimized parameter individuals;
obtaining the trained initial photovoltaic power generation prediction model according to the model parameters corresponding to the optimized parameter individuals;
retraining the trained initial photovoltaic power generation prediction model until the trained initial photovoltaic power generation prediction model meets the preset training condition.
In one embodiment, the performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain optimized parameter individuals includes:
selecting the parameter individuals according to the fitness corresponding to each parameter individual to obtain eliminated parameter individuals;
performing cross operation on the eliminated parameter individuals according to the model parameters corresponding to the eliminated parameter individuals to obtain crossed parameter individuals;
and performing variation operation on the crossed parameter individuals according to model parameters corresponding to the crossed parameter individuals to obtain the optimized parameter individuals.
In one embodiment, the method further comprises:
obtaining a difference value between a node predicted value corresponding to the neuron node and a node expected value corresponding to the neuron node;
and when the difference value is smaller than the preset difference threshold value, judging that the trained initial photovoltaic power generation prediction model meets the preset training condition.
A training apparatus for a photovoltaic power generation prediction model, the apparatus comprising:
the acquisition module is used for acquiring photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has corresponding sampling time points;
the preprocessing module is used for smoothing the similar solar photovoltaic data and extracting photovoltaic data trend characteristics of the similar days;
the training module is used for training the initial photovoltaic power generation prediction model according to the photovoltaic data trend characteristics of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method described above when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method.
According to the training method, the training device, the computer equipment and the storage medium of the photovoltaic power generation prediction model, the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be tested is sampled at intervals to obtain the photovoltaic data of the similar day, the photovoltaic data of the similar day is subjected to smoothing processing to filter interference data and noise data in the photovoltaic data of the similar day, so that the trend characteristic of the photovoltaic data of the similar day is accurately extracted, the initial photovoltaic power generation prediction model is trained by adopting the photovoltaic data trend characteristic of the similar day, the photovoltaic power generation prediction result of the photovoltaic array on the day to be tested can be accurately output by the optimized photovoltaic power generation prediction model obtained through training, and the output power prediction precision of the photovoltaic array is improved.
Drawings
FIG. 1 is a diagram of an application environment of a method for training a photovoltaic power generation prediction model according to an embodiment;
FIG. 2 is a schematic flow chart illustrating a method for training a photovoltaic power generation prediction model according to an embodiment;
FIG. 3 is a schematic flow chart illustrating a method for training a photovoltaic power generation prediction model according to another embodiment;
FIG. 4 is a block diagram of an exemplary training apparatus for a photovoltaic power generation prediction model;
FIG. 5 is a flow chart illustrating a method for training a photovoltaic power generation prediction model according to one embodiment;
FIG. 6 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The training method of the photovoltaic power generation prediction model can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The server 110 first obtains photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has a corresponding sampling time point. Then, the server 110 performs smoothing processing on the photovoltaic data of the similar day, and extracts the trend feature of the photovoltaic data of the similar day. Finally, the server 110 trains an initial photovoltaic power generation prediction model according to the photovoltaic data trend characteristics of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected. In practical applications, the server 110 may be implemented by a stand-alone server or a server cluster composed of a plurality of servers.
In an embodiment, as shown in fig. 2, a method for training a photovoltaic power generation prediction model is provided, which is described by taking the method as an example for being applied to the terminal in fig. 1, and includes the following steps:
step S210, acquiring photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; similar day photovoltaic data has corresponding sampling time points.
The photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals.
Wherein, the photovoltaic array can be the connection of a plurality of photovoltaic modules.
Wherein, the photovoltaic data may refer to meteorological data and power data of the photovoltaic array on similar days.
Wherein, meteorological data may refer to irradiance data and ambient temperature data, ambient humidity data, wind data, etc. for similar days. In practical applications, the irradiance data may refer to coplanar irradiation data measured by an irradiator having a same installation angle as the inclination angle of the photovoltaic cell panel.
Of course, the photovoltaic data may also include current data and voltage data of the photovoltaic array.
In a specific implementation, the server 110 obtains photovoltaic data in similar days according to a certain time interval to sample the photovoltaic data in the similar days, so as to obtain the photovoltaic data in the similar days.
And S220, smoothing the similar solar photovoltaic data, and extracting photovoltaic data trend characteristics of similar days.
The photovoltaic data trend characteristic may refer to a characteristic corresponding to a change trend of the photovoltaic data.
In a specific implementation, after the server 110 obtains the similar-day photovoltaic data, the server 110 may perform smoothing on the similar-day photovoltaic data according to a preset data smoothing algorithm, filter interference data and noise data in the similar-day photovoltaic data, and obtain the smoothed similar-day photovoltaic data. Finally, the server 110 determines the trend feature of the photovoltaic data corresponding to the similar day according to the smoothed photovoltaic data of the similar day.
Step S230, training the initial photovoltaic power generation prediction model according to the trend characteristics of the photovoltaic data of the similar days to obtain an optimized photovoltaic power generation prediction model; and the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected.
The photovoltaic power generation prediction result can refer to the power generation power of the photovoltaic array in a certain period of time of a day to be detected, and can also refer to the power generation amount of the photovoltaic array in a certain period of time of the day to be detected.
In a specific implementation, after the server 110 extracts and determines the photovoltaic data trend feature corresponding to the similar day, the server 110 may train the initial photovoltaic power generation prediction model by using the photovoltaic data trend feature to obtain an optimized photovoltaic power generation prediction model for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be measured.
Specifically, the server 110 may use the smoothed photovoltaic data of the similar day as the photovoltaic data trend feature corresponding to the similar day. Wherein the smoothed similar solar volt data may include smoothed similar diurnal meteorological data and smoothed similar diurnal power data. Then, the server 110 may construct an initial photovoltaic power generation prediction model, take the smoothed similar solar weather data as training sample characteristics, take the smoothed similar solar power data as training sample labels, train the initial photovoltaic power generation prediction model, and further obtain an optimized photovoltaic power generation prediction model that may be used to output a photovoltaic power generation prediction result of the photovoltaic array on the day to be measured.
In practical applications, the server 110 may input weather data of a day to be measured, such as coplanar irradiation data of the day to be measured, environmental temperature data of the day to be measured, and the like, to the optimized photovoltaic power generation prediction model, and output a power generation power of the day to be measured within a certain period of time of the day to be measured, that is, a short-term power prediction result, through the optimized photovoltaic power generation prediction model.
According to the training method of the photovoltaic power generation prediction model, the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be tested is sampled at intervals to obtain the photovoltaic data of the similar day, the photovoltaic data of the similar day is subjected to smoothing processing to filter interference data and noise data in the photovoltaic data of the similar day, so that the trend characteristic of the photovoltaic data of the similar day is accurately extracted, the initial photovoltaic power generation prediction model is trained by adopting the trend characteristic of the photovoltaic data of the similar day, the photovoltaic power generation prediction result of the photovoltaic array on the day to be tested can be accurately output by the optimized photovoltaic power generation prediction model obtained through training, and the output power prediction precision of the photovoltaic array is improved.
In another embodiment, smoothing the similar solar photovoltaic data to extract photovoltaic data trend characteristics of similar days includes: acquiring photovoltaic data observation time sequences of similar days according to the photovoltaic data of the similar days; substituting the photovoltaic data observation time sequence into a preset weighted moving average model to determine a time sequence prediction equation of a similar day; determining a photovoltaic data smoothing time sequence of similar days according to a time sequence prediction equation; the photovoltaic data smoothing time series are used for representing photovoltaic data trend characteristics of similar days.
The weighted moving average model may be a mathematical model that performs a weighted moving average method on the data. In practical applications, the weighted moving average method may be an exponential smoothing method based on time series.
In a specific implementation, the server 110 obtains photovoltaic data observation time sequences of similar days according to the photovoltaic data of the similar days. Then, the server 110 substitutes the photovoltaic data observation time series into a preset weighted moving average model to realize moving smoothing processing on the photovoltaic data observation time series by adopting an exponential smoothing method, and determines a time series prediction equation of a similar day.
Specifically, the server 110 may set the photovoltaic data observation time sequence to { y }tTaking the number of terms of the moving average as n, the predicted value of the t +1 th stage can be expressed as:
wherein, ytIs the t-th actual value, Mt (1)Represents the tth primary slip mean;
the predicted value is t +1 (t is more than or equal to n); n is { ytThe number of original data contained.
The server 110 may then determine the tth quadratic moving average Mt (2)Wherein the t-th quadratic moving average Mt (2)Can be expressed as:
then, the server 110 sets a time series ytCan be expressed as a linear model of time t, i.e.
yt＝a+bt+c
Wherein c is a random term, which can be omitted, then the t-th primary slip average value Mt (1)Can be expressed as:
thus, the server 110 can obtain the tth quadratic moving average Mt (2)Can be expressed as:
thus, server 110 calculates:
wherein the content of the first and second substances,
and
can be written as:
in summary, the server 110 determines that the time series prediction equation of the similar day is;
finally, the server 110 determines a photovoltaic data smoothing time sequence for representing photovoltaic data trend characteristics of the similar day according to the time sequence prediction equation, so that the photovoltaic data of the similar day is quickly and accurately smoothed, interference data and noise data in the photovoltaic data of the similar day are filtered, and the smoothed photovoltaic data of the similar day is obtained.
In another embodiment, the photovoltaic data includes meteorological data and power data, and the training of the initial photovoltaic power generation prediction model is performed according to photovoltaic data trend characteristics of similar days to obtain an optimized photovoltaic power generation prediction model, including: obtaining a model training sample; the model training sample comprises a meteorological data smoothing time sequence on a similar day and a power data smoothing time sequence on the similar day; training an initial photovoltaic power generation prediction model based on the meteorological data smoothing time sequence and the power data smoothing time sequence; and when the trained initial photovoltaic power generation prediction model meets the preset training condition, obtaining an optimized photovoltaic power generation prediction model.
Wherein the photovoltaic data comprises meteorological data and power data. In practical applications, when the photovoltaic data includes meteorological data and power data, the photovoltaic data smoothing time series includes meteorological data smoothing time series and power data smoothing time series.
Wherein the meteorological data comprises coplanar irradiation data and ambient temperature data. In practical applications, when the meteorological data includes coplanar irradiation data and ambient temperature data, the meteorological data smoothing time series includes irradiation data smoothing time series and temperature data smoothing time series.
The model training sample comprises a meteorological data smoothing time sequence on a similar day and a power data smoothing time sequence on a similar day. It should be noted that, the meteorological data smooth time series can be used as the sample characteristics of the model training sample; the smoothed time series of power data may be used as a sample label for the model training samples.
In the specific implementation, in the process that the server 110 trains the initial photovoltaic power generation prediction model according to the trend characteristics of the photovoltaic data of the similar day to obtain the optimized photovoltaic power generation prediction model, the specific implementation may include: the server 110 first obtains the meteorological data smoothing time series as the sample features of the model training samples, and obtains the power data smoothing time series as the sample labels of the model training samples. The server 110 will then train the initial photovoltaic power generation prediction model based on the meteorological data smoothed time series and the power data smoothed time series. Specifically, the server 110 may input the meteorological data smoothed time series to the initial photovoltaic power generation prediction model, and determine the photovoltaic power generation prediction result corresponding to the meteorological data smoothed time series through processing of the initial photovoltaic power generation prediction model. Then, the server 110 determines a photovoltaic power generation expected result corresponding to the power data smoothing time series according to the power data smoothing time series. Finally, the server 110 optimizes the model parameters of the initial photovoltaic power generation prediction model according to the difference between the photovoltaic power generation prediction result and the photovoltaic power generation expected result. In practical application, the server 110 determines the model loss of the initial photovoltaic power generation prediction model according to the difference between the photovoltaic power generation prediction result and the photovoltaic power generation expected result, and optimizes the model parameters of the initial photovoltaic power generation prediction model based on a preset model optimization algorithm so as to train the initial photovoltaic power generation prediction model. When the server 110 judges that the trained initial photovoltaic power generation prediction model meets the preset training condition, an optimized photovoltaic power generation prediction model is obtained. In practical applications, the model optimization algorithm may be at least one of a gradient descent algorithm and a genetic algorithm.
According to the technical scheme, in the process of training the initial photovoltaic power generation prediction model according to the trend characteristics of the photovoltaic data of the similar days to obtain the optimized photovoltaic power generation prediction model, the initial photovoltaic power generation prediction model is trained by adopting the meteorological data smoothing time sequence and the power data smoothing time sequence of the similar days, so that the optimized photovoltaic power generation prediction model obtained by training can accurately output the photovoltaic power generation prediction result of the photovoltaic array on the day to be detected, and the output power prediction precision of the photovoltaic array is improved.
In another embodiment, the photovoltaic power generation prediction model includes a plurality of neuron nodes, and the training of the initial photovoltaic power generation prediction model based on the meteorological data smoothing time series and the power data smoothing time series includes: inputting the meteorological data smooth time sequence into an initial photovoltaic power generation prediction model, and determining a node prediction value corresponding to each neuron node; based on an error reverse propagation algorithm, according to the power data smooth time sequence, determining a node expected value corresponding to each neuron node; and optimizing the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node.
The photovoltaic power generation prediction model comprises a plurality of neuron nodes.
In a specific implementation, the training of the initial photovoltaic power generation prediction model by the server 110 based on the meteorological data smoothing time series and the power data smoothing time series may specifically include: the server 110 may input the meteorological data smooth time series to the initial photovoltaic power generation prediction model, and further obtain a node prediction value corresponding to each neuron node. Then, the server 110 determines a node expectation value corresponding to each neuron node according to the power data smoothing time series based on an error back propagation algorithm.
Specifically, the server 110 may determine a loss of the photovoltaic power generation prediction model, i.e., a prediction error, based on the prediction value of the photovoltaic power generation prediction model output layer and the power data smoothing time series. Then, the server 110 determines the expected node value corresponding to each neuron node according to the loss of the photovoltaic power generation prediction model based on an error back propagation algorithm (back propagation algorithm).
Finally, the server 110 optimizes the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node. In practical applications, the server 110 may optimize model parameters of the initial photovoltaic power generation prediction model based on a difference between a node prediction value corresponding to the neuron node and a node expected value corresponding to the neuron node by using a genetic algorithm.
According to the technical scheme, in the process of training the initial photovoltaic power generation prediction model, the node prediction values corresponding to the neuron nodes and the node expected values corresponding to the neuron nodes are determined, and model parameters of the initial photovoltaic power generation prediction model are accurately optimized based on the difference between the node prediction values corresponding to the neuron nodes and the node expected values corresponding to the neuron nodes, so that the optimized photovoltaic power generation prediction model obtained through training can accurately output the photovoltaic power generation prediction result of the photovoltaic array on the day to be measured, and the output power prediction precision of the photovoltaic array is improved.
In another embodiment, optimizing the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node prediction value corresponding to the neuron node and the node expectation value corresponding to the neuron node includes: acquiring a parameter population corresponding to model parameters of the initial photovoltaic power generation prediction model based on a genetic algorithm; the parameter population comprises a plurality of parameter individuals; each individual parameter corresponds to a set of model parameters; according to the difference between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node, determining the fitness corresponding to each parameter individual; performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain optimized parameter individuals; obtaining a trained initial photovoltaic power generation prediction model according to the model parameters corresponding to the optimized parameter individuals; and retraining the trained initial photovoltaic power generation prediction model until the trained initial photovoltaic power generation prediction model meets the preset training condition.
Wherein the parameter population comprises a plurality of parameter individuals. Each individual parameter corresponds to a set of model parameters.
Wherein the genetic operator operation may comprise at least one of a selection operation, a crossover operation, and a mutation operation.
In a specific implementation, the server 110 may specifically include, in the process of optimizing the model parameter of the initial photovoltaic power generation prediction model according to a difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node: the server 110 obtains a parameter population corresponding to the model parameters of the initial photovoltaic power generation prediction model based on a genetic algorithm. Then, the server 110 determines the fitness corresponding to each parameter individual according to the difference between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node based on a preset fitness function F.
The fitness function can distinguish the quality of an individual and is used for natural selection. There are many choices for the fitness function, and generally the smaller the function value, the higher the fitness, and generally the better the individual.
In practical applications, the fitness function F may be:
where k is a coefficient, which can be selected according to the actual situation, fiIs the predicted output of the ith node, oiIs the expected output of the ith node, and n is the number of output nodes.
After the server 110 determines the fitness corresponding to each parameter individual, the server 110 may perform genetic operator operations such as selection operation, crossover operation, mutation operation and the like on the parameter individual according to the fitness corresponding to each parameter individual to obtain an optimized parameter individual; then, the server 110 obtains the trained initial photovoltaic power generation prediction model according to the model parameters corresponding to the optimized parameter individuals. Specifically, the server 110 may decode the optimized parameter population formed by the optimized parameter individuals to obtain decoded model parameters. Then, the server 110 uses the decoded model parameters as model parameters corresponding to the trained initial photovoltaic power generation prediction model.
Finally, the server 110 retrains the trained initial photovoltaic power generation prediction model until the trained initial photovoltaic power generation prediction model meets the preset training conditions.
According to the technical scheme, the parameter population corresponding to the model parameters of the initial photovoltaic power generation prediction model is obtained based on a genetic algorithm, and the fitness corresponding to each parameter individual in the parameter population is determined according to the difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node, so that genetic operation can be performed on each parameter individual based on the fitness corresponding to each parameter individual, and further, the model parameters of the initial photovoltaic power generation prediction model can be accurately and quickly optimized.
In another embodiment, performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain optimized parameter individuals, including: selecting the parameter individuals according to the fitness corresponding to each parameter individual to obtain eliminated parameter individuals; performing cross operation on the eliminated parameter individuals according to model parameters corresponding to the eliminated parameter individuals to obtain crossed parameter individuals; and performing variation operation on the crossed parameter individuals according to the model parameters corresponding to the crossed parameter individuals to obtain optimized parameter individuals.
In a specific implementation, in the process that the server 110 performs the genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain the optimized parameter individuals, the server 110 may perform the selection operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain the eliminated parameter individuals. In particular, the selection operation is a process of optimizing from the old population. The probability that the individual parameter is selected is related to the fitness function, and the higher the fitness function value is, the higher the probability of being selected is. Probability p of the ith individual being selectediCan be expressed as:
wherein, FiIs the fitness value of the ith parameter individual, and N is the number of the parameter individuals.
Then, the server 110 performs the model parameter adjustment on the eliminated parameter individuals according to the model parameters corresponding to the eliminated parameter individualsAnd performing cross operation to obtain crossed parameter individuals. Specifically, the crossover operation is to randomly select two individuals from the eliminated parameter individuals, inherit excellent characteristics to offspring through crossover combination, and generate excellent individuals, i.e. the l, m chromosomes al，amThe intersection at j bits can be expressed as:
wherein b is a random value of 0 to 1.
Finally, the server 110 performs variation operation on the crossed parameter individuals according to the model parameters corresponding to the crossed parameter individuals, so as to obtain optimized parameter individuals. Specifically, a preset variation formula can be adopted to perform variation operation on the crossed parameter individuals, so as to obtain optimized parameter individuals.
Wherein, the jth gene a of the ith individualijThe variation formula is as follows:
f(g)＝r2(1-g/Gmax)2
wherein, amax，aminAre respectively aijR is a random number between 0 and 1; r is2Is an arbitrary number, G is the number of iterations, GmaxThe maximum number of evolutions;
according to the technical scheme of the embodiment, in the process of obtaining the optimized parameter individuals by performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual, the optimal model parameters of the photovoltaic power generation prediction model can be rapidly determined by sequentially performing genetic operations such as selection operation, cross operation and mutation operation on the parameter individuals, so that the training of the initial photovoltaic power generation prediction model is rapidly completed.
In another embodiment, the method further comprises: obtaining a difference value between a node predicted value corresponding to the neuron node and a node expected value corresponding to the neuron node; and when the difference value is smaller than a preset difference threshold value, judging that the trained initial photovoltaic power generation prediction model meets a preset training condition.
In a specific implementation, after the server 110 obtains the node predicted values corresponding to the plurality of neuron nodes in the photovoltaic power generation prediction model and the node expected values corresponding to the plurality of neuron nodes, the server 110 may obtain a difference value between the node predicted values corresponding to the neuron nodes and the node expected values corresponding to the neuron nodes. Then, the server 110 determines whether the difference value is smaller than a preset difference threshold value; when the difference value is smaller than the preset difference threshold value, the server 110 determines that the trained initial photovoltaic power generation prediction model meets the preset training condition.
According to the technical scheme, the difference value between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node is obtained, and the trained initial photovoltaic power generation prediction model is timely and accurately judged to meet the preset training condition through the difference value.
In another embodiment, as shown in fig. 3, a method for training a photovoltaic power generation prediction model is provided, which is described by taking the method as an example for being applied to the terminal in fig. 1, and includes the following steps: step S310, photovoltaic data of similar days are obtained; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has a corresponding sampling time point. And S320, acquiring the photovoltaic data observation time sequence of the similar day according to the photovoltaic data of the similar day. And step S330, substituting the photovoltaic data observation time sequence into a preset weighted moving average model, and determining a time sequence prediction equation of the similar day. Step S340, determining a photovoltaic data smoothing time sequence of the similar day according to the time sequence prediction equation; the photovoltaic data smoothing time sequence is used for representing photovoltaic data trend characteristics of the similar days; the photovoltaic data includes meteorological data and power data. Step S350, obtaining a model training sample; the model training samples comprise meteorological data smoothing time series of the similar days and power data smoothing time series of the similar days. And S360, training the initial photovoltaic power generation prediction model based on the meteorological data smooth time sequence and the power data smooth time sequence. Step S370, when the trained initial photovoltaic power generation prediction model meets preset training conditions, obtaining the optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected. The specific definition of the steps can be referred to the above specific definition of the training method of the photovoltaic power generation prediction model.
It should be understood that although the steps in the flowcharts of fig. 2 and 3 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2 and 3 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternately with other steps or at least some of the other steps.
In one embodiment, as shown in fig. 4, there is provided a training apparatus for a photovoltaic power generation prediction model, the apparatus including:
an obtaining module 410, configured to obtain photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has corresponding sampling time points;
the preprocessing module 420 is configured to perform smoothing processing on the similar solar photovoltaic data, and extract trend characteristics of the similar solar photovoltaic data;
the training module 430 is used for training the initial photovoltaic power generation prediction model according to the photovoltaic data trend characteristics of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected.
In one embodiment, the preprocessing module 420 is specifically configured to obtain, according to the similar day photovoltaic data, a photovoltaic data observation time sequence of the similar day; substituting the photovoltaic data observation time sequence into a preset weighted moving average model to determine a time sequence prediction equation of the similar day; determining the photovoltaic data smoothing time sequence of the similar days according to the time sequence prediction equation; the photovoltaic data smooth time series are used for representing photovoltaic data trend characteristics of the similar days.
In one embodiment, the photovoltaic data includes meteorological data and power data, and the training module 430 is specifically configured to obtain a model training sample; the model training sample comprises a meteorological data smoothing time sequence of the similar day and a power data smoothing time sequence of the similar day; training the initial photovoltaic power generation prediction model based on the meteorological data smoothing time sequence and the power data smoothing time sequence; and when the trained initial photovoltaic power generation prediction model meets the preset training condition, obtaining the optimized photovoltaic power generation prediction model.
In one embodiment, the photovoltaic power generation prediction model includes a plurality of neuron nodes, and the training module 430 is specifically configured to input the meteorological data smooth time series to the initial photovoltaic power generation prediction model, and determine a node prediction value corresponding to each neuron node; based on an error reverse propagation algorithm, according to the power data smooth time sequence, determining a node expected value corresponding to each neuron node; and optimizing the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node.
In one embodiment, the training module 430 is specifically configured to obtain a parameter population corresponding to model parameters of the initial photovoltaic power generation prediction model based on a genetic algorithm; the parameter population comprises a plurality of parameter individuals; each individual parameter corresponds to a set of model parameters; determining the fitness corresponding to each parameter individual according to the difference between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node; performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain optimized parameter individuals; obtaining the trained initial photovoltaic power generation prediction model according to the model parameters corresponding to the optimized parameter individuals; retraining the trained initial photovoltaic power generation prediction model until the trained initial photovoltaic power generation prediction model meets the preset training condition.
In one embodiment, the training module 430 is specifically configured to perform a selection operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain eliminated parameter individuals; performing cross operation on the eliminated parameter individuals according to the model parameters corresponding to the eliminated parameter individuals to obtain crossed parameter individuals; and performing variation operation on the crossed parameter individuals according to model parameters corresponding to the crossed parameter individuals to obtain the optimized parameter individuals.
In one embodiment, the training device for the photovoltaic power generation prediction model further includes:
a difference value obtaining module, configured to obtain a difference value between a node predicted value corresponding to the neuron node and a node expected value corresponding to the neuron node;
and the judging module is used for judging that the trained initial photovoltaic power generation prediction model meets the preset training condition when the difference value is smaller than the preset difference threshold value.
For specific definition of the training device for the photovoltaic power generation prediction model, reference may be made to the above definition of the training method for the photovoltaic power generation prediction model, and details are not repeated here. All or part of each module in the training device of the photovoltaic power generation prediction model can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
To facilitate understanding by those skilled in the art, fig. 5 provides a training flow diagram of a training method of a photovoltaic power generation prediction model; as shown in fig. 5, the server 110 acquires similar day photovoltaic data input by the user; the server 110 then smoothes the similar solar photovoltaic data, training the initial BP neural network. Wherein the initial BP neural network has an initial weight and a threshold. Specifically, in the process of training the initial BP neural network, the server 110 may obtain a training test error in the training process, and then calculate the fitness of each parameter individual according to the test error; and carrying out selection operation, cross operation and mutation operation on each parameter individual so as to obtain a new parameter population. Then, the server 110 adjusts the BP neural network based on the new parameter population until a preset termination condition is met, and specifically, the server 110 may decode the new parameter population to obtain an optimized model parameter of the BP neural network, so as to obtain the optimized BP neural network, and further, the optimized BP neural network is used for determining the short-term power of the photovoltaic array on the day to be measured.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 6. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing training data of the photovoltaic power generation prediction model. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of training a photovoltaic power generation prediction model.
Those skilled in the art will appreciate that the architecture shown in fig. 6 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In an embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of a method of training a photovoltaic power generation prediction model as described above. Here, the steps of a training method of a photovoltaic power generation prediction model may be steps in a training method of a photovoltaic power generation prediction model according to the above embodiments.
In one embodiment, a computer-readable storage medium is provided, in which a computer program is stored, which, when being executed by a processor, causes the processor to carry out the steps of the above-mentioned training method of a photovoltaic power generation prediction model. Here, the steps of a training method of a photovoltaic power generation prediction model may be steps in a training method of a photovoltaic power generation prediction model according to the above embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A method for training a photovoltaic power generation prediction model, the method comprising:
acquiring photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has corresponding sampling time points;
smoothing the similar solar photovoltaic data, and extracting photovoltaic data trend characteristics of the similar days;
training an initial photovoltaic power generation prediction model according to the trend characteristics of the photovoltaic data of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected.
2. The method according to claim 1, wherein the smoothing of the similar solar photovoltaic data to extract photovoltaic data trend features of the similar days comprises:
acquiring photovoltaic data observation time sequences of the similar days according to the photovoltaic data of the similar days;
substituting the photovoltaic data observation time sequence into a preset weighted moving average model to determine a time sequence prediction equation of the similar day;
determining the photovoltaic data smoothing time sequence of the similar days according to the time sequence prediction equation; the photovoltaic data smooth time series are used for representing photovoltaic data trend characteristics of the similar days.
3. The method of claim 2, wherein the photovoltaic data comprises meteorological data and power data, and the training of the initial photovoltaic power generation prediction model to obtain the optimized photovoltaic power generation prediction model according to the photovoltaic data trend characteristics of the similar days comprises:
obtaining a model training sample; the model training sample comprises a meteorological data smoothing time sequence of the similar day and a power data smoothing time sequence of the similar day;
training the initial photovoltaic power generation prediction model based on the meteorological data smoothing time sequence and the power data smoothing time sequence;
and when the trained initial photovoltaic power generation prediction model meets the preset training condition, obtaining the optimized photovoltaic power generation prediction model.
4. The method of claim 1, wherein the photovoltaic power generation prediction model comprises a plurality of neuron nodes, and wherein training the initial photovoltaic power generation prediction model based on the meteorological data smoothing time series and the power data smoothing time series comprises:
inputting the meteorological data smooth time sequence into the initial photovoltaic power generation prediction model, and determining a node prediction value corresponding to each neuron node;
based on an error reverse propagation algorithm, according to the power data smooth time sequence, determining a node expected value corresponding to each neuron node;
and optimizing the model parameters of the initial photovoltaic power generation prediction model according to the difference between the node prediction value corresponding to the neuron node and the node expected value corresponding to the neuron node.
5. The method of claim 4, wherein optimizing the model parameters of the initial photovoltaic power generation prediction model based on the difference between the predicted values of the nodes corresponding to the neuron nodes and the expected values of the nodes corresponding to the neuron nodes comprises:
acquiring a parameter population corresponding to the model parameters of the initial photovoltaic power generation prediction model based on a genetic algorithm; the parameter population comprises a plurality of parameter individuals; each individual parameter corresponds to a set of model parameters;
determining the fitness corresponding to each parameter individual according to the difference between the node predicted value corresponding to the neuron node and the node expected value corresponding to the neuron node;
performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain optimized parameter individuals;
obtaining the trained initial photovoltaic power generation prediction model according to the model parameters corresponding to the optimized parameter individuals;
retraining the trained initial photovoltaic power generation prediction model until the trained initial photovoltaic power generation prediction model meets the preset training condition.
6. The method according to claim 5, wherein the performing genetic operator operation on the parameter individuals according to the fitness corresponding to each parameter individual to obtain the optimized parameter individual comprises:
selecting the parameter individuals according to the fitness corresponding to each parameter individual to obtain eliminated parameter individuals;
performing cross operation on the eliminated parameter individuals according to the model parameters corresponding to the eliminated parameter individuals to obtain crossed parameter individuals;
and performing variation operation on the crossed parameter individuals according to model parameters corresponding to the crossed parameter individuals to obtain the optimized parameter individuals.
7. The method of claim 4, further comprising:
obtaining a difference value between a node predicted value corresponding to the neuron node and a node expected value corresponding to the neuron node;
and when the difference value is smaller than the preset difference threshold value, judging that the trained initial photovoltaic power generation prediction model meets the preset training condition.
8. An apparatus for training a photovoltaic power generation prediction model, the apparatus comprising:
the acquisition module is used for acquiring photovoltaic data of similar days; the photovoltaic data of the similar day is obtained by sampling the photovoltaic data of the photovoltaic array on the similar day corresponding to the day to be detected at intervals; the similar day photovoltaic data has corresponding sampling time points;
the preprocessing module is used for smoothing the similar solar photovoltaic data and extracting photovoltaic data trend characteristics of the similar days;
the training module is used for training the initial photovoltaic power generation prediction model according to the photovoltaic data trend characteristics of the similar days to obtain an optimized photovoltaic power generation prediction model; the optimized photovoltaic power generation prediction model is used for outputting a photovoltaic power generation prediction result of the photovoltaic array on the day to be detected.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010425653.XA CN111612648B (en) | 2020-05-19 | 2020-05-19 | Training method and device for photovoltaic power generation prediction model and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010425653.XA CN111612648B (en) | 2020-05-19 | 2020-05-19 | Training method and device for photovoltaic power generation prediction model and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111612648A true CN111612648A (en) | 2020-09-01 |
| CN111612648B CN111612648B (en) | 2024-01-19 |
Family
ID=72201241
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010425653.XA Active CN111612648B (en) | 2020-05-19 | 2020-05-19 | Training method and device for photovoltaic power generation prediction model and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111612648B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113705840A (en) * | 2021-09-23 | 2021-11-26 | 重庆允成互联网科技有限公司 | Equipment predictive maintenance method and device, computer equipment and storage medium |
| CN117578597A (en) * | 2024-01-19 | 2024-02-20 | 杭州利沃得电源有限公司 | Energy-saving control method and system for photovoltaic inverter system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104463356A (en) * | 2014-11-27 | 2015-03-25 | 国网浙江省电力公司嘉兴供电公司 | Photovoltaic power generation power prediction method based on multi-dimension information artificial neural network algorithm |
| CN105631558A (en) * | 2016-03-22 | 2016-06-01 | 国家电网公司 | BP neural network photovoltaic power generation system power prediction method based on similar day |
| CN108197744A (en) * | 2018-01-02 | 2018-06-22 | 华北电力大学（保定） | A kind of determining method and system of photovoltaic generation power |
| CN110121171A (en) * | 2019-05-10 | 2019-08-13 | 青岛大学 | Trust prediction technique based on exponential smoothing and gray model |
-
2020
- 2020-05-19 CN CN202010425653.XA patent/CN111612648B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104463356A (en) * | 2014-11-27 | 2015-03-25 | 国网浙江省电力公司嘉兴供电公司 | Photovoltaic power generation power prediction method based on multi-dimension information artificial neural network algorithm |
| CN105631558A (en) * | 2016-03-22 | 2016-06-01 | 国家电网公司 | BP neural network photovoltaic power generation system power prediction method based on similar day |
| CN108197744A (en) * | 2018-01-02 | 2018-06-22 | 华北电力大学（保定） | A kind of determining method and system of photovoltaic generation power |
| CN110121171A (en) * | 2019-05-10 | 2019-08-13 | 青岛大学 | Trust prediction technique based on exponential smoothing and gray model |
Non-Patent Citations (3)
| Title |
|---|
| 付宗见: "基于遗传算法优化BP神经网络的光伏阵列短期功率预测" * |
| 叶林: "基于WRF模式输出的光伏发电量预测研究" * |
| 张久菊: "基于组合模型的光伏发电功率短期预测技术研究" * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113705840A (en) * | 2021-09-23 | 2021-11-26 | 重庆允成互联网科技有限公司 | Equipment predictive maintenance method and device, computer equipment and storage medium |
| CN117578597A (en) * | 2024-01-19 | 2024-02-20 | 杭州利沃得电源有限公司 | Energy-saving control method and system for photovoltaic inverter system |
| CN117578597B (en) * | 2024-01-19 | 2024-04-05 | 杭州利沃得电源有限公司 | Energy-saving control method and system for photovoltaic inverter system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111612648B (en) | 2024-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108416695B (en) | Power load probability density prediction method, system and medium based on deep learning | |
| CN111260030B (en) | A-TCN-based power load prediction method and device, computer equipment and storage medium | |
| CN108280552B (en) | Power load prediction method and system based on deep learning and storage medium | |
| JP5888640B2 (en) | Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program | |
| CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
| CN111382929A (en) | Method and device for constructing river diatom bloom early warning model | |
| CN113326969A (en) | Short-term wind speed prediction method and system based on improved whale algorithm for optimizing ELM | |
| CN112381673B (en) | Park electricity utilization information analysis method and device based on digital twin | |
| CN111612648B (en) | Training method and device for photovoltaic power generation prediction model and computer equipment | |
| CN114065653A (en) | Construction method of power load prediction model and power load prediction method | |
| CN114676923A (en) | Method and device for predicting generated power, computer equipment and storage medium | |
| CN111753461A (en) | Tidal water level correction method, target residual water level acquisition method, device and equipment | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| CN114971090A (en) | Electric heating load prediction method, system, equipment and medium | |
| CN114897204A (en) | Method and device for predicting short-term wind speed of offshore wind farm | |
| CN116739172B (en) | Method and device for ultra-short-term prediction of offshore wind power based on climbing identification | |
| CN117557375A (en) | Transaction evaluation method and related device based on virtual power plant | |
| CN114611799B (en) | Time sequence neural network new energy output multi-step prediction method based on supervised learning | |
| CN114169416B (en) | Short-term load prediction method based on migration learning under small sample set | |
| CN113610665B (en) | Wind power generation power prediction method based on multi-delay output echo state network | |
| CN112581311B (en) | Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants | |
| CN115345343A (en) | Method and device for predicting turbidity of water supply pipe network | |
| CN113642784A (en) | Wind power ultra-short term prediction method considering fan state | |
| CN111899123A (en) | Electric quantity prediction method, electric quantity prediction device and computer readable storage medium | |
| CN115222024B (en) | Short-term photovoltaic power generation prediction method and system based on depth feature selection network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |