CN116388279B - Grid-connected control method and control system for solar photovoltaic power generation system - Google Patents
Grid-connected control method and control system for solar photovoltaic power generation system Download PDFInfo
- Publication number
- CN116388279B CN116388279B CN202310581426.XA CN202310581426A CN116388279B CN 116388279 B CN116388279 B CN 116388279B CN 202310581426 A CN202310581426 A CN 202310581426A CN 116388279 B CN116388279 B CN 116388279B
- Authority
- CN
- China
- Prior art keywords
- power
- curve
- historical
- data
- grid
- 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.)
- Active
Links
- 238000010248 power generation Methods 0.000 title claims abstract description 135
- 238000000034 method Methods 0.000 title claims abstract description 50
- 230000002787 reinforcement Effects 0.000 claims abstract description 118
- 238000003062 neural network model Methods 0.000 claims abstract description 81
- 238000012549 training Methods 0.000 claims description 94
- 238000004146 energy storage Methods 0.000 claims description 37
- 125000004122 cyclic group Chemical group 0.000 claims description 22
- 238000007726 management method Methods 0.000 claims description 9
- 238000005286 illumination Methods 0.000 claims description 8
- 238000013480 data collection Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000013500 data storage Methods 0.000 claims description 3
- 230000000306 recurrent effect Effects 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/381—Dispersed generators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/28—Arrangements for balancing of the load in a network by storage of energy
- H02J3/32—Arrangements for balancing of the load in a network by storage of energy using batteries with converting means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/22—The renewable source being solar energy
- H02J2300/24—The renewable source being solar energy of photovoltaic origin
-
- 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
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
- Y02E10/56—Power conversion systems, e.g. maximum power point trackers
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Power Engineering (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Public Health (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Water Supply & Treatment (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a grid-connected control method and a control system thereof in a solar photovoltaic power generation system, which relate to the technical field of grid-connected control, and are characterized in that a neural network model for predicting a future power generation power curve is trained by collecting photovoltaic power generation historical data and grid-connected historical data in advance, a circulating neural network model for generating a predicted power consumption curve is trained, a first deep reinforcement learning model for determining whether grid-connected is performed or not in real time is trained, a second deep reinforcement learning model for determining stored power value or released power value in real time is trained, and a decision for determining whether grid-connected decision and generating stored power value or released power value during grid-connected is generated for the photovoltaic power generation system based on the neural network model, the circulating neural network model, the first deep reinforcement learning model and the second deep reinforcement learning model; the electric energy stability in the grid connection process of the power grid is guaranteed, and the problem that grid connection cannot be achieved due to fluctuation of photovoltaic power generation is avoided.
Description
Technical Field
The invention relates to the technical field of grid-connected control, in particular to a grid-connected control method and a grid-connected control system in a solar photovoltaic power generation system.
Background
Photovoltaic power generation systems are becoming a clean, green, renewable energy source, and have been widely used in power systems worldwide. With the continuous development of photovoltaic power generation technology, the scale of the photovoltaic power generation technology in a power grid is continuously enlarged, but the photovoltaic power generation technology also brings about challenges in the aspect of power grid safety. Particularly, in the low-charge and high-charge period, the unpredictability and the variability of the photovoltaic power generation system can cause great fluctuation of the power grid, so that the safe and stable operation of the power grid is affected; therefore, in order to solve these problems, a technique for dynamically adjusting the power value of the grid connection in real time is required.
Therefore, the invention provides a grid-connected control method and a control system thereof in a solar photovoltaic power generation system.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems existing in the prior art. Therefore, the grid-connected control method and the control system thereof in the solar photovoltaic power generation system dynamically ensure the electric energy stability in the grid-connected process, and avoid the problem that grid connection cannot be realized due to the fluctuation of photovoltaic power generation.
In order to achieve the above object, according to the present invention, a grid-connected control method in a solar photovoltaic power generation system is provided, comprising the steps of:
step one: collecting photovoltaic power generation historical data and grid connection historical data in advance; the photovoltaic power generation historical data comprises historical environment data, historical power data and historical energy storage data;
step two: training a neural network model of a future power generation curve estimated according to the historical environment data and the historical power data; marking the neural network model as M1;
step three: training a cyclic neural network model according to historical power consumption data in the historical power data to generate an estimated power consumption curve for estimating future power consumption data; marking the recurrent neural network model as M2;
step four: constructing a first four-tuple set for training a first deep reinforcement learning model according to the historical power data and the historical energy storage data;
step five: training a first deep reinforcement learning model for deciding whether to perform grid connection in real time according to a future power generation power curve and an estimated power consumption curve by taking the first four-tuple set as the input of the first deep reinforcement learning model; marking the first deep reinforcement learning model as S1;
Step six: generating a residual power curve according to the historical power data;
step seven: constructing a second four-tuple set for training a second deep reinforcement learning model according to the residual power curve and grid-connected historical data;
step eight: the second four-element set is used as the input of a second deep reinforcement learning model, and the second deep reinforcement learning model of the stored power value or the released power value is determined in real time according to the residual power curve of the photovoltaic power generation system and grid connection history data when the power grid is connected is trained, so that the stability of the grid connection power is ensured; marking the second deep reinforcement learning model as S2;
step nine: based on the neural network model, the cyclic neural network model M2, the first deep reinforcement learning model and the second deep reinforcement learning model, generating a decision of whether to carry out grid connection decision of the photovoltaic power generation system and generating a stored power value or a released power value when the grid connection is carried out;
the historical environment data comprise a historical temperature curve, a historical illumination intensity curve and a historical humidity curve in a historical environment where the photovoltaic power generation system is located;
the historical power data comprise a historical power generation power curve and a power consumption power curve which are generated by the photovoltaic power generation system under a corresponding historical environment;
The historical energy storage data comprise a residual electric quantity curve of an energy storage power supply in a photovoltaic power generation system;
the grid-connected historical data comprise grid-connected power curves stored or released in real time by an energy storage power supply when the photovoltaic power generation system performs grid connection;
the method for training the neural network model of the future generation power curve estimated according to the historical environment data comprises the following steps:
taking a historical temperature curve, a historical illumination intensity curve and a historical humidity curve in the historical environment data as inputs of a multi-feature time series prediction neural network model, taking a power generation curve as a prediction curve of the multi-feature time series prediction neural network model, and training the multi-feature time series prediction neural network model;
the method for generating the estimated power consumption curve for estimating the future power consumption data comprises the following steps:
presetting a predicted time step T, a first sliding step length and a sliding window length according to actual experience; converting the historical power consumption data into a plurality of training samples by using a sliding window method, taking the training samples as input of a cyclic neural network model, taking the historical power consumption data after a prediction time step T as output, taking the subsequent historical power consumption data of each training sample as a prediction target, taking the prediction accuracy rate as a training target, and training the cyclic neural network model; generating a neural network model for predicting an estimated power consumption curve according to the historical power consumption data;
The method for constructing the first four-tuple set for training the first deep reinforcement learning model is as follows:
in the power generation power curve, the power consumption power curve and the residual electric quantity curve of the historical electric power data, taking a predicted time step length T as the duration of a sliding window, presetting a second sliding step length b according to actual experience, sliding the power generation power curve, the power consumption power curve and the residual electric quantity curve in time sequence, and generating a plurality of groups of original data combinations;
each group of the original data combination comprises a generating power curve with the duration of T, a power consumption curve with the duration of T and a residual electric quantity curve with the duration of T;
at the beginning of each set of said original data combinations:
taking a generated power curve and a consumed power curve in the original data combination as a first initial state;
taking whether the starting time of the original data combination is in the time of the subsequent prediction time step T or not for grid connection as a first action of selection;
calculating the total power generation amount according to the power generation power curve in the original data combination; calculating the total power consumption according to the power consumption curve in the original data combination; marking the total power generation amount as F, the total power consumption amount as H, the residual power of the energy storage power supply at the end time of the original data combination as C, and the maximum storage power and the minimum storage power threshold of the energy storage power supply as D1 and D2 respectively; counting the time length T1 of the residual electric quantity equal to the maximum stored electric quantity D1 and the time length T2 of the residual electric quantity lower than the minimum stored electric quantity threshold value from the residual electric quantity curve; the maximum storage electric quantity is the maximum electric quantity capacity of the energy storage power supply, and the minimum storage electric quantity threshold value is an electric quantity value which is set according to practical experience and is necessary for guaranteeing emergency power utilization; calculating a first reward value Q1 corresponding to the original data set, wherein the expression form of the first reward value Q1 is as follows: q1=a1×f-a2×h+ea3× (T1-T2) +a4×c; wherein a1, a2, a3 and a4 are respectively preset proportionality coefficients;
Combining the next set of raw data as a first next state;
then for each original data combination, < first initial state, selected first action, first prize value Q1, first next state > as its first quadruple;
the first quaternion of all the original data combinations jointly form a first quaternion set;
the method for training the first deep reinforcement learning model for deciding whether to carry out grid connection or not in real time according to the future power generation power curve and the estimated power consumption curve comprises the following steps:
taking the first four-tuple set as input of a first deep reinforcement learning model, wherein the first deep reinforcement learning model carries out training by randomly extracting a plurality of first four-tuple from the first four-tuple set, and learns whether grid connection is carried out or not under different initial states so as to obtain a strategy of a maximum rewarding value Q1;
the method for generating the surplus power curve according to the historical power data is as follows:
the method comprises the steps of subtracting real-time power consumption in a power consumption curve from real-time power generation in a power generation curve in time sequence to obtain real-time residual power, and connecting the real-time residual power in time sequence to obtain a residual power curve;
The second four-tuple set for training the second deep reinforcement learning model is constructed by the following steps:
taking the real-time surplus power of the surplus power curve as a second initial state;
a second action of selecting the real-time storage or release power of the grid-connected power curve;
the power values in the residual power curve and the grid-connected power curve at the previous moment are marked as X1 and Y1 respectively, and the power values in the residual power curve and the grid-connected power curve at the current moment are marked as X2 and Y2 respectively;
calculating the average power V of the subsequent residual power curve according to the subsequent residual power curve at the current moment;
presetting a fluctuation power threshold Z according to actual experience of grid connection, setting a second rewarding value Q2, and setting the second rewarding value Q2 as a preset maximum punishment value if the value is (X2 + Y2) - (X1 + Y1) | > Z, wherein the maximum punishment value is a negative value; if | (x2+y2) - (x1+y1) | is less than or equal to Z, the calculation formula of the second prize value Q2 is q2= -e1|y2| -e2| (x2+y2) - (x1+y1) | -e3| (x2+y2) -v|; wherein e1, e2 and e3 are respectively preset proportionality coefficients;
taking the real-time surplus power at the next moment of the surplus power curve as a second next state;
then < second initial state, selected second action, second prize value Q1, second next state > for each moment as a second quadruple;
The second four-tuple at all moments jointly form a second four-tuple set;
the training method is that when the power grid is connected, according to the residual power curve of the photovoltaic power generation system and the grid connection history data, a second deep reinforcement learning model of the stored power value or the released power value is decided in real time, wherein the second deep reinforcement learning model comprises the following steps:
taking the second four-tuple set as the input of a second deep reinforcement learning model, wherein the second deep reinforcement learning model carries out training by randomly extracting a plurality of second four-tuples from the second four-tuple set, and learns the power values stored or released by the energy storage power supply under different initial states so as to obtain a strategy of a maximum rewarding value Q2;
the mode of generating the decision whether to carry out grid connection decision or not for the photovoltaic power generation system and generating the stored power value or the released power value during grid connection is as follows:
loading the trained neural network model M1, the first deep reinforcement learning model S1 and the second deep reinforcement learning model S2 into an energy management system of the photovoltaic power generation system;
the energy management system acquires predicted environment data, power consumption data and energy storage data in real time, and acquires a future power generation power curve by using a neural network model M1 based on the predicted environment data;
Based on the power consumption data, a cyclic neural network model M2 is used for obtaining a predicted power consumption curve; based on a future power generation power curve and an estimated power consumption curve, a first deep reinforcement learning model S1 is used for generating a decision of whether to carry out grid connection or not;
after generating a grid-connected decision, the energy management system generates a predicted residual power curve according to a future power generation power curve and a predicted power consumption curve, and generates a stored power value or a released power value decision based on the residual power curve by using a second deep reinforcement learning model;
the invention provides a grid-connected control system in a solar photovoltaic power generation system, which comprises a historical data collection module, a model training module and a model application module; wherein, each module is connected by an electric and/or wireless network mode;
the historical data collection module is mainly used for collecting photovoltaic power generation historical data and grid-connected historical data in advance and sending the collected photovoltaic power generation historical data and the collected grid-connected historical data to the model training module;
the model training module comprises a neural network model training unit, a first deep reinforcement learning model training unit and a second deep reinforcement learning model training unit;
The neural network model training unit trains a neural network model of a future power generation curve estimated according to the historical environment data and the historical power data, trains a cyclic neural network model according to the historical power consumption data in the historical power data, and generates an estimated power consumption curve estimated for the future power consumption data;
the first deep reinforcement learning model training unit constructs a first four-tuple set for training a first deep reinforcement learning model according to the historical power data and the historical energy storage data;
the first four-element set is used as the input of a first deep reinforcement learning model, and the first deep reinforcement learning model for deciding whether to carry out grid connection or not is trained according to a future power generation power curve and an estimated power consumption curve in real time;
the second deep reinforcement learning model training unit generates a residual power curve according to the historical power data, constructs a second four-tuple set for training a second deep reinforcement learning model according to the residual power curve and grid-connected historical data, trains out the power value stored or the power value released in real time according to the residual power curve of the photovoltaic power generation system and the grid-connected historical data when the grid is connected by taking the second four-tuple set as the input of the second deep reinforcement learning model, so as to ensure the stability of the grid-connected power;
The model training module sends the neural network model, the cyclic neural network model, the first deep reinforcement learning model and the second deep reinforcement learning model to the model application module;
the model training module is used for deciding whether the photovoltaic power generation system is grid-connected or not based on the neural network model, the circulating neural network model M2, the first deep reinforcement learning model and the second deep reinforcement learning model, and generating a decision of a stored power value or a released power value after the grid-connected decision is generated.
Compared with the prior art, the invention has the beneficial effects that:
(1) According to the method, based on the historical data collected in advance, a neural network model of a future power generation curve estimated according to the historical environmental data, an estimated power consumption curve estimated for the future power consumption data according to the historical power consumption data in the historical power data and a first deep reinforcement learning model for deciding whether to grid-connect or not in real time according to the future power generation curve and the estimated power consumption curve are generated, the first deep reinforcement learning model can dynamically decide whether to grid-connect or not in real time according to the future power generation and power consumption conditions of the photovoltaic power generation system, flexibility of grid-connect of the photovoltaic system is improved, and utilization rate of electric quantity of the photovoltaic power generation system is guaranteed;
(2) According to the method, a second four-tuple set for training a second deep reinforcement learning model is constructed according to the residual power curve and grid-connected historical data of the power grid, and then the second deep reinforcement learning model for deciding the stored power value or the released power value in real time according to the residual power curve of the photovoltaic power generation system and the grid-connected historical data of the power grid when the power grid is connected is trained based on the second four-tuple set.
Drawings
Fig. 1 is a flowchart of a grid-connected control method in embodiment 1 of the present invention;
fig. 2 is a block diagram of a grid-connected control system in embodiment 2 of the present invention.
Detailed Description
The technical solutions of the present invention will be clearly and completely described in connection with the embodiments, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
As shown in fig. 1, the grid-connected control method in the solar photovoltaic power generation system includes the following steps:
step one: collecting photovoltaic power generation historical data and grid connection historical data in advance; the photovoltaic power generation historical data comprises historical environment data, historical power data and historical energy storage data;
step two: training a neural network model of a future power generation curve estimated according to the historical environment data and the historical power data; marking the neural network model as M1;
step three: training a cyclic neural network model according to historical power consumption data in the historical power data to generate an estimated power consumption curve for estimating future power consumption data; marking the recurrent neural network model as M2;
step four: constructing a first four-tuple set for training a first deep reinforcement learning model according to the historical power data and the historical energy storage data;
step five: training a first deep reinforcement learning model for deciding whether to perform grid connection in real time according to a future power generation power curve and an estimated power consumption curve by taking the first four-tuple set as the input of the first deep reinforcement learning model; marking the first deep reinforcement learning model as S1;
Step six: generating a residual power curve according to the historical power data;
step seven: constructing a second four-tuple set for training a second deep reinforcement learning model according to the residual power curve and grid-connected historical data;
step eight: the second four-element set is used as the input of a second deep reinforcement learning model, and the second deep reinforcement learning model of the stored power value or the released power value is determined in real time according to the residual power curve of the photovoltaic power generation system and grid connection history data when the power grid is connected is trained, so that the stability of the grid connection power is ensured; marking the second deep reinforcement learning model as S2;
step nine: based on the neural network model, the cyclic neural network model M2, the first deep reinforcement learning model and the second deep reinforcement learning model, generating a decision of whether to carry out grid connection decision of the photovoltaic power generation system and generating a stored power value or a released power value when the grid connection is carried out;
the historical environment data comprise a historical temperature curve, a historical illumination intensity curve and a historical humidity curve in a historical environment where the photovoltaic power generation system is located; it can be understood that the historical temperature curve, the historical illumination intensity curve and the historical humidity curve can be obtained through real-time induction of a temperature sensor, an illumination intensity sensor and a humidity sensor respectively;
The historical power data comprise a historical power generation power curve and a power consumption power curve which are generated by the photovoltaic power generation system under a corresponding historical environment; it can be understood that the generated power and the consumed power of the photovoltaic power generation system can be obtained through real-time measurement of a power sensor;
the historical energy storage data comprise a residual electric quantity curve of an energy storage power supply in a photovoltaic power generation system; it can be understood that the remaining power of the energy storage power supply can be measured in real time by using an electric energy meter;
the grid-connected historical data comprise grid-connected power curves stored or released in real time by an energy storage power supply when the photovoltaic power generation system performs grid connection; in the grid-connected power curve, the power value is released when the positive value is given, and the power value is stored when the negative value is given;
further, the method for training the neural network model of the future generation power curve estimated according to the historical environment data is as follows:
taking a historical temperature curve, a historical illumination intensity curve and a historical humidity curve in the historical environment data as inputs of a multi-feature time series prediction neural network model, taking a power generation curve as a prediction curve of the multi-feature time series prediction neural network model, and training the multi-feature time series prediction neural network model; in a preferred embodiment, the multi-feature time series prediction neural network model may be an LSTM neural network model;
It should be noted that, as the prior art in the field, the multi-feature time series prediction neural network is essentially a tool, and given a specific input and output task, a specific training process and parameter settings depend on a specific engineering implementation situation;
the method for generating the estimated power consumption curve for estimating the future power consumption data comprises the following steps:
presetting a predicted time step T, a first sliding step length and a sliding window length according to actual experience; converting the historical power consumption data into a plurality of training samples by using a sliding window method, taking the training samples as input of a cyclic neural network model, taking the historical power consumption data after a prediction time step T as output, taking the subsequent historical power consumption data of each training sample as a prediction target, taking the prediction accuracy rate as a training target, and training the cyclic neural network model; generating a neural network model for predicting an estimated power consumption curve according to the historical power consumption data; the cyclic neural network model may be an RNN neural network model;
specifically, a simple example of the sliding window method is as follows: assuming that a time prediction model is to be trained using historical data [1,2,3,4,5,6], predicting values for 1 time step into the future, a sliding window of length 6 and a first sliding step of length 1 may be used to generate [1,2,3], [2,3,4] and [3,4,5] as training data, with [4], [5] and [6] as prediction targets, respectively;
The method for constructing the first four-tuple set for training the first deep reinforcement learning model is as follows:
in the power generation power curve, the power consumption power curve and the residual electric quantity curve of the historical electric power data, taking a predicted time step length T as the duration of a sliding window, presetting a second sliding step length b according to actual experience, sliding the power generation power curve, the power consumption power curve and the residual electric quantity curve in time sequence, and generating a plurality of groups of original data combinations; it can be understood that the predicted time step T is taken as the duration of the sliding window, because the duration of predicting the future generated power is the predicted time step T when the actual grid-connected decision is performed later;
each group of the original data combination comprises a generating power curve with the duration of T, a power consumption curve with the duration of T and a residual electric quantity curve with the duration of T;
at the beginning of each set of said original data combinations:
taking a generated power curve and a consumed power curve in the original data combination as a first initial state;
taking whether the starting time of the original data combination is in the time of the subsequent prediction time step T or not for grid connection as a first action of selection;
Calculating the total power generation amount according to the power generation power curve in the original data combination; calculating the total power consumption according to the power consumption curve in the original data combination; marking the total power generation amount as F, the total power consumption amount as H, the residual power of the energy storage power supply at the end time of the original data combination as C, and the maximum storage power and the minimum storage power threshold of the energy storage power supply as D1 and D2 respectively; counting the time length T1 of the residual electric quantity equal to the maximum stored electric quantity D1 and the time length T2 of the residual electric quantity lower than the minimum stored electric quantity threshold value from the residual electric quantity curve; the maximum storage electric quantity is the maximum electric quantity capacity of the energy storage power supply, and the minimum storage electric quantity threshold value is an electric quantity value which is set according to practical experience and is necessary for guaranteeing emergency power utilization; calculating a first prize value Q1 corresponding to the original data set, whereinThe expression of the first prize value Q1 is: q1=a1×f-a2×h+e a3*(T1-T2) +a4×c; wherein a1, a2, a3 and a4 are respectively preset proportionality coefficients;
it can be understood that in the first reward value Q1, the influence of the difference value between the time periods T1 and T2 on the decision is exponential, which indicates that when the electric quantity of the stored power supply is too much, grid connection should be performed in time, so as to prevent electric power waste; when the storage power supply is insufficient, grid connection should not be performed, so that unnecessary safety risks caused by the use of an emergency power supply are prevented;
Combining the next set of raw data as a first next state;
then for each original data combination, < first initial state, selected first action, first prize value Q1, first next state > as its first quadruple;
the first quaternion of all the original data combinations jointly form a first quaternion set;
the method for training the first deep reinforcement learning model for deciding whether to carry out grid connection or not in real time according to the future power generation power curve and the estimated power consumption curve comprises the following steps:
the first four-tuple set is used as input of a first deep reinforcement learning model, the first deep reinforcement learning model is trained by randomly extracting a plurality of first four-tuple from the first four-tuple set, and a strategy of whether grid connection is performed or not is selected to obtain a maximum rewarding value Q1 in different initial states; preferably, the first deep reinforcement learning model is a deep Q network model;
it can be understood that in the actual application process of the first deep reinforcement learning model, the decision of whether to perform grid connection affects the future grid connection condition, so that a future power generation power curve and an estimated power consumption curve are required to be used as initial states;
The method for generating the surplus power curve according to the historical power data is as follows:
the method comprises the steps of subtracting real-time power consumption in a power consumption curve from real-time power generation in a power generation curve in time sequence to obtain real-time residual power, and connecting the real-time residual power in time sequence to obtain a residual power curve; it can be understood that the residual power curve is the real-time power which can be used by the photovoltaic power generation system to participate in grid connection in real time;
the second four-tuple set for training the second deep reinforcement learning model is constructed by the following steps:
taking the real-time surplus power of the surplus power curve as a second initial state;
a second action of selecting the real-time storage or release power of the grid-connected power curve;
the power values in the residual power curve and the grid-connected power curve at the previous moment are marked as X1 and Y1 respectively, and the power values in the residual power curve and the grid-connected power curve at the current moment are marked as X2 and Y2 respectively;
calculating the average power V of the subsequent residual power curve according to the subsequent residual power curve at the current moment; it can be understood that the average power calculation mode can be obtained by dividing the area formed by the subsequent residual power curve and the coordinate axis calculated by using the calculus method by the subsequent total duration;
Presetting a fluctuation power threshold Z according to actual experience of grid connection, setting a second rewarding value Q2, and setting the second rewarding value Q2 as a preset maximum punishment value if the value is (X2 + Y2) - (X1 + Y1) | > Z, wherein the maximum punishment value is a negative value; if | (x2+y2) - (x1+y1) | is less than or equal to Z, the calculation formula of the second prize value Q2 is q2= -e1|y2| -e2| (x2+y2) - (x1+y1) | -e3| (x2+y2) -v|; wherein e1, e2 and e3 are respectively preset proportionality coefficients;
it can be understood that, in the calculation formula of the second prize value Q2, -e1|y2| is used to reduce the intervention of the energy storage power source, -e2| (x2+y2) - (x1+y1) | is used to minimize the real-time power fluctuation, -e3| (x2+y2) -v| is used to reduce the future power fluctuation;
taking the real-time surplus power at the next moment of the surplus power curve as a second next state;
then < second initial state, selected second action, second prize value Q1, second next state > for each moment as a second quadruple;
the second four-tuple at all moments jointly form a second four-tuple set;
the training method is that when the power grid is connected, according to the residual power curve of the photovoltaic power generation system and the grid connection history data, a second deep reinforcement learning model of the stored power value or the released power value is decided in real time, wherein the second deep reinforcement learning model comprises the following steps:
The second four-tuple set is used as input of a second deep reinforcement learning model, the second deep reinforcement learning model is trained by randomly extracting a plurality of second four-tuples from the second four-tuple set, and a strategy of selecting stored or released power values by an energy storage power supply under different initial states is learned, so that a maximum rewarding value Q2 can be obtained; preferably, the second deep reinforcement learning model is a deep Q network model;
the mode of generating the decision whether to carry out grid connection decision or not for the photovoltaic power generation system and generating the stored power value or the released power value during grid connection is as follows:
loading the trained neural network model M1, the first deep reinforcement learning model S1 and the second deep reinforcement learning model S2 into an energy management system of the photovoltaic power generation system;
the energy management system acquires predicted environment data, power consumption data and energy storage data in real time, and acquires a future power generation power curve by using a neural network model M1 based on the predicted environment data;
based on the power consumption data, a cyclic neural network model M2 is used for obtaining a predicted power consumption curve; based on a future power generation power curve and an estimated power consumption curve, a first deep reinforcement learning model S1 is used for generating a decision of whether to carry out grid connection or not;
After generating a grid-connected decision, the energy management system generates a predicted residual power curve according to a future power generation power curve and a predicted power consumption curve, and generates a stored power value or a released power value decision based on the residual power curve by using a second deep reinforcement learning model;
it should be noted that, the prediction environment data may be obtained by a weather prediction model, and the power consumption data and the energy storage data are historical power consumption data and historical energy storage data before the current moment.
Example 2
As shown in fig. 2, the grid-connected control system in the solar photovoltaic power generation system comprises a historical data collection module, a model training module and a model application module; wherein, each module is connected by an electric and/or wireless network mode;
the historical data collection module is mainly used for collecting photovoltaic power generation historical data and grid-connected historical data in advance and sending the collected photovoltaic power generation historical data and the collected grid-connected historical data to the model training module;
the model training module comprises a neural network model training unit, a first deep reinforcement learning model training unit and a second deep reinforcement learning model training unit;
The neural network model training unit trains a neural network model of a future power generation curve estimated according to the historical environment data and the historical power data, trains a cyclic neural network model according to the historical power consumption data in the historical power data, and generates an estimated power consumption curve estimated for the future power consumption data;
the first deep reinforcement learning model training unit constructs a first four-tuple set for training a first deep reinforcement learning model according to the historical power data and the historical energy storage data;
the first four-element set is used as the input of a first deep reinforcement learning model, and the first deep reinforcement learning model for deciding whether to carry out grid connection or not is trained according to a future power generation power curve and an estimated power consumption curve in real time;
the second deep reinforcement learning model training unit generates a residual power curve according to the historical power data, constructs a second four-tuple set for training a second deep reinforcement learning model according to the residual power curve and grid-connected historical data, trains out the power value stored or the power value released in real time according to the residual power curve of the photovoltaic power generation system and the grid-connected historical data when the grid is connected by taking the second four-tuple set as the input of the second deep reinforcement learning model, so as to ensure the stability of the grid-connected power;
The model training module sends the neural network model, the cyclic neural network model, the first deep reinforcement learning model and the second deep reinforcement learning model to the model application module;
the model training module is used for deciding whether the photovoltaic power generation system is grid-connected or not based on the neural network model, the circulating neural network model M2, the first deep reinforcement learning model and the second deep reinforcement learning model, and generating a decision of a stored power value or a released power value after generating the grid-connected decision;
in this embodiment, all parameters preset according to practical experience are preferential settings by those skilled in the art according to the specific photovoltaic power generation system and grid connection condition.
The above embodiments are only for illustrating the technical method of the present invention and not for limiting the same, and it should be understood by those skilled in the art that the technical method of the present invention may be modified or substituted without departing from the spirit and scope of the technical method of the present invention.
Claims (8)
1. The grid-connected control method in the solar photovoltaic power generation system is characterized by comprising the following steps of:
Step one: collecting photovoltaic power generation historical data and grid connection historical data in advance; the photovoltaic power generation historical data comprises historical environment data, historical power data and historical energy storage data;
step two: training a neural network model of a future power generation curve estimated according to the historical environment data and the historical power data; marking the neural network model as M1;
step three: training a cyclic neural network model according to historical power consumption data in the historical power data to generate an estimated power consumption curve for estimating future power consumption data; marking the recurrent neural network model as M2;
step four: constructing a first four-tuple set for training a first deep reinforcement learning model according to the historical power data and the historical energy storage data;
step five: training a first deep reinforcement learning model according to a future power generation power curve and an estimated power consumption curve by taking the first four-element set as the input of the first deep reinforcement learning model so as to determine whether to carry out grid connection in real time; marking the first deep reinforcement learning model as S1;
step six: generating a residual power curve according to the historical power data;
Step seven: constructing a second four-tuple set for training a second deep reinforcement learning model according to the residual power curve and grid-connected historical data;
step eight: training a second deep reinforcement learning model for deciding a stored power value or a released power value in real time according to a residual power curve of the photovoltaic power generation system and grid-connected historical data when the power grid is connected by taking the second four-tuple set as the input of the second deep reinforcement learning model; marking the second deep reinforcement learning model as S2;
step nine: based on the neural network model, the cyclic neural network model M2, the first deep reinforcement learning model and the second deep reinforcement learning model, generating a decision of whether to carry out grid connection decision of the photovoltaic power generation system and generating a stored power value or a released power value when the grid connection is carried out;
the historical environment data comprise a historical temperature curve, a historical illumination intensity curve and a historical humidity curve in a historical environment where the photovoltaic power generation system is located;
the historical power data comprise a historical power generation power curve and a power consumption power curve which are generated by the photovoltaic power generation system under a corresponding historical environment;
The historical energy storage data comprises a residual electric quantity curve of an energy storage power supply in the photovoltaic power generation system;
the grid-connected historical data comprise grid-connected power curves stored or released in real time by an energy storage power supply when the photovoltaic power generation system performs grid connection;
the method for constructing the first four-tuple set for training the first deep reinforcement learning model is as follows:
in the power generation power curve, the power consumption power curve and the residual electric quantity curve of the historical electric power data, taking a predicted time step length T as the duration of a sliding window, and sliding the power generation power curve, the power consumption power curve and the residual electric quantity curve according to a preset second sliding step length b in time sequence to generate a plurality of groups of original data combinations;
each group of the original data combination comprises a generating power curve with the duration of T, a power consumption curve with the duration of T and a residual electric quantity curve with the duration of T;
at the starting time of each group of the original data combination, taking a generated power curve and a consumed power curve in the original data combination as a first initial state;
taking the starting time of the original data combination as a first action of selecting whether grid connection is carried out in the time of a subsequent prediction time step T;
Calculating the total power generation amount according to the power generation power curve in the original data combination; calculating the total power consumption according to the power consumption curve in the original data combination; marking the total power generation amount as F, the total power consumption amount as H, the residual power amount of the energy storage power supply at the end time of the original data combination as C, and the maximum storage power amount and the minimum storage power amount threshold value of the energy storage power supply as D1 and D2 respectively; counting the time length T1 of the residual electric quantity equal to the maximum stored electric quantity D1 and the time length T2 of the residual electric quantity lower than the minimum stored electric quantity threshold value from the residual electric quantity curve; the maximum storage electric quantity is the maximum electric quantity capacity of the energy storage power supply, and the minimum storage electric quantity threshold value is a preset electric quantity value which is used for guaranteeing emergency power utilization; calculating a first reward value Q1 corresponding to the original data set, wherein the expression form of the first reward value Q1 is as follows: q1=a1×f-a2×h+e a3*(T1-T2) +a4×c; wherein a1, a2, a3 and a4 are respectively preset proportionality coefficients;
combining the next set of raw data as a first next state;
then for each original data combination, < first initial state, selected first action, first prize value Q1, first next state > as its first quadruple;
The first quaternion of all the original data combinations jointly form a first quaternion set;
the second four-tuple set for training the second deep reinforcement learning model is constructed by the following steps:
taking the real-time surplus power of the surplus power curve as a second initial state;
a second action of selecting the real-time storage or release power of the grid-connected power curve;
the power values in the residual power curve and the grid-connected power curve at the previous moment are marked as X1 and Y1 respectively, and the power values in the residual power curve and the grid-connected power curve at the current moment are marked as X2 and Y2 respectively;
calculating the average power V of the subsequent residual power curve according to the subsequent residual power curve at the current moment;
presetting a fluctuation power threshold Z, setting a second rewarding value Q2, and setting the second rewarding value Q2 as a preset maximum punishment value if the value is (X2 +Y2) - (X1 +Y1) | > Z, wherein the maximum punishment value is a negative value; if | (x2+y2) - (x1+y1) | is less than or equal to Z, the calculation formula of the second prize value Q2 is q2= -e1|y2| -e2| (x2+y2) - (x1+y1) | -e3| (x2+y2) -v|; wherein e1, e2 and e3 are respectively preset proportionality coefficients;
taking the real-time surplus power at the next moment of the surplus power curve as a second next state;
Then < second initial state, selected second action, second prize value Q1, second next state > for each moment as a second quadruple;
the second quaternions at all times together form a second quaternion set.
2. The grid-connected control method in a solar photovoltaic power generation system according to claim 1, wherein the method for training a neural network model of a future power generation curve estimated according to historical environmental data is as follows:
and taking the historical temperature curve, the historical illumination intensity curve and the historical humidity curve in the historical environment data as inputs of the multi-feature time series prediction neural network model, taking the generated power curve as a prediction curve of the multi-feature time series prediction neural network model, and training the multi-feature time series prediction neural network model.
3. The grid-connected control method in a solar photovoltaic power generation system according to claim 2, wherein the method for generating the estimated power consumption curve estimated for future power consumption data is as follows:
according to a preset prediction time step T, a first sliding step length and a sliding window length; converting the historical power consumption data into a plurality of training samples by using a sliding window method, taking the training samples as input of a cyclic neural network model, taking the historical power consumption data after a prediction time step T as output, taking the subsequent historical power consumption data of each training sample as a prediction target, taking the prediction accuracy rate as a training target, and training the cyclic neural network model; and generating a neural network model for predicting an estimated power consumption curve according to the historical power consumption data.
4. The grid-connected control method in a solar photovoltaic power generation system according to claim 3, wherein the training of the first deep reinforcement learning model for deciding whether to perform grid-connection in real time according to the future power generation power curve and the estimated power consumption curve is:
and taking the first four-tuple set as the input of a first deep reinforcement learning model, wherein the first deep reinforcement learning model carries out training by randomly extracting a plurality of first four-tuple from the first four-tuple set, and learns whether grid connection is carried out or not under different initial states so as to obtain a strategy of a maximum rewarding value Q1.
5. The grid-tie control method in a solar photovoltaic power generation system according to claim 4, wherein the method for generating the surplus power curve according to the historical power data is as follows:
and subtracting the real-time power consumption power in the power consumption power curve from the real-time power generation power in the power generation power curve in time sequence to obtain real-time residual power, and connecting the real-time residual power in time sequence to obtain a residual power curve.
6. The grid-connected control method in a solar photovoltaic power generation system according to claim 5, wherein the training is that, when the grid is connected, according to the remaining power curve of the photovoltaic power generation system and the grid-connected historical data, the mode of the second deep reinforcement learning model for deciding the stored power value or the released power value in real time is as follows:
And taking the second four-tuple set as the input of a second deep reinforcement learning model, wherein the second deep reinforcement learning model is trained by randomly extracting a plurality of second four-tuples from the second four-tuple set, and learns the power values stored or released by the energy storage power supply under different initial states so as to obtain a strategy of the maximum rewarding value Q2.
7. The grid-tie control method in a solar photovoltaic power generation system according to claim 6, wherein the manner of making a decision on whether to perform grid-tie and a decision on whether to generate a stored power value or a released power value when tie is made for the photovoltaic power generation system is:
loading the trained neural network model M1, the first deep reinforcement learning model S1 and the second deep reinforcement learning model S2 into an energy management system of the photovoltaic power generation system;
the energy management system acquires predicted environment data, power consumption data and energy storage data in real time, and acquires a future power generation power curve by using a neural network model M1 based on the predicted environment data;
based on the power consumption data, a cyclic neural network model M2 is used for obtaining a predicted power consumption curve; based on a future power generation power curve and an estimated power consumption curve, a first deep reinforcement learning model S1 is used for generating a decision of whether to carry out grid connection or not;
After generating a grid-connected decision, the energy management system generates a predicted residual power curve according to a future power generation power curve and a predicted power consumption curve, and generates a stored power value or a released power value decision based on the residual power curve by using a second deep reinforcement learning model.
8. The grid-connected control system in the solar photovoltaic power generation system is realized based on the grid-connected control method in the solar photovoltaic power generation system according to any one of claims 1 to 7, and is characterized by comprising a historical data collection module, a model training module and a model application module; wherein, each module is connected by an electric and/or wireless network mode;
the historical data collection module is mainly used for collecting photovoltaic power generation historical data and grid-connected historical data in advance and sending the collected photovoltaic power generation historical data and the collected grid-connected historical data to the model training module;
the model training module comprises a neural network model training unit, a first deep reinforcement learning model training unit and a second deep reinforcement learning model training unit;
the neural network model training unit trains a neural network model of a future power generation curve estimated according to the historical environment data and the historical power data, trains a cyclic neural network model according to the historical power consumption data in the historical power data, and generates an estimated power consumption curve estimated for the future power consumption data;
The first deep reinforcement learning model training unit constructs a first four-tuple set for training a first deep reinforcement learning model according to the historical power data and the historical energy storage data;
the first four-element set is used as the input of a first deep reinforcement learning model, and the first deep reinforcement learning model for deciding whether to carry out grid connection or not is trained according to a future power generation power curve and an estimated power consumption curve in real time;
the second deep reinforcement learning model training unit generates a residual power curve according to the historical power data, constructs a second four-tuple set for training a second deep reinforcement learning model according to the residual power curve and the grid-connected historical data, takes the second four-tuple set as the input of the second deep reinforcement learning model, trains out the power value stored or the power value released in real time according to the residual power curve of the photovoltaic power generation system and the grid-connected historical data when the grid is connected;
the model training module sends the neural network model, the cyclic neural network model, the first deep reinforcement learning model and the second deep reinforcement learning model to the model application module;
the model training module is used for deciding whether the photovoltaic power generation system is grid-connected or not based on the neural network model, the circulating neural network model M2, the first deep reinforcement learning model and the second deep reinforcement learning model, and generating a decision of a stored power value or a released power value after the grid-connected decision is generated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310581426.XA CN116388279B (en) | 2023-05-23 | 2023-05-23 | Grid-connected control method and control system for solar photovoltaic power generation system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310581426.XA CN116388279B (en) | 2023-05-23 | 2023-05-23 | Grid-connected control method and control system for solar photovoltaic power generation system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116388279A CN116388279A (en) | 2023-07-04 |
| CN116388279B true CN116388279B (en) | 2024-01-23 |
Family
ID=86969631
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310581426.XA Active CN116388279B (en) | 2023-05-23 | 2023-05-23 | Grid-connected control method and control system for solar photovoltaic power generation system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116388279B (en) |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109002926A (en) * | 2018-07-27 | 2018-12-14 | 中国计量大学 | The photovoltaic power generation quantity prediction model and its construction method of a kind of high accuracy and application |
| CN109347149A (en) * | 2018-09-20 | 2019-02-15 | 国网河南省电力公司电力科学研究院 | Micro-capacitance sensor energy storage dispatching method and device based on depth Q value network intensified learning |
| CN109934423A (en) * | 2019-04-25 | 2019-06-25 | 山东大学 | Predicting power of photovoltaic plant method and system based on gird-connected inverter operation data |
| CN110070226A (en) * | 2019-04-24 | 2019-07-30 | 河海大学 | Photovoltaic power prediction technique and system based on convolutional neural networks and meta learning |
| CN111668873A (en) * | 2020-06-28 | 2020-09-15 | 合肥阳光新能源科技有限公司 | Energy storage and charging power regulation and control method and energy management system for light storage grid-connected power generation system |
| CN111884213A (en) * | 2020-07-27 | 2020-11-03 | 国网北京市电力公司 | Power distribution network voltage adjusting method based on deep reinforcement learning algorithm |
| CN112117760A (en) * | 2020-08-13 | 2020-12-22 | 国网浙江省电力有限公司台州供电公司 | Micro-grid energy scheduling method based on double-Q-value network deep reinforcement learning |
| CN112186799A (en) * | 2020-09-22 | 2021-01-05 | 中国电力科学研究院有限公司 | Distributed energy system autonomous control method and system based on deep reinforcement learning |
| CN112330021A (en) * | 2020-11-05 | 2021-02-05 | 中国电子科技集团公司第十八研究所 | Network coordination control method of distributed optical storage system |
| CN112614009A (en) * | 2020-12-07 | 2021-04-06 | 国网四川省电力公司电力科学研究院 | Power grid energy management method and system based on deep expected Q-learning |
| CN113743674A (en) * | 2021-09-10 | 2021-12-03 | 中国电力科学研究院有限公司 | Energy storage output prediction method, system, equipment and medium based on deep learning |
| CN113837434A (en) * | 2021-08-16 | 2021-12-24 | 同盾科技有限公司 | Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium |
| CN113919232A (en) * | 2021-10-25 | 2022-01-11 | 北京航天创智科技有限公司 | Photovoltaic power station power prediction method and system based on recurrent neural network |
| CN113935463A (en) * | 2021-09-30 | 2022-01-14 | 南方电网数字电网研究院有限公司 | Microgrid controller based on artificial intelligence control method |
| CN114123273A (en) * | 2021-11-12 | 2022-03-01 | 青海综合能源服务有限公司 | Control method and system of wind power-photovoltaic-energy storage combined system |
| CN114169627A (en) * | 2021-12-14 | 2022-03-11 | 湖南工商大学 | Deep reinforcement learning distributed photovoltaic power generation excitation method |
| CN114285075A (en) * | 2021-12-22 | 2022-04-05 | 浙江工业大学 | Micro-grid energy online optimization method based on distributed deep reinforcement learning |
| CN115986834A (en) * | 2022-12-07 | 2023-04-18 | 北京交通大学 | Near-end strategy optimization algorithm-based optical storage charging station operation optimization method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112072655B (en) * | 2020-09-10 | 2022-05-27 | 天津大学 | Hybrid energy storage optimal configuration method for grid-connected wind energy storage power generation system |
-
2023
- 2023-05-23 CN CN202310581426.XA patent/CN116388279B/en active Active
Patent Citations (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109002926A (en) * | 2018-07-27 | 2018-12-14 | 中国计量大学 | The photovoltaic power generation quantity prediction model and its construction method of a kind of high accuracy and application |
| CN109347149A (en) * | 2018-09-20 | 2019-02-15 | 国网河南省电力公司电力科学研究院 | Micro-capacitance sensor energy storage dispatching method and device based on depth Q value network intensified learning |
| CN110070226A (en) * | 2019-04-24 | 2019-07-30 | 河海大学 | Photovoltaic power prediction technique and system based on convolutional neural networks and meta learning |
| CN109934423A (en) * | 2019-04-25 | 2019-06-25 | 山东大学 | Predicting power of photovoltaic plant method and system based on gird-connected inverter operation data |
| WO2020215405A1 (en) * | 2019-04-25 | 2020-10-29 | 山东大学 | Photovoltaic power station power prediction method and system based on grid-tie inverter operating data |
| CN111668873A (en) * | 2020-06-28 | 2020-09-15 | 合肥阳光新能源科技有限公司 | Energy storage and charging power regulation and control method and energy management system for light storage grid-connected power generation system |
| CN111884213A (en) * | 2020-07-27 | 2020-11-03 | 国网北京市电力公司 | Power distribution network voltage adjusting method based on deep reinforcement learning algorithm |
| CN112117760A (en) * | 2020-08-13 | 2020-12-22 | 国网浙江省电力有限公司台州供电公司 | Micro-grid energy scheduling method based on double-Q-value network deep reinforcement learning |
| CN112186799A (en) * | 2020-09-22 | 2021-01-05 | 中国电力科学研究院有限公司 | Distributed energy system autonomous control method and system based on deep reinforcement learning |
| CN112330021A (en) * | 2020-11-05 | 2021-02-05 | 中国电子科技集团公司第十八研究所 | Network coordination control method of distributed optical storage system |
| CN112614009A (en) * | 2020-12-07 | 2021-04-06 | 国网四川省电力公司电力科学研究院 | Power grid energy management method and system based on deep expected Q-learning |
| CN113837434A (en) * | 2021-08-16 | 2021-12-24 | 同盾科技有限公司 | Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium |
| CN113743674A (en) * | 2021-09-10 | 2021-12-03 | 中国电力科学研究院有限公司 | Energy storage output prediction method, system, equipment and medium based on deep learning |
| CN113935463A (en) * | 2021-09-30 | 2022-01-14 | 南方电网数字电网研究院有限公司 | Microgrid controller based on artificial intelligence control method |
| CN113919232A (en) * | 2021-10-25 | 2022-01-11 | 北京航天创智科技有限公司 | Photovoltaic power station power prediction method and system based on recurrent neural network |
| CN114123273A (en) * | 2021-11-12 | 2022-03-01 | 青海综合能源服务有限公司 | Control method and system of wind power-photovoltaic-energy storage combined system |
| CN114169627A (en) * | 2021-12-14 | 2022-03-11 | 湖南工商大学 | Deep reinforcement learning distributed photovoltaic power generation excitation method |
| CN114285075A (en) * | 2021-12-22 | 2022-04-05 | 浙江工业大学 | Micro-grid energy online optimization method based on distributed deep reinforcement learning |
| CN115986834A (en) * | 2022-12-07 | 2023-04-18 | 北京交通大学 | Near-end strategy optimization algorithm-based optical storage charging station operation optimization method and system |
Non-Patent Citations (2)
| Title |
|---|
| 基于深度学习的超短期光伏精细化预测模型研究;史佳琪;张建华;;电力建设(第06期);全文 * |
| 基于深度强化学习的微电网复合储能协调控制方法;张自东;邱才明;张东霞;徐舒玮;贺兴;;电网技术(第06期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116388279A (en) | 2023-07-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104538999B (en) | Photovoltaic generating energy storage system and energy scheduling method thereof | |
| CN111581850A (en) | Full-period power battery management system applying digital twinning technology | |
| CN110854932A (en) | Multi-time scale optimization scheduling method and system for AC/DC power distribution network | |
| EP3640869A1 (en) | Method for predicting an energy demand, data processing system and renewable power plant with a storage | |
| CN108227759B (en) | Solar tracking control system based on neural network prediction technology | |
| CN115425680A (en) | Power prediction model construction and prediction method of multi-energy combined power generation system | |
| CN116599151A (en) | Source network storage safety management method based on multi-source data | |
| CN111553750A (en) | Energy storage bidding strategy method considering power price uncertainty and loss cost | |
| CN116436019B (en) | Multi-resource coordination optimization method, device and storage medium | |
| CN116388279B (en) | Grid-connected control method and control system for solar photovoltaic power generation system | |
| CN116722605B (en) | Power distribution network scheduling optimization method based on Internet of things | |
| CN114154790A (en) | Industrial park light storage capacity configuration method based on demand management and flexible load | |
| CN117060386A (en) | Micro-grid energy storage scheduling optimization method based on value distribution depth Q network | |
| CN114814350B (en) | Energy monitoring method, system, device and medium based on park electric data prediction | |
| CN106355297A (en) | Electrolytic aluminum load characteristic-based power grid loss reduction optimization method | |
| CN112731803B (en) | Energy storage system charging and discharging control method, device, equipment and readable storage medium | |
| CN114904655A (en) | Single-electric-field energy-saving control method and device | |
| CN113935182A (en) | Quantitative battery life attenuation prediction method based on mixed integer programming model | |
| CN114298429A (en) | Power distribution network scheme aided decision-making method, system, device and storage medium | |
| CN114530872A (en) | Multilateral shared energy storage optimization configuration and cost sharing method thereof | |
| CN114689122B (en) | Equipment fault monitoring method, device, equipment and medium | |
| CN113013905B (en) | Scheduling method and device for wind-solar micro-energy collection wireless sensing system | |
| CN115663921B (en) | Method and system for determining regulation and control plan of wind-solar storage and charging micro-grid | |
| CN117996858A (en) | Intelligent optimization method and device for power distribution network | |
| Zhou et al. | Application of genetic neural network in power battery Charging State-of-Charge Estimation |
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 |