Intelligent prediction method for regulation and control capability of optical storage and charging station

Abstract

The invention discloses an intelligent prediction method for the regulation and control capability of an optical storage and charging station, relates to the field of power prediction of the optical storage and charging station, and provides an innovative solution for the prediction of the regulation and control capability of the optical storage and charging station by combining a generative adversarial network with correlation analysis and Wasserstein distance optimization. The correlation analysis enables the model to accurately identify and utilize the complex coupling relationship between the photovoltaic output and the charging demand of the electric vehicle, thereby improving the prediction accuracy; the generative adversarial network can generate a prediction result very similar to real station operation data through a unique optical storage and charging station regulation and control capability prediction scene generation model and an optical storage and charging station regulation and control capability prediction scene evaluation model mechanism, and the reliability of the model is improved; the generative adversarial network is optimized through the Wasserstein distance, and the stability and training effect of the model are enhanced, so that the prediction model not only can process a large-scale data set, but also can adapt to the characteristics of highly dynamic changes of the power market and renewable energy sources.

Description

Technical Field

[0001] The present invention relates to the technical field of power prediction of photovoltaic storage and charging stations, and specifically to an intelligent prediction method for the control capability of photovoltaic storage and charging stations. Background Art

[0002] Under the current background of energy transformation, photovoltaic storage and charging stations are the key link in the integration of new energy and power systems. Their control capabilities are directly related to the stability of the power grid and the effective use of energy. With the continuous increase in the proportion of renewable energy, the complexity of power grid control is also increasing. Traditional energy control methods can no longer meet the needs of high-proportion renewable energy grid connection. Therefore, accurate prediction of the coordinated control capabilities of photovoltaic storage and charging stations can not only optimize power grid operation, but also improve energy utilization efficiency. It has important practical significance and broad application prospects.

[0003] Although there are currently a variety of methods that attempt to solve the problem of predicting the regulation capacity of photovoltaic storage and charging stations, these methods often ignore the interaction between the internal components of photovoltaic storage and charging stations and the influence of external factors. For example, few methods in existing research can fully consider the correlation between photovoltaic output and market charging prices, and the impact of this correlation on the prediction model. In addition, most models do not handle nonlinear and non-stationary data characteristics well, which means that the accuracy and reliability of predictions still need to be improved.

[0004] In view of this, the present application proposes a new prediction method to solve the problem of inaccurate prediction of the control capacity of photovoltaic storage and charging stations in the prior art. Summary of the invention

[0005] The purpose of the present invention is to provide a method for intelligently predicting the control capability of a photovoltaic storage and charging station to solve the problems raised in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for intelligently predicting the control capability of a photovoltaic storage and charging station, the method comprising the following steps:

[0007] S1. Mining the coupling relationship between the control capabilities of photovoltaic storage and charging stations based on the correlation analysis method;

[0008] S2. Construct a prediction model for the control capability of photovoltaic storage and charging stations based on generative adversarial networks;

[0009] S3, optimize the prediction model constructed in S2 based on Wasserstein distance;

[0010] S4: Predict the control capacity of the photovoltaic storage and charging station based on the prediction model optimized by S3.

[0011] According to the above technical solution, mining the coupling relationship of the control capability of the photovoltaic storage and charging station in step S1 specifically includes the following steps:

[0012] S101, collecting and preprocessing historical operation data of the photovoltaic storage and charging station;

[0013] S102, using the Pearson correlation coefficient to measure the correlation between photovoltaic output and electric vehicle charging demand;

[0014] S103. Predicting the response of the electric vehicle charging demand to the photovoltaic output based on the correlation between the photovoltaic output and the electric vehicle charging demand.

[0015] According to the above technical solution, the historical data includes: photovoltaic power generation output, energy storage unit status, charging equipment usage and external factors;

[0016] The preprocessing of the collected historical operation data includes: data cleaning, standardization and normalization to ensure that the data quality meets the requirements of subsequent analysis;

[0017] The correlation in step S102 is expressed as:

[0018]

[0019] Where PV represents photovoltaic power output, EV represents electric vehicle charging demand; cov(PV, EV) represents the covariance between photovoltaic power output and electric vehicle charging demand; σ PV and σ EV represent the standard deviation of PV output and EV charging demand, respectively;

[0020] The response of the electric vehicle charging demand to the photovoltaic output in step S103 is expressed as:

[0021] EV=β 0 +β 1 *PV+ε;

[0022] Among them, β 0 and β 1 represents the response coefficient of electric vehicle charging demand to photovoltaic output; ε represents the error value of the model.

[0023] According to the above technical solution, in step S2, the establishment of the prediction model specifically includes the following steps:

[0024] S201. Construct a scenario generation model for predicting the control capability of photovoltaic storage and charging stations;

[0025] S202. Construct a scenario assessment model for predicting the control capability of photovoltaic storage and charging stations;

[0026] S203: Update the network parameters of the PV-storage-charging station control capability prediction scenario generation model and the PV-storage-charging station control capability prediction scenario evaluation model.

[0027] According to the above technical solution, in step S201, the network structure of the photovoltaic storage and charging station control capability prediction scenario generation model is set, including multiple fully connected layers or convolutional layers; the specific architecture depends on the dimension and complexity of the input data; the goal of the photovoltaic storage and charging station control capability prediction scenario generation model is to generate actual possible operation scenarios by learning the control data of the photovoltaic storage and charging station; the photovoltaic storage and charging station control capability prediction scenario generation model receives the noise vector z and generates a predicted control scenario through a neural network; the photovoltaic storage and charging station control capability prediction scenario generation model is:

[0028] G(z)=f (L) (...f (2) (f (1) (z;θ (1) );θ (2) )...;θ (L) );

[0029] Among them, G(z) represents the output of the prediction scenario generation model of the photovoltaic storage and charging station regulation capability, that is, the intelligently generated regulation capability scenario, including the regulation capability of photovoltaic, charging station, and energy storage; f G (L) represents the Lth in the prediction scenario generation model of the photovoltaic storage and charging station regulation capability G The neural network function of the layer; θ G (L) represents the Lth in the prediction scenario generation model of the photovoltaic storage and charging station regulation capability G The parameters of the layer.

[0030] According to the above technical solution, in step S202, the network structure of the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model is set, including a fully connected layer or a convolutional layer; the purpose of the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model is to accurately determine whether the input data is actual data or data generated by the photovoltaic storage and charging station regulation capacity prediction scenario generation model; the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model receives data from the photovoltaic storage and charging station regulation capacity prediction scenario generation model and real data respectively, and outputs real probability evaluation data; the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model is:

[0031] D(x)=σ(f (L) (...f (2) (f (1) (x;θ (1) );θ (2) )...;θ (L) ));

[0032] Where D(x) represents the output of the PV-storage-charging station control capability prediction scenario evaluation model; σ is the activation function of the PV-storage-charging station control capability prediction scenario evaluation model, which is used to compress the output into the range of (0,1); f (L) represents the network function of the Lth layer in the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; x represents the input of the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; θ (L) Represents the parameters of the Lth layer in the scenario evaluation model for predicting the control capability of photovoltaic storage and charging stations.

[0033] According to the above technical solution, in step S203, the network parameters are updated as follows:

[0034]

[0035] Among them, θ D ,θ G are the parameters of the PV-storage-charging station regulation capability prediction scenario evaluation model and the PV-storage-charging station regulation capability prediction scenario generation model, α is the learning rate, are the gradients of the loss function with respect to their respective parameters.

[0036] According to the above technical solution, step S3 specifically includes the following steps:

[0037] S301. Establishing a distance model for the real data distribution and generated data distribution of the power prediction of the photovoltaic storage and charging station based on the Wasserstein distance;

[0038] S302, using a weight clipping method to constrain the gradient of the PV-storage-charging station regulation capability prediction scenario evaluation model to ensure that the solution of the PV-storage-charging station regulation capability prediction scenario evaluation model is continuous;

[0039] S303. Use gradient penalty to constrain the gradient of the PV-storage-charging-station regulation capability prediction scenario evaluation model to ensure that the PV-storage-charging-station regulation capability prediction scenario evaluation model does not have problems such as gradient explosion and training instability.

[0040] According to the above technical solution, the distance model in step S301 is specifically:

[0041]

[0042] Where f(x) represents the 1-Lipschitz function, ‖f‖ L ≤1 means that the Lipschitz constant of function f(x) does not exceed 1; the sup operation means maximizing the expected difference for all 1-Lipschitz functions;

[0043] In step S302, the constraints are specifically:

[0044] ‖θ C ‖ ∞ ≤c;

[0045] Among them, θ C represents the weight of the PV-storage-charging station control capability prediction scenario evaluation model, and c represents a constant. In this way, the gradient of the PV-storage-charging station control capability prediction scenario evaluation model is kept in a reasonable range at each update, thereby ensuring 1-Lipschitz continuity.

[0046] In step S303, the constraints are specifically:

[0047]

[0048] in, is a sample that linearly interpolates between the real data and the generated data, It is the gradient of the PV-storage-charging-station control capability prediction scenario evaluation model for these interpolation samples; in this way, the gradient of the PV-storage-charging-station control capability prediction scenario evaluation model is forced to remain within a reasonable range, avoiding the problems of gradient explosion and training instability.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention can not only improve the energy management efficiency of the photovoltaic storage and charging station, but also provide strong support for the sustainable development of the power system, especially showing its unique advantages in improving the utilization rate of renewable energy and reducing the risk of power grid operation;

[0051] Correlation analysis enables the model to accurately identify and utilize the complex coupling relationship between factors such as photovoltaic output and electric vehicle charging demand, thereby improving the accuracy of the prediction. The generative adversarial network can generate prediction results that are very close to the actual station operation data through its unique photovoltaic storage and charging station regulation capability prediction scenario generation model and photovoltaic storage and charging station regulation capability prediction scenario evaluation model mechanism, which greatly improves the realism and reliability of the model. In addition, the generative adversarial network is optimized through the Wasserstein distance, which further enhances the stability and training effect of the model, so that the prediction model can not only process large-scale data sets, but also adapt to the highly dynamic characteristics of the electricity market and renewable energy. In general, the present invention provides an efficient, accurate and stable prediction tool, which provides strong support for the formulation and optimization of control strategies for photovoltaic storage and charging stations;

[0052] The present invention aims to provide a more accurate and robust regulation capability prediction tool for photovoltaic storage and charging stations, supporting the sustainable development of power systems in an environment with high renewable energy penetration. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a learning curve diagram of the method proposed in the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] This embodiment provides the following implementation method: a method for intelligently predicting the control capability of a photovoltaic storage and charging station, the prediction method comprising the following steps:

[0056] S1. Mining the coupling relationship between the control capabilities of photovoltaic storage and charging stations based on the correlation analysis method;

[0057] Specific:

[0058] In step S1, mining the coupling relationship of the control capability of the photovoltaic storage and charging station specifically includes the following steps:

[0059] S101, collecting and preprocessing historical operation data of the photovoltaic storage and charging station;

[0060] S102. Use the Pearson correlation coefficient to measure the correlation between PV output and EV charging demand;

[0061] S103, predicting the response of electric vehicle charging demand to photovoltaic output based on the correlation between photovoltaic output and electric vehicle charging demand;

[0062] Furthermore, the historical data includes: photovoltaic power generation output, energy storage unit status, charging equipment usage and external factors, such as weather conditions;

[0063] The preprocessing of the collected historical operation data includes: data cleaning, standardization and normalization to ensure that the data quality meets the requirements of subsequent analysis;

[0064] The correlation in step S102 is expressed as:

[0065]

[0066] Where PV represents photovoltaic power output, EV represents electric vehicle charging demand; cov(PV, EV) represents the covariance between photovoltaic power output and electric vehicle charging demand; σ PV and σ EV represent the standard deviation of PV output and EV charging demand, respectively;

[0067] The response of the electric vehicle charging demand to the photovoltaic output in step S103 is expressed as:

[0068] EV=β 0 +β 1 *PV+ε;

[0069] Among them, β 0 and β 1 represents the response coefficient of electric vehicle charging demand to photovoltaic output; ε represents the error value of the model.

[0070] The correlation here can reflect the relationship between photovoltaic output and electric vehicle charging demand, and can be used as one of the inputs of the subsequent model. Compared with this method of considering correlation, the prediction of the traditional method is independent and uncoupled, so there may be a problem of large prediction randomness. By considering the correlation between different sites, this prediction error can be constrained within a certain range.

[0071] S2. Construct a prediction model for the control capability of photovoltaic storage and charging stations based on generative adversarial networks;

[0072] Specific:

[0073] In step S2, the establishment of the prediction model specifically includes the following steps:

[0074] S201. Construct a scenario generation model for predicting the control capability of photovoltaic storage and charging stations;

[0075] Furthermore, in step S201, the network structure of the photovoltaic storage and charging station control capability prediction scenario generation model is set, including multiple fully connected layers or convolutional layers; the specific architecture depends on the dimension and complexity of the input data; the goal of the photovoltaic storage and charging station control capability prediction scenario generation model is to generate actual possible operation scenarios by learning the control data of the photovoltaic storage and charging station; the photovoltaic storage and charging station control capability prediction scenario generation model receives a Gaussian white noise vector z, which is randomly input to the model to improve the generalization ability of the model. Generate predicted control capability scenarios through neural networks; specifically how to generate predicted control capability scenarios through neural networks. The photovoltaic storage and charging station control capability prediction scenario generation model is:

[0076]

[0077] Among them, G(z) represents the output of the prediction scenario generation model of the photovoltaic storage and charging station regulation capability, that is, the intelligently generated regulation capability scenario, including the regulation capability of photovoltaic, charging station, and energy storage; f G (L) represents the Lth in the prediction scenario generation model of the photovoltaic storage and charging station regulation capability G The neural network function of the layer; θ G(L) represents the Lth in the prediction scenario generation model of the photovoltaic storage and charging station regulation capability G The parameters of the layer.

[0078] S202. Construct a scenario assessment model for predicting the control capability of photovoltaic storage and charging stations;

[0079] Further, in step S202, the network structure of the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model is set, including a fully connected layer or a convolutional layer; the purpose of the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model is to accurately determine whether the input data is actual data or data generated by the photovoltaic storage and charging station regulation capacity prediction scenario generation model; the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model receives data from the photovoltaic storage and charging station regulation capacity prediction scenario generation model and real data respectively, and outputs real probability evaluation data; the photovoltaic storage and charging station regulation capacity prediction scenario evaluation model is:

[0080] D(x)=σ(f (L) (...f (2) (f (1) (x;θ (1) );θ (2) )...;θ (L) ));

[0081] Where D(x) represents the output of the PV storage and charging station control capability prediction scenario evaluation model. σ is the activation function of the PV storage and charging station control capability prediction scenario evaluation model, which is used to compress the output into the range of (0,1); f (L) represents the network function of the Lth layer in the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station. x represents the input of the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; θ (L) Represents the parameters of the Lth layer in the scenario evaluation model for predicting the control capability of photovoltaic storage and charging stations.

[0082] The evaluation model of the scenario prediction of the control capability of the photovoltaic storage and charging station is shown in the formula. Note that x represents the input of the evaluation model of the scenario prediction of the control capability of the photovoltaic storage and charging station, that is, the output of the generation model of the scenario prediction of the control capability of the photovoltaic storage and charging station. The output of the generation model of the scenario prediction of the control capability of the photovoltaic storage and charging station is put into the model, and the output is output using a multi-layer network. The output quantity is the evaluation of the scenario prediction of the control capability of the photovoltaic storage and charging station.

[0083] S203, updating the network parameters of the PV-storage-charging station regulation capability prediction scenario generation model and the PV-storage-charging station regulation capability prediction scenario evaluation model;

[0084] Furthermore, in step S203, the network parameters are updated as follows:

[0085]

[0086] Among them, θ D ,θ G are the parameters of the PV-storage-charging station regulation capability prediction scenario evaluation model and the PV-storage-charging station regulation capability prediction scenario generation model, α is the learning rate, are the gradients of the loss function with respect to their respective parameters.

[0087] S3, optimize the prediction model constructed in S2 based on Wasserstein distance;

[0088] Specifically, step S3 includes the following steps:

[0089] S301. Establishing a distance model for the real data distribution and generated data distribution of the power prediction of the photovoltaic storage and charging station based on the Wasserstein distance;

[0090] S302, using a weight clipping method to constrain the gradient of the PV-storage-charging station regulation capability prediction scenario evaluation model to ensure that the solution of the PV-storage-charging station regulation capability prediction scenario evaluation model is continuous;

[0091] S303, using gradient penalty to constrain the gradient of the solar-storage-charging-station regulation capability prediction scenario evaluation model to ensure that the solar-storage-charging-station regulation capability prediction scenario evaluation model does not have problems such as gradient explosion and unstable training;

[0092] Furthermore, the distance model in step S301 is specifically:

[0093]

[0094] Where f(x) represents the 1-Lipschitz function, ‖f‖L≤1 The Lipschitz constant of the function f(x) does not exceed 1; the sup operation represents maximizing the expected difference for all 1-Lipschitz functions;

[0095] In step S302, the constraints are specifically:

[0096] ‖θ C ‖ ∞ ≤c;

[0097] Among them, θ C represents the weight of the PV-storage-charging station control capability prediction scenario evaluation model, and c represents a constant. In this way, the gradient of the PV-storage-charging station control capability prediction scenario evaluation model is kept in a reasonable range at each update, thereby ensuring 1-Lipschitz continuity.

[0098] In step S303, the constraints are specifically:

[0099]

[0100] in, is a sample that linearly interpolates between the real data and the generated data, It is the gradient of the PV-storage-charging-station control capability prediction scenario evaluation model for these interpolation samples; in this way, the gradient of the PV-storage-charging-station control capability prediction scenario evaluation model is forced to remain within a reasonable range, avoiding the problems of gradient explosion and training instability.

[0101] S4: Predict the control capacity of the photovoltaic storage and charging station based on the prediction model optimized by S3.

[0102] In one embodiment, the performance of the present invention is verified by selecting a photovoltaic storage and charging station of a regional power grid as an application example system;

[0103] The system includes a 400kW photovoltaic station, a 200kW electric vehicle station, and a 220kW energy storage station. The required system data include: Photovoltaic output data: including hourly photovoltaic power generation data.

[0104] Energy storage status data: the charging and discharging status data of the energy storage unit, including the charging and discharging power, charge state, etc. of the energy storage battery.

[0105] Electric vehicle charging demand data: including electric vehicle charging capacity, charging time, etc.

[0106] External environmental data: such as meteorological data and power grid electricity price data; meteorological data such as temperature, light intensity, wind speed, etc.

[0107] In order to further evaluate the superiority of the proposed method, we use the following indicators for comparison: Mean square error: measures the difference between the predicted value and the true value. Mean absolute error: measures the absolute error of the predicted result. Correlation coefficient: measures the correlation between the predicted value and the true value. In order to reflect the superiority of the present invention, the following methods are used for comparison:

[0108] M1: A method for predicting the photovoltaic storage and charging regulation capability based on traditional prediction algorithms.

[0109] M2: A method for predicting the photovoltaic storage and charging control capability based on the method proposed in the present invention.

[0110] The comparison results are shown in Table 1. From the table, it can be seen that the generative adversarial network (WGAN) method combined with the Wasserstein distance bureau of the present invention is superior to the traditional method in various evaluation indicators, especially in mean square error and correlation coefficient, which shows the high accuracy and superiority of the WGAN model in the prediction of regulation ability.

[0111] Table 1

[0112]

[0113] like Figure 1 The figure shows the learning curve of the method proposed in the present invention, wherein the horizontal axis is the number of iterations, the vertical axis is the learning benefit value, the red curve is the curve of the solar-storage-charging station regulation capability prediction scenario generation model (prediction curve), and the blue curve is the curve of the solar-storage-charging station regulation capability prediction scenario evaluation model (actual scenario curve).

[0114] It can be seen that the prediction model of the method proposed in the present invention can effectively track the actual control capability of the photovoltaic storage and charging station, and realize the coordinated prediction of the control capability of the photovoltaic storage and charging station.

[0115] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for intelligently predicting the control capability of a photovoltaic storage and charging station, characterized in that: The prediction method includes the following steps: S1. Mining the coupling relationship between the control capabilities of photovoltaic storage and charging stations based on the correlation analysis method; S2. Construct a prediction model for the control capability of photovoltaic storage and charging stations based on generative adversarial networks; S3, optimize the prediction model constructed in S2 based on Wasserstein distance; S4: Predict the control capacity of the photovoltaic storage and charging station based on the prediction model optimized by S3.

2. According to claim 1, a method for intelligently predicting the control capability of a photovoltaic storage and charging station is characterized in that: In step S1, mining the coupling relationship of the control capability of the photovoltaic storage and charging station specifically includes the following steps: S101, collecting and preprocessing historical operation data of the photovoltaic storage and charging station; S102. Use the Pearson correlation coefficient to measure the correlation between PV output and EV charging demand; S103. Predicting the response of the electric vehicle charging demand to the photovoltaic output based on the correlation between the photovoltaic output and the electric vehicle charging demand.

3. According to claim 2, a method for intelligently predicting the control capability of a photovoltaic storage and charging station is characterized in that: The correlation in step S102 is expressed as: Where PV represents photovoltaic power output, EV represents electric vehicle charging demand; cov(PV,EV) represents the covariance between photovoltaic power output and electric vehicle charging demand; σ PV and σ EV represent the standard deviation of PV output and EV charging demand, respectively; The response of the electric vehicle charging demand to the photovoltaic output in step S103 is expressed as: EV=β0+β1*PV+ε; Among them, β0 and β1 represent the response coefficients of electric vehicle charging demand to photovoltaic output; ε represents the error value of the model.

4. According to claim 1, a method for intelligently predicting the control capability of a photovoltaic storage and charging station is characterized in that: In step S2, the establishment of the prediction model specifically includes the following steps: S201. Construct a scenario generation model for predicting the control capability of photovoltaic storage and charging stations; S202. Construct a scenario assessment model for predicting the control capability of photovoltaic storage and charging stations; S203: Update the network parameters of the PV-storage-charging station control capability prediction scenario generation model and the PV-storage-charging station control capability prediction scenario evaluation model.

5. According to claim 4, a method for intelligently predicting the control capability of a photovoltaic storage and charging station is characterized in that: In step S201, the network structure of the photovoltaic storage and charging station control capability prediction scenario generation model is set, including multiple fully connected layers or convolutional layers; the goal of the photovoltaic storage and charging station control capability prediction scenario generation model is to generate actual possible operation scenarios by learning the control data of the photovoltaic storage and charging station; the photovoltaic storage and charging station control capability prediction scenario generation model receives the noise vector z and generates a predicted control scenario through a neural network; the photovoltaic storage and charging station control capability prediction scenario generation model is: Among them, G(z) represents the output of the prediction scenario generation model of the photovoltaic storage and charging station regulation capability, that is, the intelligently generated regulation capability scenario, including the regulation capability of photovoltaic, charging station, and energy storage; f G (L) represents the Lth in the prediction scenario generation model of the photovoltaic storage and charging station regulation capability G The neural network function of the layer; θ G (L) represents the Lth in the prediction scenario generation model of the photovoltaic storage and charging station regulation capability G The parameters of the layer.

6. According to claim 4, a method for intelligently predicting the control capability of a photovoltaic storage and charging station is characterized in that: In step S202, the network structure of the photovoltaic storage and charging station control capability prediction scenario evaluation model is set, including a fully connected layer or a convolutional layer; the purpose of the photovoltaic storage and charging station control capability prediction scenario evaluation model is to accurately determine whether the input data is actual data or data generated by the photovoltaic storage and charging station control capability prediction scenario generation model; the photovoltaic storage and charging station control capability prediction scenario evaluation model receives data from the photovoltaic storage and charging station control capability prediction scenario generation model and real data respectively, and outputs real probability evaluation data; the photovoltaic storage and charging station control capability prediction scenario evaluation model is: D(x)=σ(f (L) (...f (2) (f (1) (x;θ (1) );i (2) )...;i (L) )); Where D(x) represents the output of the PV-storage-charging station control capability prediction scenario evaluation model; σ is the activation function of the PV-storage-charging station control capability prediction scenario evaluation model, which is used to compress the output into the range of (0,1); f (L) represents the network function of the Lth layer in the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; x represents the input of the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; θ (L) Represents the parameters of the Lth layer in the scenario evaluation model for predicting the control capability of photovoltaic storage and charging stations.

7. According to claim 4, a method for intelligently predicting the control capability of a photovoltaic storage and charging station is characterized in that: In step S203, the network parameters are updated as follows: Among them, θ D ,θ G are the parameters of the PV-storage-charging station regulation capability prediction scenario evaluation model and the PV-storage-charging station regulation capability prediction scenario generation model, α is the learning rate, are the gradients of the loss function with respect to their respective parameters.

8. The intelligent prediction method for the control capability of a photovoltaic storage and charging station according to claim 4 is characterized in that: Step S3 specifically includes the following steps: S301. Establishing a distance model for the real data distribution and generated data distribution of the power prediction of the photovoltaic storage and charging station based on the Wasserstein distance; S302, using a weight clipping method to constrain the gradient of a scenario evaluation model for predicting the control capability of a photovoltaic storage and charging station; S303. Use gradient penalty to constrain the gradient of the photovoltaic storage and charging station regulation capability prediction scenario evaluation model.

9. The intelligent prediction method for the control capability of a photovoltaic storage and charging station according to claim 8 is characterized in that: The distance model in step S301 is specifically: Where f(x) represents the 1-Lipschitz function, ‖f‖L≤1 The Lipschitz constant of the function f(x) does not exceed 1; the sup operation represents maximizing the expected difference for all 1-Lipschitz functions; p r ,p g The data distribution of the real data of the control capability of the photovoltaic storage and charging station and the data distribution generated by the prediction model of the control capability of the photovoltaic storage and charging station are respectively; Respectively expressed in p r ,p g The mathematical expectation of a distribution. In step S302, the constraints are specifically: ‖θ C ‖ ∞ ≤c; Among them, θ C represents the weight of the prediction scenario evaluation model of the control capability of the photovoltaic storage and charging station, and c represents a constant; In step S303, the constraints are specifically: Among them, L GP It is the gradient penalty constraint term of the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; is a sample that linearly interpolates between the real data and the generated data, is the gradient of these interpolation samples by the scenario evaluation model for predicting the control capability of the photovoltaic storage and charging station; for obey The mathematical expectation of a distribution.

Images