A simulation method, device and medium for adding a new grid connection to a photovoltaic energy storage system

Abstract

The present invention discloses a simulation method, device, and medium for the addition of a photovoltaic energy storage system to the grid, relating to the technical field of electric energy storage systems. The method comprises the following steps: collecting state data dynamically changing over time from each target data source in a target photovoltaic power station and classifying it based on various electricity price levels and power restriction scenarios; extracting the time series features of the target photovoltaic power station's state data after the addition of energy storage to the grid using a base learner; dynamically adjusting the weight distribution of the target time series features under various scenarios using an attention mechanism based on the correlation between the target time series features and various electricity price levels and power restriction scenarios; using the time series features after the dynamic weight distribution as input, simulating the operation of the photovoltaic energy storage system under various electricity price levels and power restriction scenarios using a meta-learner, and outputting target scheduling parameters; and evaluating the simulation accuracy of the meta-learner based on a preset loss function and outputting the meta-learner. The present invention ensures the stability and reliability of power grid operation through precise simulation.

Description

Technical Field

[0001] The present invention relates to the technical field of electric energy storage systems, and in particular to a simulation method, device and medium for adding a new grid connection to a photovoltaic energy storage system. Background Art

[0002] With the increasing global demand for clean energy, photovoltaic power generation and energy storage technologies are playing an increasingly important role in the energy transition. However, photovoltaic power generation is characterized by significant volatility and intermittency, which poses challenges to the stable operation of the power grid. The power generated by photovoltaic systems depends not only on weather conditions but also on diurnal variations, resulting in unstable output power. This instability places an additional burden on the power grid, especially when a high proportion of renewable energy is connected to the grid. It can cause voltage fluctuations, frequency deviations, and other problems, affecting the safety and reliability of the power system.

[0003] At the same time, as a key link in balancing electricity supply and demand, the scheduling strategy of the energy storage system directly affects the economic viability of the entire project. Energy storage systems can store energy during periods of excess power and release it during periods of shortage, thereby smoothing the volatility of photovoltaic power generation. However, optimizing the scheduling strategy of energy storage systems to maximize economic benefits under different electricity price levels and power curtailment scenarios is a complex and urgent problem. Traditional simulation methods and scheduling strategies often fail to accurately predict the actual power generation of photovoltaic power plants and the optimal charging and discharging strategies of energy storage systems, resulting in an inability to effectively guarantee the robustness and stability of project investment returns.

[0004] In addition, the operating environment of photovoltaic energy storage systems is complex and changeable, involving the interaction of multiple factors, such as the spatiotemporal distribution of light resources, the technical parameters of energy storage equipment, and changes in market electricity prices. These all increase the difficulty of model construction and optimization. Summary of the Invention

[0005] In order to optimize the scheduling strategy of photovoltaic energy storage systems and ensure the stability and robustness of project investment returns, the present invention proposes a simulation method for the new grid connection of photovoltaic energy storage systems, comprising the following steps:

[0006] S1: Collect the status data of each target data source in the target PV power station that changes dynamically over time and classify it based on various electricity price levels and power restriction scenarios;

[0007] S2: Extracting the time series features of the target PV power station’s status data after the new energy storage is connected to the grid through the base learner;

[0008] S3: Based on the correlation between the target time series features and various electricity price levels and power curtailment scenarios, the attention mechanism is used to dynamically adjust the weight distribution of the target time series features in each scenario;

[0009] S4: Using the target time series features after dynamically adjusting the weight distribution as input, the meta-learner simulates the operation of the PV energy storage system under various electricity price levels and power curtailment scenarios, and outputs the target scheduling parameters.

[0010] S5: Evaluate the simulation accuracy of the meta-learner based on the preset loss function, adjust the meta-learner parameters before the simulation accuracy exceeds the preset threshold, and return to step S4. After the simulation accuracy exceeds the preset threshold, output the meta-learner with adjusted parameters.

[0011] Furthermore, in the step S1, the status data to be collected also includes energy storage capacity, installed capacity of photovoltaic power stations, spare capacity, and light resources that change dynamically over time.

[0012] Furthermore, in step S2, the base learner is based on a recursive neural network to extract the time series features corresponding to the state data of the target photovoltaic power station after the new energy storage is connected to the grid.

[0013] Furthermore, in step S3, the weight distribution is dynamically adjusted by the following formula to obtain the adjusted target time series characteristics:

[0014]

[0015] Where, is the weight of the current target time series feature at time t in the current scene, To transform from The calculated attention score is is the adjusted target time series feature, is the total duration, The value range is 1 to The constant, is the current target temporal feature at time t in the current scene, for The attention score at the moment.

[0016] Furthermore, in the step S4, the target scheduling parameters include the actual utilization hours after grid connection. and grid-connected power generation curve .

[0017] Furthermore, the actual utilization hours after grid connection and the grid-connected power generation curve Obtained through the following formula:

[0018]

[0019] Where, is the adjusted target time series feature, For To optimize the target, adjust the actual utilization hours based on real-time gradient information A meta-learner that forms a closed-loop optimization to ultimately influence the grid-connected power generation curve.

[0020] Furthermore, in the step S5, the preset loss function adopts mean square error, which is expressed as the following formula:

[0021]

[0022] Where, is the error loss, is the total duration, is the real target timing characteristic at time t in actual operation, Simulate the target time series features at time t obtained by the meta-learner simulation.

[0023] Furthermore, in step S5, the gradient descent method is used to adjust the parameters of the meta-learner.

[0024] The present invention also includes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the simulation method for adding a photovoltaic energy storage system to the grid.

[0025] Also included is a device for processing data, comprising:

[0026] a memory having a computer program stored thereon;

[0027] A processor is used to execute the computer program in the memory to implement the steps of the simulation method for adding a photovoltaic energy storage system to the grid.

[0028] Compared with the prior art, the present invention has at least the following beneficial effects:

[0029] (1) The present invention proposes a simulation method for the grid connection of new photovoltaic energy storage systems. Through this multi-level model structure, especially the application of the attention mechanism, the model can focus on the most relevant state features and suppress irrelevant information, thereby improving the accuracy of the prediction. This enables the system to accurately simulate the photovoltaic power generation and energy storage scheduling process in real time under different electricity price levels and power restriction scenarios, ensuring the stability and reliability of the grid operation.

[0030] (2) Through in-depth analysis and processing of these multi-source parameters, the model can better understand and simulate actual operating conditions and optimize the scheduling strategy of the photovoltaic energy storage system. This not only improves the economic benefits of the project, but also enhances the risk resistance in the face of various uncertainties, ensuring the stability and robustness of the project investment return rate;

[0031] (3) By smoothing the volatility of photovoltaic power generation, reducing dependence on traditional energy, and increasing the proportion of renewable energy in the energy structure, it helps promote the global transition to a low-carbon economy;

[0032] (4) Models based on deep learning have good generalization capabilities and can maintain high prediction accuracy in the face of unknown situations. By continuously verifying and optimizing model parameters, the model can be continuously improved and adapted to the needs of different environments and conditions, ensuring its long-term effectiveness and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A step-by-step diagram of a simulation method for adding a photovoltaic energy storage system to the grid. DETAILED DESCRIPTION

[0034] The following are specific embodiments of the present invention and the accompanying drawings to further describe the technical solutions of the present invention, but the present invention is not limited to these embodiments.

[0035] With the growing global demand for clean energy, photovoltaic power generation and energy storage technologies play a crucial role in the energy transition. However, the volatility and intermittent nature of photovoltaic power generation, coupled with the complexity of energy storage system scheduling strategies, pose significant challenges to the stable operation of the power grid. To address these challenges and optimize the operational efficiency of photovoltaic energy storage systems, this paper proposes a simulation method for the grid-connected addition of photovoltaic energy storage systems.

[0036] This example will elaborate on the specific implementation steps and technical points of this method, aiming to provide a complete solution for researchers, engineers and operators in related fields to achieve high-precision simulation and optimized scheduling of photovoltaic energy storage systems. Figure 1 As shown, the main technical steps of the method are as follows:

[0037] S1: Collect the status data of each target data source in the target PV power station that changes dynamically over time and classify it based on various electricity price levels and power restriction scenarios;

[0038] S2: Extracting the time series features of the target PV power station’s status data after the new energy storage is connected to the grid through the base learner;

[0039] S3: Based on the correlation between the target time series features and various electricity price levels and power curtailment scenarios, the attention mechanism is used to dynamically adjust the weight distribution of the target time series features in each scenario;

[0040] S4: Using the dynamically adjusted weighted time series features as input, the meta-learner simulates the operation of the PV energy storage system under various electricity price levels and power curtailment scenarios, and outputs the target scheduling parameters.

[0041] S5: Evaluate the simulation accuracy of the meta-learner based on the preset loss function, adjust the meta-learner parameters before the simulation accuracy exceeds the preset threshold, and return to step S4. After the simulation accuracy exceeds the preset threshold, output the meta-learner with adjusted parameters.

[0042] Specifically, in the virtual precise simulation method of photovoltaic energy storage based on the attention mechanism deep learning model, data collection and classification are the basic links of the entire process. First, through an efficient data collection system, the state data of each data source in the target photovoltaic power station that changes dynamically over time is obtained. These data include light resources, energy storage capacity, photovoltaic power station installed capacity, and backup capacity. In addition, it is also necessary to collect time series data that changes dynamically over time. ,in Represents the state data of the multi-target data source at time t, where T is the total duration.

[0043] To ensure data integrity and accuracy, the data acquisition system must be highly accurate and real-time. Specifically, light resource data can be obtained through meteorological stations or satellite remote sensing technology, energy storage capacity relies on real-time monitoring systems for energy storage equipment, and information on installed capacity and backup capacity of photovoltaic power plants can be extracted from power plant management systems. All of this data is organized into a comprehensive and detailed historical database, laying a solid foundation for subsequent modeling.

[0044] Next, the collected data is categorized based on various electricity price levels and power curtailment scenarios. Dispatch strategies vary significantly across different electricity price levels and curtailment scenarios, necessitating detailed data classification and processing. For example, during periods of high electricity prices, energy storage systems should prioritize releasing power to maximize economic benefits; during periods of low electricity prices, excess photovoltaic power generation should be stored as much as possible. Furthermore, curtailment scenarios require special consideration. When curtailment occurs on the grid, energy storage systems can serve as a buffer, avoiding power waste and maintaining stable system operation.

[0045] This refined data classification not only improves the quality of model input data but also enhances the accuracy and reliability of model predictions. By categorizing data by electricity price tier and power curtailment scenario, the model can better understand operating modes in various complex situations and generate more accurate scheduling plans accordingly. This helps improve project investment returns and enhance risk resilience in the face of market fluctuations and policy changes.

[0046] Furthermore, data classification helps optimize energy storage system dispatch strategies. By conducting in-depth analysis of data across different electricity price levels and curtailment scenarios, operators can develop optimal charging and discharging strategies, thereby reducing operating costs and improving economic efficiency. For example, releasing stored power during peak electricity price periods and storing excess power during low-price periods can not only balance grid loads but also generate higher profits.

[0047] After data collection and classification, the next step is feature extraction to capture the temporal characteristics of the PV power plant's status data after the addition of energy storage. This process utilizes a two-layer stacking model structure from deep ensemble learning. The first layer of the model is responsible for extracting valuable information from a large number of multi-source parameters through a base learner.

[0048] To effectively process time series data, this paper uses a combination of Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Simple Recurrent Neural Network (Simple RNN), and Bidirectional Recurrent Neural Network (Bi-RNN) algorithms, depending on the specific use case, as the base learners for the first-layer model. These algorithms are all variants of Recurrent Neural Networks (RNNs) and are particularly well-suited for processing time-dependent sequence data. Specifically:

[0049] LSTM effectively captures dependencies over long time spans by introducing mechanisms such as the forget gate, input gate, and output gate. The forget gate determines which information should be discarded, the input gate determines which new information should be stored, and the output gate controls the final output state. The state update equation is as follows:

[0050] Forget Gate:

[0051] Input Gate:

[0052] Status candidate values:

[0053] Status Update:

[0054] Output Gate:

[0055] Hide status update:

[0056] in, Output of the forget gate, which determines which information needs to be discarded. is the weight matrix of the forget gate, is the hidden state at time t-1, is the input data at time, is the bias term of the forget gate, Sigmoid activation function compresses the value to the (0, 1) interval; Output of the input gate, which determines what new information needs to be stored. is the weight matrix of the input gate, is the bias term of the input gate; is the new candidate state value, is the weight matrix of candidate state values, is the bias term of the candidate state value, It is a hyperbolic tangent activation function that compresses the value to the (-1, 1) interval; is the cell state at the current time t, is the cell state at time t-1, is element-wise multiplication (Hadamard product); Output gate output, controls the state of the final output, is the weight matrix of the output gate, is the bias term of the output gate; is the hidden state at time t.

[0057] GRU simplifies the LSTM structure by combining the forget gate and input gate into an update gate and introducing a reset gate to control the degree of retention of historical information. Compared to LSTM, GRU typically has fewer parameters and is more computationally efficient. The state update equation is as follows:

[0058] Update Gate:

[0059] Reset the gate:

[0060] Candidate Activation:

[0061] Hide status update:

[0062] in, To update the gate output, decide how much historical information to keep, is the weight matrix of the update gate, is the bias term of the update gate; To reset the gate output, decide how much historical information to discard, To reset the gate weight matrix, is the bias term for resetting the gate; is the new candidate hidden state, is the weight matrix of the candidate hidden state, is the bias term of the candidate hidden state; is the hidden state at time t.

[0063] During feature extraction, these base learners can extract valuable information from time series data. For example, LSTM or GRU can capture the changing trends in a photovoltaic power plant after adding energy storage, identifying patterns in power demand over different time periods and the optimal charging and discharging times for energy storage devices. This capability enables the model to generate effective base learners, providing reliable input for upper-level models.

[0064] This feature extraction not only improves the model's understanding of complex temporal dependencies but also enhances its overall predictive performance. Specifically, LSTM and GRU are better able to handle the volatility and intermittency of PV plant operation, generating more accurate state features. This is crucial for subsequent model calibration and optimization.

[0065] Furthermore, high-quality feature extraction brings significant benefits. First, it improves the model's robustness and generalization capabilities, ensuring that the model maintains high prediction accuracy even in the face of unknown situations. Second, by capturing key time series features, the model can generate more accurate scheduling plans, optimize the energy storage system's charging and discharging strategies, reduce operating costs, and improve economic benefits. Finally, the feature extraction process provides solid technical support for the entire simulation method, ensuring its reliability and effectiveness in practical applications.

[0066] After feature extraction, the next step is to dynamically adjust the weights of the target time series features for each scenario using an attention mechanism based on their correlation with various electricity price levels and curtailment scenarios. This process (step S3) aims to improve the model's prediction accuracy and better understand complex temporal dependencies.

[0067] Here, the attention mechanism allows the model to focus on the most relevant state features while suppressing irrelevant information, thereby improving prediction accuracy. Specifically, the attention mechanism is introduced in the second-layer model. By dynamically adjusting the weight distribution, the model can better handle the complex situations under different electricity price levels and power curtailment scenarios.

[0068] In the attention mechanism, we first need to calculate the attention score for each moment. These scores represent the importance of each moment and are used in the subsequent weighted sum operation. The attention score is obtained by the following formula:

[0069]

[0070] Where, is the attention score at time t, is the weight matrix for attention score calculation, is the current target temporal feature at time t in the current scene (generated by the first layer model), is the bias term for attention score calculation, As the activation function, the Softmax function is usually used to ensure that the sum of the weights at all times is 1.

[0071] In order to ensure that the sum of the weights at each moment is 1, the Softmax function is usually used to normalize the attention score:

[0072]

[0073] in, is the weight of the current target time series feature at time t in the current scene, is the attention score at time i, and T is the total duration.

[0074] Finally, the new eigenvector is obtained by weighted summation of the original eigenvectors:

[0075]

[0076] Where, is the weighted feature vector, that is, the dynamically adjusted time series feature, which serves as the input of the second-layer model.

[0077] By introducing an attention mechanism and dynamically adjusting weight distribution, the model can better capture important information at different time steps. This capability is particularly important in scenarios with fluctuating electricity prices and power curtailments. This allows the model to dynamically adjust the energy storage system's charging and discharging strategies based on real-time changes in electricity prices and demand, thereby achieving higher economic benefits. For example, releasing stored power during peak electricity prices and storing excess power during low electricity prices not only balances grid load but also generates higher profits.

[0078] After feature extraction and attention weight assignment, the next step is to use the meta-learner to simulate the operation of the PV energy storage system under various electricity price levels and power curtailment scenarios, and output target scheduling parameters. This process aims to use the optimized feature vectors to generate a specific scheduling strategy to maximize the system's economic benefits and ensure stable grid operation.

[0079] During the simulation, the meta-learner receives as input the feature vectors extracted by the first-layer model and weighted by the attention mechanism. These feature vectors contain important information at each time step, helping the meta-learner better understand complex temporal dependencies. The meta-learner can use a variety of machine learning algorithms, with common choices including linear regression, support vector machines (SVMs), and random forests. The general expression for the meta-learner is as follows:

[0080]

[0081] in, is the timing feature after dynamic adjustment, For To optimize the target, adjust the control parameters according to the real-time gradient information A meta-learner that forms a closed-loop optimization to ultimately influence the grid-connected power generation curve.

[0082] In order to better handle complex nonlinear relationships, this invention uses more advanced models such as Gradient Boosting Decision Tree (GBDT) or Neural Network. Based on these input feature vectors, the meta-learner simulates the operation of the photovoltaic energy storage system under different electricity price levels and power restriction scenarios, and generates detailed scheduling parameters, such as the actual utilization hours after grid connection. and grid-connected power generation curve .

[0083] Specifically, the actual utilization hours after grid connection Indicates the length of time a photovoltaic power station actually generates electricity in a day. This parameter can be calculated using the following formula:

[0084]

[0085] in, for The grid-connected power generation at the moment, is the maximum installed capacity of the PV power station. The grid-connected power generation curve describes the hourly power generation variation of the PV power station throughout the day. This curve can be directly predicted by the meta-learner or obtained by fitting historical data. The specific formula is as follows:

[0086]

[0087] here, is the weight matrix of the meta-learner, is the bias term of the meta-learner, in this formula is the prediction function of the meta-learner.

[0088] To ensure the prediction accuracy of the meta-learner, it needs to be rigorously evaluated and continuously optimized. A preset loss function (such as mean squared error (MSE)) is used to evaluate the performance of the meta-learner. The loss function is defined as follows:

[0089]

[0090] in, is the error loss, is the real target timing characteristic at time t in actual operation, Simulate the target time series features at time t obtained by the meta-learner simulation.

[0091] To minimize the loss function and improve model performance, the meta-learner parameters need to be adjusted. Common optimization algorithms include gradient descent, stochastic gradient descent (SGD), and the Adam optimizer. Gradient descent is a commonly used optimization algorithm that calculates the gradient of the loss function with respect to the model parameters and gradually adjusts the parameters to minimize the loss function. The specific formula is as follows:

[0092]

[0093] in, are model parameters, is the learning rate, is the gradient of the loss function with respect to the parameters.

[0094] To ensure the model maintains high prediction accuracy even in the face of unknown situations, it's necessary to regularly update the dataset and retrain the model. During model retraining, hyperparameters (such as the learning rate and batch size) can also be tuned to further improve model performance. Common tuning methods include grid search and random search.

[0095] The present invention also includes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the simulation method for adding a photovoltaic energy storage system to the grid.

[0096] Also included is a device for processing data, comprising:

[0097] a memory having a computer program stored thereon;

[0098] A processor is used to execute the computer program in the memory to implement the steps of the simulation method for adding a photovoltaic energy storage system to the grid.

[0099] In summary, the proposed simulation method for newly connected photovoltaic energy storage systems demonstrates significant benefits in multiple areas within the field of virtual precision simulation of photovoltaic energy storage. First, during the data collection and classification phase, an efficient data acquisition system captures key parameters such as light resources, energy storage capacity, installed photovoltaic power station capacity, and spare capacity. This data is then refined and classified based on electricity price levels and power curtailment scenarios, ensuring data integrity and accuracy. This not only improves the foundational quality of subsequent analysis but also provides a solid foundation for generating precise scheduling plans.

[0100] During feature extraction, the use of long short-term memory (LSTM) or gated recurrent units (GRU) as base learners effectively captures complex dependencies in time series data, thereby improving forecasting accuracy. Furthermore, the introduction of variants such as bidirectional recurrent neural networks (Bi-RNN) and deep recurrent neural networks (Deep RNN) further enhances the model's robustness and generalization capabilities, making it stable in the face of noise and outliers.

[0101] The application of the attention mechanism is a key highlight of this approach. By dynamically adjusting weight distribution, the model focuses on the most relevant state features and suppresses irrelevant information, significantly improving prediction accuracy and reliability. In particular, the attention mechanism helps optimize the charging and discharging strategies of energy storage devices under varying electricity price levels and power curtailment scenarios, maximizing the project's economic benefits and ensuring grid load balance.

[0102] The use of a meta-learner in operational simulations further improved system performance. The meta-learner receives feature vectors weighted by the attention mechanism and generates highly accurate scheduling parameters, such as actual utilization hours after grid connection and the grid-connected power generation curve. These parameters not only improve the scheduling accuracy and response speed of the PV energy storage system, but also reduce operating costs and improve investment returns through reasonable scheduling strategies.

[0103] During the model validation and optimization phase, we use loss functions such as mean squared error (MSE) to evaluate model performance, and continuously adjust model parameters using gradient descent or the Adam optimizer to minimize the difference between predicted and actual values. Regularly updating the dataset and retraining the model ensures that the model maintains high prediction accuracy even in unknown situations, enhancing its generalization and stability.

[0104] This invention significantly improves the prediction accuracy and robustness of photovoltaic energy storage systems through comprehensive data processing, high-quality feature extraction, the application of an attention mechanism, and efficient meta-learner simulation and model optimization. Its optimized scheduling strategy not only reduces operating costs and improves economic efficiency, but also enhances grid stability and promotes the efficient utilization and sustainable development of clean energy. This combination of technologies provides strong support for the optimized scheduling of photovoltaic energy storage systems and has broad practical application prospects.

[0105] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0106] In addition, in the present invention, descriptions such as "first," "second," and "one" are for descriptive purposes only and should not be understood to indicate or imply their relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0107] In the present invention, unless otherwise specified or limited, the terms "connection" and "fixation" should be understood in a broad sense. For example, "fixation" can mean fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two elements or interaction between two elements, unless otherwise specified. Those skilled in the art will be able to understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0108] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in this field can implement it. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

Claims

1. A simulation method for adding a photovoltaic energy storage system to the grid, characterized in that: Including steps: S1: Collect the status data of each target data source in the target PV power station that changes dynamically over time and classify it based on various electricity price levels and power restriction scenarios; S2: Extract the time series features of the target PV power station’s status data after the new energy storage is connected to the grid through the base learner; S3: Based on the correlation between the target time series features and various electricity price levels and power curtailment scenarios, the attention mechanism is used to dynamically adjust the weight distribution of the target time series features in each scenario; S4: Using the target time series features after dynamically adjusting the weight distribution as input, the meta-learner simulates the operation of the PV energy storage system under various electricity price levels and power curtailment scenarios, and outputs the target scheduling parameters. S5: Evaluate the simulation accuracy of the meta-learner based on the preset loss function, adjust the meta-learner parameters before the simulation accuracy exceeds the preset threshold, and return to step S4. After the simulation accuracy exceeds the preset threshold, output the meta-learner with adjusted parameters.

2. A simulation method for newly added grid connection of photovoltaic energy storage system according to claim 1, characterized in that: In the step S1, the status data to be collected also includes energy storage capacity, photovoltaic power station installed capacity, spare capacity, and light resources that change dynamically over time.

3. A simulation method for newly added grid connection of photovoltaic energy storage system according to claim 1, characterized in that: In the step S2, the base learner is based on a recursive neural network to extract the time series features corresponding to the state data of the target photovoltaic power station after the new energy storage is connected to the grid.

4. A simulation method for newly added grid connection of photovoltaic energy storage system according to claim 1, characterized in that: In step S3, the weight distribution is dynamically adjusted using the following formula to obtain the adjusted target time series features: Where, is the weight of the current target time series feature at time t in the current scene, To transform from The calculated attention score is is the adjusted target time series feature, is the total duration, The value range is 1 to The constant, is the current target temporal feature at time t in the current scene, for The attention score at the moment.

5. The simulation method for newly added grid connection of photovoltaic energy storage system according to claim 1, characterized in that: In the step S4, the target scheduling parameters include the actual utilization hours after grid connection and grid-connected power generation curve .

6. A simulation method for newly added grid connection of photovoltaic energy storage system according to claim 5, characterized in that: Actual utilization hours after grid connection and the grid-connected power generation curve Obtained through the following formula: Where, is the adjusted target time series feature, For To optimize the target, adjust the actual utilization hours based on real-time gradient information A meta-learner that forms a closed-loop optimization to ultimately influence the grid-connected power generation curve.

7. A simulation method for newly added grid connection of photovoltaic energy storage system according to claim 1, characterized in that: In the step S5, the preset loss function uses mean square error, which is expressed as the following formula: Where, is the error loss, is the total duration, is the real target timing characteristic at time t in actual operation, Simulate the target time series features at time t obtained by the meta-learner simulation.

8. The simulation method for newly added grid connection of photovoltaic energy storage system according to claim 1, characterized in that: In step S5, the gradient descent method is used to adjust the parameters of the meta-learner.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a simulation method for adding a photovoltaic energy storage system to the grid as described in any one of claims 1 to 8 are implemented.

10. A device for processing data, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of a simulation method for adding a photovoltaic energy storage system to the grid as described in any one of claims 1 to 8.

Images