Power distribution network control system and method for distributed photovoltaic grid connection

Abstract

The invention relates to the technical field of photovoltaic grid-connected control, and discloses a distributed photovoltaic grid-connected-oriented power distribution network control system and method, and the system consists of a data collection layer, a data analysis layer, an intelligent strategy generation layer and a control execution layer. The data acquisition layer monitors photovoltaic and power grid states in real time; the data analysis layer preprocesses the original data; the intelligent strategy generation layer performs mode recognition and trend prediction by using a deep learning module, and generates an optimal control strategy in combination with a strategy optimization module; and the control execution layer adjusts photovoltaic grid-connected parameters according to the instruction. According to the system, through real-time data monitoring, deep learning and strategy optimization, dynamic adaptation to the power generation characteristics of the distributed photovoltaic system is realized, the stability and power supply quality of a power distribution network are effectively improved, the requirements of stable operation and efficient scheduling of the power grid are met, and the system is suitable for popularization and application. And powerful support is provided for efficient utilization of renewable energy sources and sustainable development of a power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic grid-connected control, and in particular to a distribution network control system and method for distributed photovoltaic grid-connected. Background Art

[0002] With the transformation of the global energy structure and the rapid development of renewable energy, the penetration rate of distributed photovoltaic systems in distribution networks has been increasing as an important component of clean energy. However, the large-scale grid connection of distributed photovoltaic systems has brought new challenges to the stable operation of distribution networks. Since the power generation characteristics of photovoltaic systems are affected by environmental factors such as light and temperature, their output power has significant volatility and uncertainty, which may lead to voltage fluctuations, frequency deviations, and power factor reduction in the distribution network, thus affecting the stability of the power grid and the quality of power supply.

[0003] Current distribution network control systems are often based on fixed control strategies and rules, which are difficult to adapt to the rapid changes in the power generation characteristics of distributed photovoltaic systems. These systems lack real-time data monitoring, analysis and processing capabilities, cannot accurately predict the power generation trend of photovoltaic systems, and cannot dynamically adjust control strategies according to the actual operating status of the power grid. Therefore, in the context of the continuous expansion of the scale of distributed photovoltaic systems connected to the grid, traditional distribution network control systems have been unable to meet the needs of stable operation and efficient dispatch of the power grid. Summary of the invention

[0004] The purpose of the present invention is to provide a distribution network control system and method for distributed photovoltaic grid-connected power generation, so as to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above-mentioned object, the present invention provides the following technical solutions: a distribution network control system for distributed photovoltaic grid connection, comprising a data acquisition layer, a data analysis layer, an intelligent strategy generation layer and a control execution layer; The data acquisition layer includes a photovoltaic state monitoring module and a power grid state monitoring module; the photovoltaic state monitoring module is used to collect the operating parameters of the distributed photovoltaic system in real time, and send the operating parameters to the data analysis layer; the power grid state monitoring module is used to monitor the voltage, current, frequency and power factor of the distribution network in real time, and send the power grid parameters to the data analysis layer; The data analysis layer includes a data preprocessing module, which is responsible for processing the original data transmitted by the data acquisition layer, including data verification, outlier removal and data normalization, and the processed data is transmitted to the intelligent strategy generation layer; The intelligent strategy generation layer includes a deep learning module and a strategy optimization module; the deep learning module is used to perform pattern recognition and trend prediction based on the received data, and automatically adjust the control strategy of photovoltaic grid connection; the deep learning module includes a model construction submodule and a model application submodule, the model construction submodule uses historical data to perform model training, and the model application submodule performs prediction based on real-time data; the strategy optimization module generates the optimal control strategy based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system, and sends the control strategy to the control execution layer; The control execution layer includes a strategy analysis module and a photovoltaic grid-connected control module; the strategy analysis module is used to receive and analyze the control strategy, and convert the analyzed strategy into specific control instructions; the photovoltaic grid-connected control module is used to adjust the grid-connected parameters of the distributed photovoltaic system according to the received control instructions, and the photovoltaic grid-connected control module includes an active power control submodule and a reactive power control submodule, which are respectively used to adjust the active power output and reactive power compensation of the photovoltaic system.

[0006] Preferably, the model building submodule builds and trains a control strategy prediction model, and the specific steps include: Step 1: Collect historical operation data of distributed photovoltaic systems and distribution networks, including photovoltaic state parameters, grid state parameters and corresponding control strategies; Step 2: Preprocess the collected data, including data cleaning, data standardization and normalization, to form a standardized data set; divide the processed data set into training set, validation set and test set; Step 3: Select a deep learning model that is used to predict the optimal control strategy based on the input photovoltaic state parameters and grid state parameters; Step 4: Initialize the parameters of the deep learning model and prepare for training; Step 5: Use the training set to train the control strategy prediction model, adjust the model parameters through the optimization algorithm to improve the prediction accuracy; use the validation set to verify the model, adjust the model structure or parameters to optimize the performance; use the test set to evaluate the generalization ability of the model; Step 6: Save the trained and verified control strategy prediction model for subsequent real-time prediction of the control strategy of the distributed photovoltaic system in the intelligent strategy generation layer.

[0007] Preferably, the model building submodule uses a long short-term memory network (LSTM) model to build a control strategy prediction model.

[0008] Preferably, the structural design of the LSTM model includes an input layer, an LSTM layer, a fully connected layer and an output layer; The input layer is used to receive preprocessed photovoltaic state parameters and grid state parameters, including the voltage, current, power, temperature, irradiance of the photovoltaic array, and the voltage amplitude, voltage phase, current amplitude, current phase, frequency and power factor of the distribution network, and use these parameters as input features of the LSTM model; The LSTM layer is composed of one or more layers of LSTM units, each of which contains a forget gate, an input gate, and an output gate, which are used to process long-term dependencies in time series data and capture the dynamic characteristics of photovoltaic system and grid state parameters changing over time; the LSTM layer determines whether to retain or discard historical information through the forget gate, determines the importance of new information through the input gate, and generates the hidden state at the current moment through the output gate; The fully connected layer is connected behind the LSTM layer and is used to map the hidden state output by the LSTM layer to the predicted value of the control strategy; the fully connected layer converts the hidden state into a specific predicted value of the control strategy through a weight matrix and a bias vector, including the adjustment amount of the grid-connected voltage, grid-connected current, power factor and reactive compensation amount; The output layer is used to output the predicted value of the control strategy, and further process the predicted value generated by the fully connected layer, including rounding, limiting or normalizing.

[0009] Preferably, during the model training process, the model building submodule uses the training set to iteratively train the LSTM model, and adjusts the weights and bias parameters in the model through the back propagation algorithm and the optimizer to minimize the loss function between the predicted value and the actual value; at the same time, the model is verified using the validation set to avoid overfitting, and the generalization ability of the model is evaluated through the test set; after the training is completed, the parameters and structure of the LSTM model are saved.

[0010] Preferably, the operating parameters collected in real time by the photovoltaic status monitoring module include the voltage, current, power, temperature and irradiance of the photovoltaic array; the grid parameters monitored in real time by the grid status monitoring module include the voltage amplitude, voltage phase, current amplitude, current phase, frequency and power factor of the distribution network.

[0011] Preferably, the photovoltaic grid-connected parameters adjusted by the photovoltaic grid-connected control module include grid-connected voltage, grid-connected current, power factor and reactive power compensation amount.

[0012] Preferably, the evaluation indicators of the control strategy prediction model include root mean square error, mean absolute percentage error and accuracy.

[0013] Preferably, the data preprocessing module receives the original data transmitted by the data acquisition layer, sets a reasonable threshold range for each type of parameter, detects and eliminates abnormal data that exceeds the threshold range; after eliminating the abnormal data, the data preprocessing module performs time series analysis on the remaining data, detects and processes duplicate values ​​and missing values ​​in the data.

[0014] Preferably, a distribution network control method for distributed photovoltaic grid connection is provided, the method comprising: S1: The photovoltaic status monitoring module of the data acquisition layer collects the operating parameters of the distributed photovoltaic system in real time, including the voltage, current, power, temperature and irradiance of the photovoltaic array, and the power grid status monitoring module monitors the voltage, current, frequency and power factor of the distribution network in real time; S2: The collected photovoltaic state parameters and grid state parameters are transmitted to the data preprocessing module of the data analysis layer to perform data verification, outlier removal and data normalization to generate processed data; S3: The processed data is transmitted to the deep learning module of the intelligent strategy generation layer, wherein the model application submodule uses the trained LSTM model to perform pattern recognition and trend prediction on the received data, and automatically adjusts the control strategy of photovoltaic grid connection; the LSTM model is trained in the model construction submodule based on historical data, and can capture the dynamic characteristics of the photovoltaic system and grid state parameters changing over time, and predict the future state; S4: The strategy optimization module of the intelligent strategy generation layer generates the optimal control strategy based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system; S5: Transmitting the optimal control strategy to the strategy analysis module of the control execution layer, the strategy analysis module receives and analyzes the control strategy, and converts the analyzed strategy into specific control instructions; S6: The photovoltaic grid-connected control module of the control execution layer adjusts the grid-connected parameters of the distributed photovoltaic system according to the received control instructions, adjusts the active power output of the photovoltaic system through the active power control submodule, and adjusts the reactive power compensation of the photovoltaic system through the reactive power control submodule to achieve stable grid-connected operation of the photovoltaic system and the distribution network.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The system obtains the operating status parameters of the photovoltaic system and distribution network in real time through the data acquisition layer, and uses the data analysis layer to perform data preprocessing and pattern recognition and trend prediction of the deep learning module. It can accurately predict the power generation trend of the photovoltaic system and the future state of the power grid, providing data support for the formulation of precise control strategies.

[0016] The intelligent strategy generation layer dynamically generates the optimal control strategy based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system. This dynamic adjustment mechanism can effectively cope with the rapid changes in the power generation characteristics of the photovoltaic system and ensure the stable operation and power supply quality of the distribution network.

[0017] By controlling the real-time adjustment of photovoltaic grid-connected parameters by the execution layer, including precise control of active output and reactive compensation, the present invention can significantly reduce problems such as voltage fluctuation, frequency deviation and power factor reduction in the distribution network, improve the stability and power supply quality of the power grid, and meet users' demand for high-quality electricity. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a working principle diagram of the present invention; Figure 2 Flowchart for building a predictive model for control strategies; Figure 3 Flowchart for data preprocessing of raw data. DETAILED DESCRIPTION

[0019] 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.

[0020] See also Figure 1-3 The present invention provides a technical solution: a distribution network control system for distributed photovoltaic grid-connected power generation, including a data acquisition layer, a data analysis layer, an intelligent strategy generation layer and a control execution layer.

[0021] Data acquisition layer: This layer consists of a photovoltaic status monitoring module and a power grid status monitoring module. The photovoltaic status monitoring module collects the operating parameters of the distributed photovoltaic system in real time through sensors and monitoring equipment, such as voltage, current, power, temperature, and irradiance, and sends these parameters to the data analysis layer. The power grid status monitoring module is responsible for real-time monitoring of key parameters such as voltage, current, frequency, and power factor of the distribution network, and also sends these data to the data analysis layer.

[0022] Data analysis layer: This layer contains the data preprocessing module, which receives the raw data transmitted by the data acquisition layer and performs data verification, outlier removal and data normalization to ensure the accuracy and consistency of the data. The processed data is transmitted to the intelligent strategy generation layer for further analysis and prediction.

[0023] Intelligent strategy generation layer: This layer consists of a deep learning module and a strategy optimization module. The deep learning module contains a model building submodule and a model application submodule. The model building submodule uses historical data (including photovoltaic system operation data and distribution network status data) to train the LSTM model. The LSTM model can capture long-term dependencies in time series data and is suitable for predicting the power generation characteristics of photovoltaic systems. The model application submodule uses the trained LSTM model to perform pattern recognition and trend prediction based on real-time data, and automatically adjusts the control strategy for photovoltaic grid connection. The strategy optimization module generates the optimal control strategy based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system, and sends these strategies to the control execution layer.

[0024] Control execution layer: This layer includes the strategy analysis module and the photovoltaic grid-connected control module. The strategy analysis module receives and analyzes the control strategy from the intelligent strategy generation layer and converts it into specific control instructions. The photovoltaic grid-connected control module adjusts the grid-connected parameters of the distributed photovoltaic system according to these control instructions. This module includes an active power control submodule and a reactive power control submodule, which are responsible for adjusting the active power output and reactive power compensation of the photovoltaic system, respectively, to achieve stable grid-connected operation of the photovoltaic system and the distribution network.

[0025] In the specific implementation, data is first collected in real time through the photovoltaic status monitoring module and the power grid status monitoring module of the data acquisition layer, and the data is cleaned and preprocessed through the data preprocessing module of the data analysis layer. Then, these data are input into the deep learning module of the intelligent strategy generation layer for pattern recognition and trend prediction. After the LSTM model is trained in the model construction submodule, it can make predictions based on real-time data in the model application submodule and automatically adjust the photovoltaic grid-connected control strategy. The strategy optimization module generates the optimal control strategy based on these prediction results, and implements specific control operations through the strategy analysis module and photovoltaic grid-connected control module of the control execution layer.

[0026] The present invention will be further described below in conjunction with Examples 1 to 3: Embodiment 1: The present invention provides a specific implementation method of a model building submodule in a distribution network control system for distributed photovoltaic grid-connected power grids. The submodule is responsible for building and training a control strategy prediction model based on a long short-term memory network (LSTM). The following are the detailed implementation steps and structural design: 1. Data Collection and Preprocessing Step 1: Collect historical operating data of distributed photovoltaic systems and distribution networks. These data include photovoltaic state parameters (such as voltage, current, power, temperature, and irradiance of the photovoltaic array) and grid state parameters (such as voltage amplitude, voltage phase, current amplitude, current phase, frequency, and power factor of the distribution network), as well as corresponding control strategies (such as grid-connected voltage, grid-connected current, power factor, and adjustment of reactive power compensation).

[0027] Step 2: Preprocess the collected data. First, clean the data to remove outliers and missing values; then standardize and normalize the data to convert parameters of different dimensions and value ranges to a unified scale to eliminate the impact of data bias on model training. The processed data set is divided into training set, validation set and test set for model training, validation and evaluation.

[0028] 2. Model selection and structural design Step 3: Select the LSTM model as the control strategy prediction model. The LSTM model is suitable for processing time series data and can capture the dynamic characteristics of the photovoltaic system and grid state parameters changing over time.

[0029] Step 4: Design the structure of the LSTM model. The model includes input layer, LSTM layer, fully connected layer and output layer.

[0030] Input layer: receives the preprocessed photovoltaic state parameters and grid state parameters as input features of the LSTM model.

[0031] LSTM layer: It consists of one or more layers of LSTM units. Each LSTM unit contains a forget gate, an input gate, and an output gate to process long-term dependencies in time series data. The forget gate determines whether to retain or discard historical information, the input gate determines the importance of new information, and the output gate generates the hidden state at the current moment.

[0032] Fully connected layer: Connected to the back of the LSTM layer, it maps the hidden state output by the LSTM layer to the predicted value of the control strategy. Through the weight matrix and bias vector, the hidden state is converted into a specific control strategy predicted value.

[0033] Output layer: Outputs the predicted value of the control strategy and performs further processing, such as rounding, limiting or normalization, to meet actual control requirements.

[0034] 3. Model training and validation Step 5: Use the training set to iteratively train the LSTM model. Through the back-propagation algorithm and optimizer (such as Adam, SGD, etc.), adjust the weight and bias parameters in the model to minimize the loss function (such as mean square error, cross entropy, etc.) between the predicted value and the actual value.

[0035] During the training process, the validation set is used to verify the model to avoid overfitting. By monitoring the performance indicators (such as accuracy, loss, etc.) on the validation set, the model structure (such as increasing or decreasing the number of LSTM layers, adjusting the number of neurons, etc.) or parameters (such as learning rate, batch size, etc.) are adjusted to optimize performance.

[0036] Use the test set to evaluate the generalization ability of the model and ensure that the model can perform well on unseen data.

[0037] 4. Model preservation and application Step 6: After training and verification, save the parameters and structure of the LSTM model. These parameters and structures will be loaded into the model application submodule of the intelligent strategy generation layer for subsequent real-time prediction of the control strategy of the distributed photovoltaic system.

[0038] In actual applications, the model application submodule receives the real-time collected photovoltaic state parameters and grid state parameters and inputs them into the LSTM model for prediction. Based on the prediction results, the intelligent strategy generation layer generates the optimal control strategy and implements specific control operations through the control execution layer.

[0039] Embodiment 2: In order to comprehensively evaluate the performance of the control strategy prediction model, the present invention adopts the following three evaluation indicators: Root Mean Square Error (RMSE): It is used to measure the square root of the average of the sum of squares of the differences between the predicted values ​​and the actual values. The smaller the RMSE, the higher the prediction accuracy of the model.

[0040] Mean absolute percentage error (MAPE): Calculates the average value of the absolute value of the difference between the predicted value and the actual value as a percentage of the actual value. MAPE can intuitively reflect the relative error of the predicted value. The smaller its value, the better the prediction effect of the model.

[0041] Accuracy: In classification problems, accuracy refers to the ratio of correctly predicted samples to the total number of samples. For control strategy prediction models, although the main focus is on regression problems, the prediction results can be converted into classification problems by setting a certain error threshold, thereby calculating the accuracy. For example, samples whose difference between the predicted value and the actual value is less than a certain threshold can be considered to be correctly predicted.

[0042] The data preprocessing module is an important part of the model building submodule. Its main task is to clean and normalize the raw data transmitted by the data acquisition layer to ensure the quality and consistency of the data. The specific implementation method is as follows: Abnormal data detection and elimination: The data preprocessing module first sets a reasonable threshold range for each type of parameter. These threshold ranges are determined based on the actual operation of the photovoltaic system and the power grid and empirical data. Then, the module traverses the raw data, detects and eliminates abnormal data that exceeds the threshold range. For example, for the voltage parameters of the photovoltaic array, a reasonable voltage range (such as 0-500V) can be set, and any data outside this range will be regarded as abnormal data and eliminated.

[0043] Duplicate and missing value processing: After removing abnormal data, the data preprocessing module will perform time series analysis on the remaining data to detect and process duplicate and missing values ​​in the data. For duplicate values, the module will only retain one record to avoid the impact of data redundancy on model training. For missing values, the module will use interpolation or mean filling methods to fill in the missing values ​​to ensure the integrity and continuity of the data. For example, for missing data at a certain time point, the average value of the data at the two previous and next time points can be used to fill in the missing data.

[0044] Embodiment 3: The present invention also includes a distribution network control method for distributed photovoltaic grid-connected power generation. The method realizes intelligent adjustment of distributed photovoltaic system grid-connected parameters through hierarchical module design such as data collection, data analysis, intelligent strategy generation and control execution, ensuring stable grid-connected operation of photovoltaic system and distribution network. The following are the detailed implementation steps: S1: Data Collection Photovoltaic status monitoring module: real-time collection of key operating parameters of distributed photovoltaic systems, including but not limited to the voltage, current, power, temperature and irradiance of the photovoltaic array. These parameters are measured by high-precision sensors and converted into digital signals for subsequent processing.

[0045] Grid status monitoring module: Synchronously monitors the voltage, current, frequency, power factor and other status parameters of the distribution network. These parameters reflect the real-time operating status of the distribution network and are crucial for the subsequent generation of control strategies.

[0046] S2: Data preprocessing Data verification: After receiving the photovoltaic state parameters and grid state parameters from the data acquisition layer, the data preprocessing module first performs data verification to ensure the integrity and accuracy of the data.

[0047] Outlier removal: By setting a reasonable threshold range, detect and remove data points that exceed the range to eliminate the impact of abnormal data on subsequent analysis.

[0048] Data normalization: Normalize parameters of different dimensions and value ranges to convert them to a unified scale to facilitate subsequent model training and prediction.

[0049] S3: Deep Learning Model Prediction Model application submodule: Load the trained LSTM model, which is trained based on a large amount of historical data in the model construction submodule. The LSTM model can capture the dynamic characteristics of the PV system and grid state parameters changing over time and predict future states.

[0050] Pattern recognition and trend prediction: The processed data is input into the LSTM model, which performs pattern recognition and trend prediction and outputs the prediction results of the future photovoltaic system and power grid status.

[0051] S4: Strategy Optimization Strategy optimization module: Based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system, the strategy is optimized. Factors to be considered include but are not limited to the voltage stability, frequency stability, power factor control of the power grid, and the power generation efficiency and power regulation capability of the photovoltaic system.

[0052] Generate optimal control strategy: Generate the optimal control strategy through algorithm optimization, which aims to achieve the best match between the photovoltaic system and the distribution network to ensure stable grid-connected operation.

[0053] S5: Strategy Analysis Strategy parsing module: receives the optimal control strategy from the strategy optimization module and parses it. The parsing process includes converting the abstract control strategy into specific control instructions, which can be directly executed by the photovoltaic grid-connected control module.

[0054] S6: Control Execution Photovoltaic grid-connected control module: adjusts the grid-connected parameters of the distributed photovoltaic system according to the received control instructions. This includes adjusting the active power output of the photovoltaic system through the active power control submodule and adjusting the reactive power compensation of the photovoltaic system through the reactive power control submodule.

[0055] Achieve stable grid-connected operation: Through precise control command execution, ensure that parameters such as voltage, current, frequency and power factor between the photovoltaic system and the distribution network are kept within the specified range to achieve stable grid-connected operation.

[0056] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0057] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The distribution network control system for distributed photovoltaic grid connection is characterized by: It includes data collection layer, data analysis layer, intelligent strategy generation layer and control execution layer; The data acquisition layer includes a photovoltaic state monitoring module and a power grid state monitoring module; the photovoltaic state monitoring module is used to collect the operating parameters of the distributed photovoltaic system in real time, and send the operating parameters to the data analysis layer; the power grid state monitoring module is used to monitor the voltage, current, frequency and power factor of the distribution network in real time, and send the power grid parameters to the data analysis layer; The data analysis layer includes a data preprocessing module, which is responsible for processing the original data transmitted by the data acquisition layer, including data verification, outlier removal and data normalization, and the processed data is transmitted to the intelligent strategy generation layer; The intelligent strategy generation layer includes a deep learning module and a strategy optimization module; the deep learning module is used to perform pattern recognition and trend prediction based on the received data, and automatically adjust the control strategy of photovoltaic grid connection; the deep learning module includes a model construction submodule and a model application submodule, the model construction submodule uses historical data to perform model training, and the model application submodule performs prediction based on real-time data; The strategy optimization module generates an optimal control strategy based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system, and sends the control strategy to the control execution layer; The control execution layer includes a strategy analysis module and a photovoltaic grid-connected control module; the strategy analysis module is used to receive and analyze the control strategy, and convert the analyzed strategy into specific control instructions; the photovoltaic grid-connected control module is used to adjust the grid-connected parameters of the distributed photovoltaic system according to the received control instructions, and the photovoltaic grid-connected control module includes an active power control submodule and a reactive power control submodule, which are respectively used to adjust the active power output and reactive power compensation of the photovoltaic system.

2. The distribution network control system for distributed photovoltaic grid connection according to claim 1, characterized in that: The model building submodule builds and trains a control strategy prediction model, and the specific steps include: Step 1: Collect historical operation data of distributed photovoltaic systems and distribution networks, including photovoltaic state parameters, grid state parameters and corresponding control strategies; Step 2: Preprocess the collected data, including data cleaning, data standardization and normalization, to form a standardized data set; divide the processed data set into training set, validation set and test set; Step 3: Select a deep learning model that is used to predict the optimal control strategy based on the input photovoltaic state parameters and grid state parameters; Step 4: Initialize the parameters of the deep learning model and prepare for training; Step 5: Use the training set to train the control strategy prediction model, adjust the model parameters through the optimization algorithm to improve the prediction accuracy; use the validation set to verify the model, adjust the model structure or parameters to optimize the performance; use the test set to evaluate the generalization ability of the model; Step 6: Save the trained and verified control strategy prediction model for subsequent real-time prediction of the control strategy of the distributed photovoltaic system in the intelligent strategy generation layer.

3. The distribution network control system for distributed photovoltaic grid connection according to claim 2, characterized in that: The model building submodule uses a long short-term memory network (LSTM) model to build a control strategy prediction model.

4. The distribution network control system for distributed photovoltaic grid connection according to claim 3, characterized in that: The structural design of the LSTM model includes an input layer, an LSTM layer, a fully connected layer and an output layer; The input layer is used to receive preprocessed photovoltaic state parameters and grid state parameters, including the voltage, current, power, temperature, irradiance of the photovoltaic array, and the voltage amplitude, voltage phase, current amplitude, current phase, frequency and power factor of the distribution network, and use these parameters as input features of the LSTM model; The LSTM layer is composed of one or more layers of LSTM units, each of which contains a forget gate, an input gate, and an output gate, which are used to process long-term dependencies in time series data and capture the dynamic characteristics of photovoltaic system and grid state parameters changing over time; the LSTM layer determines whether to retain or discard historical information through the forget gate, determines the importance of new information through the input gate, and generates the hidden state at the current moment through the output gate; The fully connected layer is connected behind the LSTM layer and is used to map the hidden state output by the LSTM layer to the predicted value of the control strategy; The fully connected layer converts the hidden state into specific control strategy prediction values ​​through the weight matrix and bias vector, including the adjustment of grid-connected voltage, grid-connected current, power factor and reactive compensation; The output layer is used to output the predicted value of the control strategy, and further process the predicted value generated by the fully connected layer, including rounding, limiting or normalizing.

5. The distribution network control system for distributed photovoltaic grid connection according to claim 4 is characterized in that: During the model training process, the model building submodule uses the training set to iteratively train the LSTM model, and adjusts the weight and bias parameters in the model through the back propagation algorithm and optimizer to minimize the loss function between the predicted value and the actual value; at the same time, the model is verified using the validation set to avoid overfitting, and the generalization ability of the model is evaluated through the test set; After training is completed, save the parameters and structure of the LSTM model.

6. The distribution network control system for distributed photovoltaic grid connection according to claim 1, characterized in that: The operating parameters collected in real time by the photovoltaic status monitoring module include the voltage, current, power, temperature and irradiance of the photovoltaic array; the grid parameters monitored in real time by the grid status monitoring module include the voltage amplitude, voltage phase, current amplitude, current phase, frequency and power factor of the distribution network.

7. The distribution network control system for distributed photovoltaic grid connection according to claim 1, characterized in that: The photovoltaic grid-connected parameters adjusted by the photovoltaic grid-connected control module include grid-connected voltage, grid-connected current, power factor and reactive power compensation amount.

8. The distribution network control system for distributed photovoltaic grid connection according to claim 5, characterized in that: The evaluation indicators of the control strategy prediction model include root mean square error, mean absolute percentage error and accuracy.

9. The distribution network control system for distributed photovoltaic grid connection according to claim 1, characterized in that: The data preprocessing module receives the original data transmitted by the data acquisition layer, sets a reasonable threshold range for each type of parameter, detects and removes abnormal data that exceeds the threshold range; after removing the abnormal data, the data preprocessing module performs time series analysis on the remaining data, detects and processes duplicate values ​​and missing values ​​in the data.

10. A distribution network control method for distributed photovoltaic grid connection, characterized in that: The method comprises: S1: The photovoltaic status monitoring module of the data acquisition layer collects the operating parameters of the distributed photovoltaic system in real time, including the voltage, current, power, temperature and irradiance of the photovoltaic array, and the power grid status monitoring module monitors the voltage, current, frequency and power factor of the distribution network in real time; S2: The collected photovoltaic state parameters and grid state parameters are transmitted to the data preprocessing module of the data analysis layer to perform data verification, outlier removal and data normalization to generate processed data; S3: The processed data is transmitted to the deep learning module of the intelligent strategy generation layer, wherein the model application submodule uses the trained LSTM model to perform pattern recognition and trend prediction on the received data, and automatically adjusts the control strategy of photovoltaic grid connection; the LSTM model is trained in the model construction submodule based on historical data, and can capture the dynamic characteristics of the photovoltaic system and grid state parameters changing over time, and predict the future state; S4: The strategy optimization module of the intelligent strategy generation layer generates the optimal control strategy based on the prediction results of the deep learning module, combined with the requirements for stable operation of the power grid and the power generation characteristics of the photovoltaic system; S5: Transmitting the optimal control strategy to the strategy analysis module of the control execution layer, the strategy analysis module receives and analyzes the control strategy, and converts the analyzed strategy into specific control instructions; S6: The photovoltaic grid-connected control module of the control execution layer adjusts the grid-connected parameters of the distributed photovoltaic system according to the received control instructions, adjusts the active power output of the photovoltaic system through the active power control submodule, and adjusts the reactive power compensation of the photovoltaic system through the reactive power control submodule to achieve stable grid-connected operation of the photovoltaic system and the distribution network.

Images