Distributed photovoltaic energy storage method and system based on deep learning model

Abstract

The invention provides a distributed photovoltaic energy storage method and system based on a deep learning model, and the system comprises a data collection and preprocessing module, a deep learning model construction module, a model training and optimization module, an energy storage control strategy module, a cooperative work and optimization module, and a system monitoring and feedback module. A deep learning model can accurately predict photovoltaic power generation capacity and load requirements, realizes optimized charge and discharge control of energy storage equipment, maximally utilizes photovoltaic energy and reduces energy waste, and meanwhile, the deep learning model can quickly learn and adapt to a new operation scene and a data mode, so that the energy storage efficiency is improved. The system can cope with various complex and changeable environments and load conditions, and through accurate prediction and an intelligent control strategy, the fluctuation of photovoltaic power generation is effectively stabilized, the stable power supply of the micro-grid under different working conditions is ensured, and the fluctuation of voltage and frequency is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage and microgrid, and in particular to a distributed photovoltaic energy storage method and system based on a deep learning model. Background Art

[0002] As the global demand for clean energy continues to grow, distributed photovoltaic power generation is increasingly used in the energy field. However, photovoltaic power generation is intermittent and unstable, and its power generation is affected by factors such as weather, season and time, which brings challenges to the stable operation of the power system.

[0003] To solve this problem, distributed photovoltaic energy storage technology came into being. Traditional distributed photovoltaic energy storage methods usually rely on simple rules or experience-based strategies to store and release energy. These methods often show problems of inefficiency and lack of adaptability when dealing with complex and changeable energy supply and demand situations.

[0004] With the rapid development of information technology, deep learning technology has demonstrated powerful prediction and optimization capabilities in various fields. However, in the field of distributed photovoltaic energy storage, the application of deep learning models is still in a relatively early stage. Some existing attempts have problems such as low data quality, unreasonable model architecture, lack of effective training methods, and insufficient integration with actual systems, resulting in the fact that the effect in actual application has not met expectations.

[0005] In addition, the existing distributed photovoltaic energy storage systems still have the following deficiencies in energy management and optimization: it is impossible to accurately predict changes in photovoltaic power generation and load demand, and it is difficult to achieve optimal charging and discharging control of energy storage equipment, resulting in energy waste and system instability; the ability of different energy storage devices to work together is weak, affecting the overall energy storage effect; the system has limited adaptability to new operating scenarios and emergencies, and lacks a flexible adjustment mechanism.

[0006] Therefore, there is an urgent need for an innovative distributed photovoltaic energy storage method and system based on deep learning models to improve energy utilization efficiency, enhance system stability and adaptability, and meet the growing energy demand and higher grid operation requirements. Summary of the invention

[0007] In order to overcome the existing problems, the embodiments of the present application provide a distributed photovoltaic energy storage method and system based on a deep learning model. The deep learning model can accurately predict photovoltaic power generation and load demand, realize optimized charging and discharging control of energy storage equipment, maximize the use of photovoltaic energy, and reduce energy waste. At the same time, the deep learning model can quickly learn and adapt to new operating scenarios and data patterns, so that the system can cope with various complex and changeable environments and load conditions. Through accurate prediction and intelligent control strategies, it can effectively smooth out the fluctuations of photovoltaic power generation, ensure stable power supply of microgrids under different working conditions, and reduce voltage and frequency fluctuations.

[0008] The technical solution adopted by the embodiment of the present application to solve the technical problem is:

[0009] A distributed photovoltaic energy storage method and system based on a deep learning model, the specific steps of the distributed photovoltaic energy storage method based on the deep learning model are as follows:

[0010] Step 1: Use data collection and preprocessing technology to collect and process relevant data of photovoltaic panels, energy storage equipment and load ends to eliminate noise and outliers in the data and normalize the data. Data collection and preprocessing technology includes data cleaning, missing value filling and outlier processing;

[0011] Step 2: Build a deep learning model architecture that combines a multi-layer convolutional neural network and a long short-term memory network to accurately predict photovoltaic power generation and load demand in the future. In the deep learning model architecture, the convolutional neural network is used to extract local features in the data, and the long short-term memory network is used to capture long-term dependencies in the time series.

[0012] Step 3: Use the adaptive momentum estimation optimization algorithm and introduce regularization technology to train the model, and establish a real-time feedback mechanism to continuously optimize the model based on the actual system operation data;

[0013] Step 4: Based on the prediction results of the deep learning model, formulate a charging and discharging strategy for the energy storage device that takes into account the capacity limitation and charging and discharging efficiency factors of the energy storage device. The charging and discharging strategy for the energy storage device controls the charging of the energy storage device when it is predicted that the photovoltaic power generation is in excess, and controls the discharging of the energy storage device when it is predicted that the photovoltaic power generation is insufficient and the load demand is large.

[0014] Step 5: For multiple distributed energy storage devices, the central controller is used for collaborative management to achieve reasonable allocation and optimal utilization of energy. This also includes real-time monitoring of the system's operating status and external environmental changes, and dynamic adjustment of energy storage strategies to adapt to different operating scenarios and emergencies;

[0015] Step 6: Compare the actual operating data with the model prediction results. When the deviation exceeds the set threshold, the model is triggered to retrain and optimize to ensure the accuracy and stability of the system. The retraining and optimization of the model includes establishing a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads.

[0016] It includes data acquisition and preprocessing module, deep learning model building module, model training and optimization module, energy storage control strategy module, collaborative work and optimization module, and system monitoring and feedback module;

[0017] The data acquisition and preprocessing module collects power generation, energy storage status, and load demand data in real time through sensors installed on photovoltaic panels, energy storage devices, and load ends. These data are transmitted to the data processing center and undergo data cleaning, missing value filling, and outlier processing preprocessing operations to provide a high-quality data foundation for subsequent model training and prediction.

[0018] The deep learning model building module adopts an architecture that combines a multi-layer convolutional neural network (CNN) and a long short-term memory network (LSTM), where CNN is used to extract local features in the data and LSTM is used to capture long-term dependencies in the time series. The input of the model includes historical photovoltaic power generation, weather data, and load demand, and the output is the photovoltaic power generation and load forecast value in the future period.

[0019] The model training and optimization module uses the preprocessed data for model training, adopts the adaptive momentum estimation (Adam) optimization algorithm, automatically adjusts the learning rate according to the change of the loss function during the training process, and introduces the Dropout regularization technology to randomly shut down some neurons to prevent the model from overfitting. In addition, during the training process, the performance of the model on the validation set is continuously monitored. When the performance is no longer improved, the training is stopped in advance to avoid overtraining;

[0020] The energy storage control strategy module formulates a charging and discharging strategy for the energy storage device based on the prediction results of the deep learning model. When it is predicted that the photovoltaic power generation is in excess, the energy storage device is controlled to charge; when it is predicted that the photovoltaic power generation is insufficient and the load demand is large, the energy storage device is controlled to discharge. At the same time, the capacity limitation and charging and discharging efficiency factors of the energy storage device are considered to achieve optimal energy management.

[0021] The collaborative work and optimization module manages multiple distributed energy storage devices through a central controller, reasonably allocates charging and discharging tasks according to the location, capacity and operating status of each device, and improves the overall energy storage effect. In addition, by real-time monitoring of the system's operating status and external environmental changes, the energy storage strategy is dynamically adjusted to adapt to different operating scenarios and emergencies;

[0022] The system monitoring and feedback module establishes a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads, compares the actual operating data with the model prediction results, and triggers retraining and optimization of the model when the deviation exceeds a certain threshold to ensure the accuracy and stability of the system.

[0023] The advantages of the embodiments of the present application are:

[0024] 1. The deep learning model can accurately predict photovoltaic power generation and load demand, realize the optimized charging and discharging control of energy storage equipment, maximize the use of photovoltaic energy, and reduce energy waste. At the same time, the deep learning model can quickly learn and adapt to new operating scenarios and data patterns, enabling the system to cope with various complex and changeable environments and load conditions. Through accurate prediction and intelligent control strategies, it can effectively smooth out the fluctuations of photovoltaic power generation, ensure the stable power supply of microgrids under different working conditions, and reduce voltage and frequency fluctuations.

[0025] 2. Through deep learning models, it can quickly learn and adapt to new operating scenarios and data patterns, cope with complex and changing environments and load conditions, optimize the balance of electricity supply and demand, better match supply and demand, reduce peak-to-valley differences, improve grid operation efficiency, enhance system adaptability, achieve efficient energy utilization, reduce operation and maintenance costs, and extend equipment life, bringing significant economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 in 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. In addition, for the convenience of description below, the referenced "upper", "lower", "left", and "right" are the same as the upper, lower, left, and right directions of the drawings themselves. The "first", "second", etc. below are distinguished for the description and have no other special meanings.

[0028] The embodiments of the present application solve the problems in the prior art by providing a distributed photovoltaic energy storage method and system based on a deep learning model. The deep learning model can accurately predict photovoltaic power generation and load demand, realize optimized charging and discharging control of energy storage equipment, maximize the use of photovoltaic energy, and reduce energy waste. At the same time, the deep learning model can quickly learn and adapt to new operating scenarios and data patterns, so that the system can cope with various complex and changeable environments and load conditions. Through accurate prediction and intelligent control strategies, it can effectively smooth out the fluctuations of photovoltaic power generation, ensure the stable power supply of microgrids under different working conditions, and reduce voltage and frequency fluctuations. The deep learning model can quickly learn and adapt to new operating scenarios and data patterns, cope with complex and changeable environments and load conditions, optimize the balance of power supply and demand, better match supply and demand, reduce peak-to-valley differences, improve power grid operation efficiency, enhance system adaptability, achieve efficient energy utilization, reduce operation and maintenance costs, and extend equipment life, thereby bringing significant economic benefits.

[0029] The technical solution in the embodiment of the present application is to solve the above problems, and the overall idea is as follows:

[0030] Example 1

[0031] This embodiment provides a distributed photovoltaic energy storage method and system based on a deep learning model. Figure 1 As shown, the specific steps of the distributed photovoltaic energy storage method of the deep learning model are as follows:

[0032] Step 1: Use data collection and preprocessing technology to collect and process relevant data of photovoltaic panels, energy storage equipment and load ends to eliminate noise and outliers in the data and normalize the data. Data collection and preprocessing technology includes data cleaning, missing value filling and outlier processing;

[0033] Step 2: Build a deep learning model architecture that combines a multi-layer convolutional neural network and a long short-term memory network to accurately predict photovoltaic power generation and load demand in the future. In the deep learning model architecture, the convolutional neural network is used to extract local features in the data, and the long short-term memory network is used to capture long-term dependencies in the time series.

[0034] Step 3: Use the adaptive momentum estimation optimization algorithm and introduce regularization technology to train the model, and establish a real-time feedback mechanism to continuously optimize the model based on the actual system operation data;

[0035] Step 4: Based on the prediction results of the deep learning model, formulate a charging and discharging strategy for the energy storage device that takes into account the capacity limitation and charging and discharging efficiency factors of the energy storage device. The charging and discharging strategy for the energy storage device controls the charging of the energy storage device when it is predicted that the photovoltaic power generation is in excess, and controls the discharging of the energy storage device when it is predicted that the photovoltaic power generation is insufficient and the load demand is large.

[0036] Step 5: For multiple distributed energy storage devices, the central controller is used for collaborative management to achieve reasonable allocation and optimal utilization of energy. This also includes real-time monitoring of the system's operating status and external environmental changes, and dynamic adjustment of energy storage strategies to adapt to different operating scenarios and emergencies;

[0037] Step 6: Compare the actual operating data with the model prediction results. When the deviation exceeds the set threshold, the model is triggered to retrain and optimize to ensure the accuracy and stability of the system. The retraining and optimization of the model includes establishing a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads.

[0038] It includes data acquisition and preprocessing module, deep learning model building module, model training and optimization module, energy storage control strategy module, collaborative work and optimization module, and system monitoring and feedback module;

[0039] The data acquisition and preprocessing module collects data on power generation, energy storage status, and load demand in real time through sensors installed on photovoltaic panels, energy storage devices, and load ends. These data are transmitted to the data processing center and undergo data cleaning, missing value filling, and outlier processing preprocessing operations to provide a high-quality data foundation for subsequent model training and prediction.

[0040] The deep learning model building module adopts an architecture that combines multi-layer convolutional neural networks (CNN) and long short-term memory networks (LSTM). CNN is used to extract local features in the data, and LSTM is used to capture long-term dependencies in time series. The model input includes historical photovoltaic power generation, weather data, and load demand, and the output is the photovoltaic power generation and load forecast values ​​for a period of time in the future.

[0041] The model training and optimization module uses preprocessed data for model training and adopts the adaptive momentum estimation (Adam) optimization algorithm to automatically adjust the learning rate according to the change of the loss function during the training process. At the same time, the Dropout regularization technology is introduced to randomly close some neurons to prevent the model from overfitting. In addition, during the training process, the performance of the model on the validation set is continuously monitored. When the performance no longer improves, the training is stopped in advance to avoid overtraining.

[0042] The energy storage control strategy module formulates the charging and discharging strategy of the energy storage equipment according to the prediction results of the deep learning model. When it is predicted that the photovoltaic power generation is in excess, the energy storage equipment is controlled to charge; when it is predicted that the photovoltaic power generation is insufficient and the load demand is large, the energy storage equipment is controlled to discharge. At the same time, the capacity limitation and charging and discharging efficiency factors of the energy storage equipment are considered to achieve optimal energy management.

[0043] The collaborative work and optimization module manages multiple distributed energy storage devices through a central controller, reasonably allocates charging and discharging tasks according to the location, capacity and operating status of each device, and improves the overall energy storage effect. In addition, by real-time monitoring of the system's operating status and external environmental changes, it dynamically adjusts the energy storage strategy to adapt to different operating scenarios and emergencies;

[0044] The system monitoring and feedback module establishes a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads, and compares the actual operating data with the model prediction results. When the deviation exceeds a certain threshold, it triggers the retraining and optimization of the model to ensure the accuracy and stability of the system.

[0045] By adopting the above technical solution:

[0046] In terms of data processing, advanced data cleaning and preprocessing technologies have been introduced to improve data quality. By screening, repairing and normalizing the original data, noise and outliers in the data are eliminated to provide more accurate and reliable input for deep learning models.

[0047] In order to strengthen the integration of the model with the actual system, a real-time feedback mechanism was established. The model is continuously adjusted and optimized according to the operating data of the actual system to ensure that it can accurately reflect the real state of the system and provide precise decision-making support for energy storage control.

[0048] Example 2

[0049] This embodiment provides a distributed photovoltaic energy storage method and system based on a deep learning model. Figure 1 As shown, the specific steps of the distributed photovoltaic energy storage method of the deep learning model are as follows:

[0050] Step 1: Use data collection and preprocessing technology to collect and process relevant data of photovoltaic panels, energy storage equipment and load ends to eliminate noise and outliers in the data and normalize the data. Data collection and preprocessing technology includes data cleaning, missing value filling and outlier processing;

[0051] Step 2: Build a deep learning model architecture that combines a multi-layer convolutional neural network and a long short-term memory network to accurately predict photovoltaic power generation and load demand in the future. In the deep learning model architecture, the convolutional neural network is used to extract local features in the data, and the long short-term memory network is used to capture long-term dependencies in the time series.

[0052] Step 3: Use the adaptive momentum estimation optimization algorithm and introduce regularization technology to train the model, and establish a real-time feedback mechanism to continuously optimize the model based on the actual system operation data;

[0053] Step 4: Based on the prediction results of the deep learning model, formulate a charging and discharging strategy for the energy storage device that takes into account the capacity limitation and charging and discharging efficiency factors of the energy storage device. The charging and discharging strategy for the energy storage device controls the charging of the energy storage device when it is predicted that the photovoltaic power generation is in excess, and controls the discharging of the energy storage device when it is predicted that the photovoltaic power generation is insufficient and the load demand is large.

[0054] Step 5: For multiple distributed energy storage devices, the central controller is used for collaborative management to achieve reasonable allocation and optimal utilization of energy. This also includes real-time monitoring of the system's operating status and external environmental changes, and dynamic adjustment of energy storage strategies to adapt to different operating scenarios and emergencies;

[0055] Step 6: Compare the actual operating data with the model prediction results. When the deviation exceeds the set threshold, the model is triggered to retrain and optimize to ensure the accuracy and stability of the system. The retraining and optimization of the model includes establishing a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads.

[0056] It includes data acquisition and preprocessing module, deep learning model building module, model training and optimization module, energy storage control strategy module, collaborative work and optimization module, and system monitoring and feedback module;

[0057] The data acquisition and preprocessing module collects data on power generation, energy storage status, and load demand in real time through sensors installed on photovoltaic panels, energy storage devices, and load ends. These data are transmitted to the data processing center and undergo data cleaning, missing value filling, and outlier processing preprocessing operations to provide a high-quality data foundation for subsequent model training and prediction.

[0058] The deep learning model building module adopts an architecture that combines multi-layer convolutional neural networks (CNN) and long short-term memory networks (LSTM). CNN is used to extract local features in the data, and LSTM is used to capture long-term dependencies in time series. The model input includes historical photovoltaic power generation, weather data, and load demand, and the output is the photovoltaic power generation and load forecast values ​​for a period of time in the future.

[0059] The model training and optimization module uses preprocessed data for model training and adopts the adaptive momentum estimation (Adam) optimization algorithm to automatically adjust the learning rate according to the change of the loss function during the training process. At the same time, the Dropout regularization technology is introduced to randomly close some neurons to prevent the model from overfitting. In addition, during the training process, the performance of the model on the validation set is continuously monitored. When the performance no longer improves, the training is stopped in advance to avoid overtraining.

[0060] The energy storage control strategy module formulates the charging and discharging strategy of the energy storage equipment according to the prediction results of the deep learning model. When it is predicted that the photovoltaic power generation is in excess, the energy storage equipment is controlled to charge; when it is predicted that the photovoltaic power generation is insufficient and the load demand is large, the energy storage equipment is controlled to discharge. At the same time, the capacity limitation and charging and discharging efficiency factors of the energy storage equipment are considered to achieve optimal energy management.

[0061] The collaborative work and optimization module manages multiple distributed energy storage devices through a central controller, reasonably allocates charging and discharging tasks according to the location, capacity and operating status of each device, and improves the overall energy storage effect. In addition, by real-time monitoring of the system's operating status and external environmental changes, it dynamically adjusts the energy storage strategy to adapt to different operating scenarios and emergencies;

[0062] The system monitoring and feedback module establishes a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads, and compares the actual operating data with the model prediction results. When the deviation exceeds a certain threshold, it triggers the retraining and optimization of the model to ensure the accuracy and stability of the system.

[0063] By adopting the above technical solution:

[0064] In view of the model architecture, combined with the characteristics of photovoltaic power generation and load demand, a combination of multi-layer convolutional neural network and recurrent neural network is used to fully capture the time series characteristics and spatial correlation, improve the prediction accuracy and generalization ability of the model, and in terms of training methods, an adaptive optimization algorithm is used to dynamically adjust the training parameters according to data characteristics and model performance. At the same time, regularization technology is introduced to prevent overfitting and improve the stability of the model.

[0065] In addition, at the system level, through the coordinated control of multiple energy storage devices, the rational distribution and optimal utilization of energy are achieved. Flexible response plans are formulated for different operating scenarios and emergencies to improve the adaptability and stability of the system.

[0066] Working principle:

[0067] This system consists of multiple key parts working together. First, through high-precision sensors and monitoring devices distributed in photovoltaic panels, microgrid loads, energy storage equipment and the environment, comprehensive and real-time data is collected, including but not limited to the power generation power, voltage, and current of photovoltaic panels, the real-time power demand of microgrid loads, the power, voltage, temperature, and charge and discharge status of energy storage equipment, and multi-dimensional data such as light intensity, temperature, and humidity in the environment.

[0068] The collected raw data will be quickly transmitted to the data processing module. In the data processing module, data cleaning technology is used to remove noise and outliers, invalid or erroneous data is filtered out, and key features are extracted with the help of feature engineering to convert the raw data into a format suitable for deep learning model input to ensure the accuracy, completeness and availability of the data.

[0069] The processed data is input into a carefully constructed deep learning model. The model uses a multi-layer neural network architecture, such as a combination of convolutional neural network (CNN) and recurrent neural network (RNN), or long short-term memory network (LSTM), etc., to capture the complex spatiotemporal relationships and nonlinear characteristics in the data. The model has been fully trained in advance with a large amount of high-quality and representative historical data, and has learned the patterns and laws of photovoltaic power generation and load demand under different weather conditions, seasonal changes, and time periods.

[0070] The deep learning model performs in-depth analysis and accurate prediction based on the input data. It can not only accurately predict the subtle changes in photovoltaic power generation and load demand in the short term in the future, but also take into account the changes in long-term trends.

[0071] Based on the prediction results, the control module formulates an optimized energy storage strategy. When the model predicts that photovoltaic power generation will exceed load demand in the future and the energy storage device is not saturated, the control module issues a command to start the energy storage device to charge and efficiently store the excess electricity. At this time, the charging power and time will be reasonably allocated according to the characteristics and remaining capacity of the energy storage device. For example, deep charging of batteries with high energy density but relatively slow charging speed will be prioritized, while supercapacitors will be used to quickly absorb short-term surplus electricity.

[0072] On the contrary, when it is predicted that the load demand is about to exceed the photovoltaic power generation, or when power shortage has occurred, the control module quickly switches the energy storage device to discharge mode. According to the urgency and power demand of the load, the discharge sequence and discharge amount of different energy storage devices are flexibly adjusted to ensure that the microgrid can provide continuous and stable power supply.

[0073] During the entire operation process, the system continuously compares and analyzes the data generated by actual operation with the prediction results of the model, and uses the feedback mechanism to evaluate the prediction accuracy of the model and the effectiveness of the strategy in real time. If it is found that the actual situation deviates from the prediction, the system will automatically adjust the model parameters, or update the training data and re-train to continuously optimize the performance and decision-making ability of the deep learning model, so that it can better adapt to the ever-changing actual operating environment.

[0074] Through such a tight cycle of data collection, processing, prediction, decision-making and optimization, intelligent, efficient and precise management of distributed photovoltaic energy storage is achieved, ensuring the stable operation of the microgrid and the optimal use of energy.

[0075] Finally, it should be noted that: Obviously, the above embodiments are only examples for clearly explaining the present invention, and are not intended to limit the implementation methods. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from this are still within the scope of protection of the present invention.

Claims

1. A distributed photovoltaic energy storage method based on a deep learning model, characterized in that: The specific steps of the distributed photovoltaic energy storage method of the deep learning model are as follows: Step 1: Use data acquisition and preprocessing technology to collect and process relevant data of photovoltaic panels, energy storage equipment and load ends to eliminate noise and outliers in the data and normalize the data; Step 2: Build a deep learning model architecture that combines a multi-layer convolutional neural network and a long short-term memory network to accurately predict photovoltaic power generation and load demand in the future; Step 3: Use the adaptive momentum estimation optimization algorithm and introduce regularization technology to train the model, and establish a real-time feedback mechanism to continuously optimize the model based on the actual system operation data; Step 4: Based on the prediction results of the deep learning model, formulate a charging and discharging strategy for the energy storage equipment that takes into account the capacity limitation and charging and discharging efficiency factors of the energy storage equipment; Step 5: For multiple distributed energy storage devices, collaborative management is performed through a central controller to achieve reasonable allocation and optimal utilization of energy; Step 6: Compare the actual operation data with the model prediction results. When the deviation exceeds the set threshold, the model is retrained and optimized to ensure the accuracy and stability of the system.

2. A distributed photovoltaic energy storage method based on a deep learning model as claimed in claim 1, characterized in that: The data collection and preprocessing techniques in step one include data cleaning, missing value filling and outlier processing.

3. A distributed photovoltaic energy storage method based on a deep learning model as claimed in claim 1, characterized in that: In the deep learning model architecture in step 2, the convolutional neural network is used to extract local features in the data, and the long short-term memory network is used to capture long-term dependencies in the time series.

4. A distributed photovoltaic energy storage method based on a deep learning model as claimed in claim 1, characterized in that: The energy storage device charging and discharging strategy in step 4 controls the energy storage device to charge when it is predicted that the photovoltaic power generation is in excess, and controls the energy storage device to discharge when it is predicted that the photovoltaic power generation is insufficient and the load demand is large.

5. A distributed photovoltaic energy storage method based on a deep learning model as claimed in claim 1, characterized in that: The step five also includes dynamically adjusting the energy storage strategy by real-time monitoring the system's operating status and external environmental changes to adapt to different operating scenarios and emergencies.

6. A distributed photovoltaic energy storage method based on a deep learning model as claimed in claim 1, characterized in that: The retraining and optimization of the model in step six includes establishing a real-time monitoring system to conduct all-round monitoring of the operating status of photovoltaic power generation, energy storage equipment and loads.

7. A distributed photovoltaic energy storage system based on a deep learning model, characterized in that: It includes data acquisition and preprocessing module, deep learning model building module, model training and optimization module, energy storage control strategy module, collaborative work and optimization module, and system monitoring and feedback module; The data acquisition and preprocessing module is used to collect power generation, energy storage status and load demand data in real time, transmit the data to the data processing center, and perform data cleaning, missing value filling, and outlier processing preprocessing operations to provide a high-quality data foundation for subsequent model training and prediction. The deep learning model building module adopts a multi-layer convolutional neural network (CNN) and a long short-term memory network (LSTM) architecture to build a model for photovoltaic power generation and load forecast values ​​in the future; The model training and optimization module adopts an adaptive momentum estimation (Adam) optimization algorithm to automatically adjust the learning rate according to the change of the loss function during the training process; The energy storage control strategy module formulates a charging and discharging strategy for the energy storage device according to the prediction results of the deep learning model building module; The collaborative work and optimization modules are collaboratively managed through a central controller, and the charging and discharging tasks are reasonably allocated according to the location, capacity and operating status of each device to improve the overall energy storage effect; The system monitoring and feedback module compares the actual operation data with the model prediction results. When the deviation exceeds a certain threshold, it triggers the retraining and optimization of the model to ensure the accuracy and stability of the system.

8. A distributed photovoltaic energy storage system based on a deep learning model as claimed in claim 7, characterized in that: The distributed photovoltaic energy storage system has a complete safety protection mechanism, including overvoltage, overcurrent and overheating protection.

Images