A photovoltaic power generation energy storage control method and system based on AI group control technology

Abstract

The present invention discloses a photovoltaic power generation energy storage control method and system based on AI group control technology, which relates to the field of intelligent control, including collecting real-time data and multimodal data, integrating historical data, and performing data cleaning and preprocessing, using a deep learning model to train the processed multimodal data set, and predicting future power generation and power consumption. Based on the prediction results, a preliminary charging and discharging strategy for the energy storage system is formulated, and adaptive adjustments are made. The preliminary strategy is optimized using reinforcement learning technology, and the final strategy is executed. Real-time monitoring is performed to detect abnormal situations, and system performance is evaluated based on real-time monitoring data, and the strategy is adjusted based on the evaluation results. The present invention improves the intelligence and dynamic adaptability of the strategy, thereby reducing operating costs and improving economic benefits, ensuring the safe operation of the system, and improving the overall performance of the system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control, and in particular to a photovoltaic power generation energy storage control method and system based on AI group control technology. Background Art

[0002] With the rapid development of renewable energy technology, photovoltaic power generation has become one of the important clean energy sources. However, photovoltaic power generation systems are greatly affected by weather conditions, resulting in large fluctuations and intermittent power output. In order to improve the stability and reliability of the power grid, energy storage technology is widely used in photovoltaic power stations. Traditional energy storage system control strategies mainly rely on manually set rules, which are difficult to effectively cope with complex and changing environmental conditions. Therefore, in order to achieve more efficient and intelligent energy distribution and utilization, artificial intelligence technology has begun to be applied to photovoltaic energy storage systems to achieve higher energy efficiency and economic benefits.

[0003] At present, most photovoltaic energy storage systems adopt a single-mode data collection method, which results in limited data types obtained and makes it difficult to fully reflect the true status of the photovoltaic power station and its surrounding environment. Once a failure occurs, it is often impossible to discover and take countermeasures in time, affecting the stable operation and long-term performance of the system. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a photovoltaic power generation and energy storage control method and system based on AI group control technology to solve the problems of incomplete data collection, low prediction accuracy and insufficient intelligent and flexible charging and discharging strategies in existing photovoltaic power generation and energy storage control systems.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a photovoltaic power generation energy storage control method based on AI group control technology, which includes data collection and preprocessing, including real-time data collection, multimodal data collection, historical data integration and data preprocessing, to obtain a multimodal data set after collection and preprocessing;

[0008] Based on the multimodal data set, training and prediction are performed through a deep learning model to generate prediction results of future power generation and future power consumption;

[0009] Based on the prediction results, a preliminary energy storage system charging and discharging strategy is formulated, and adaptively adjusted to obtain an optimized energy storage system charging and discharging strategy;

[0010] Based on the optimized energy storage system charging and discharging strategy, a reinforcement learning method is used to optimize the scheduling strategy to obtain the final energy storage system charging and discharging strategy;

[0011] Implementing the final energy storage system charging and discharging strategy, while performing real-time monitoring, and performing abnormality detection and processing based on real-time monitoring data;

[0012] Based on real-time monitoring data, system evaluation and feedback are carried out to continuously optimize the charging and discharging strategy of the energy storage system.

[0013] As a preferred solution of the photovoltaic power generation energy storage control method based on AI group control technology described in the present invention, the real-time data collected is the power generation of the photovoltaic panel, the ambient temperature, and the light intensity, the multimodal data set collected is the market electricity price data, the weather forecast data, and the social media data, and the historical data integrates the historical data within the past year, including the power generation of the photovoltaic panel, the ambient temperature, and the light intensity;

[0014] All collected information is preprocessed. The preprocessing process includes data cleaning and data standardization.

[0015] After preprocessing, the real-time data and historical data are fused, and then further fused with multimodal data. The expression is:

[0016] ;

[0017] ;

[0018] in, Represents the data after the real-time data and historical data are integrated. represents the fusion coefficient, Indicates the normalized value of real-time data. Indicates the sequence number of the feature, Indicates the sequence number of the time point. Indicates the serial number of the day. represents the value of the corresponding feature in the historical data, Represents the data in the multimodal dataset after all data are fused, represents the market electricity price data, Represents weather forecast data, Represents the results of social media data analysis, Indicates the importance coefficient of the data after the fusion of real-time data and historical data in the multimodal data set, Represents the importance coefficient of market electricity price data in the multimodal data set, Represents the importance coefficient of weather forecast data in the multimodal dataset, Represents the importance coefficient of social media data in multimodal datasets.

[0019] As a preferred solution of the photovoltaic power generation energy storage control method based on AI group control technology described in the present invention, the multimodal data set is After normalization and information filtering, we get ;

[0020] Use a deep learning model to train and predict future power generation and future power consumption. The model will include a combination of convolutional neural networks and long short-term memory networks.

[0021] The predicted values ​​of future power generation and future power consumption are expressed as:

[0022] ;

[0023] ;

[0024] in, represents the predicted value of future power generation, represents the predicted value of future electricity consumption, represents the model for predicting power generation, represents a model for predicting electricity consumption, Denotes the time-integrated variable A small increment of Indicates the current time point. represents the time window length, represents the time-integrated variable Data after normalization and information filtering.

[0025] As a preferred solution of the photovoltaic power generation energy storage control method based on AI group control technology described in the present invention, a preliminary energy storage system charging and discharging strategy is formulated based on the prediction results, specifically including:

[0026] A preliminary strategy is formulated based on the prediction results of future power generation and future power consumption, and its expression is:

[0027] ;

[0028] in, Indicates the initial strategy, represents the policy parameters, Indicates the current time point. Represents the loop variable, represents the nonlinear coefficient, Indicates that at all time points The actual power generation Indicates that at all time points target power generation;

[0029] Based on the preliminary strategy, it is adaptively adjusted to obtain the optimized energy storage system charging and discharging strategy, which is expressed as:

[0030] ;

[0031] in, represents the optimized charging and discharging strategy of the energy storage system, represents the number of iterations, represents the learning rate, represents a nonlinear function, represents the cost function, Indicates passing After iterative optimization, the time point The strategy parameters, According to the optimal strategy parameters and real-time monitoring data The calculated The actual value of the performance indicator, Indicates passing strategy parameters.

[0032] As a preferred solution of the photovoltaic power generation energy storage control method based on AI group control technology described in the present invention, based on the optimized energy storage system charging and discharging strategy, the scheduling strategy is optimized by using the reinforcement learning method to obtain the final energy storage system charging and discharging strategy, which is expressed as follows:

[0033] ;

[0034] in, represents the final strategy after reinforcement learning, represents the learning rate of reinforcement learning, Indicates at a point in time Pass by Strategy parameters The rewards received, represents the average value of the reward, represents a derivative.

[0035] As a preferred solution of the photovoltaic power generation energy storage control method based on AI group control technology described in the present invention, the final energy storage system charging and discharging strategy is implemented, real-time monitoring is performed at the same time, and abnormality detection and processing are performed according to the real-time monitoring data. The expression is:

[0036] ;

[0037] in, represents the anomaly detection function, Expressed as time Real-time monitoring data, Indicates the number of historical data points used for comparison, Represents an index variable, Indicates a point in time, Represents the final strategy after reinforcement learning The calculated expected value;

[0038] Design an exception handling mechanism and preset a response threshold When the value of the anomaly detection function exceeds this threshold, it is considered that an anomaly has occurred. The anomaly is divided into three levels: level one represents a slight anomaly, level two represents a moderate anomaly, and level three represents a high anomaly; and corresponding processing is performed according to the anomaly level.

[0039] As a preferred solution of the photovoltaic power generation energy storage control method based on AI group control technology described in the present invention, wherein: a target score is set, and system evaluation and feedback are performed based on real-time monitoring data to continuously optimize the charging and discharging strategy of the energy storage system;

[0040] The system evaluation expression is:

[0041] ;

[0042] ;

[0043] in, represents the system evaluation score, Indicates The target value of the performance indicator is represents the number of performance indicators, Represents the final strategy after reinforcement learning and real-time monitoring data The calculated The actual value of the performance indicator;

[0044] The feedback level is determined based on the gap between the system evaluation score and the target score, and the system updates its strategy based on the feedback level until the expected goal is achieved.

[0045] In a second aspect, the present invention provides a photovoltaic power generation energy storage control system based on AI group control technology, comprising:

[0046] The data collection and preprocessing module is responsible for the collection of real-time data, multimodal data, and the integration of historical data. It also performs cleaning, preprocessing, and fusion to obtain a multimodal data set after all data are fused.

[0047] The predictive analysis module uses deep learning models to train the cleaned and preprocessed multimodal data to predict future power generation and consumption;

[0048] The initial strategy formulation module formulates a preliminary energy storage system charging and discharging strategy based on the prediction results, and makes adaptive adjustments to it to obtain the optimized energy storage system charging and discharging strategy;

[0049] A strategy optimization module, based on the optimized energy storage system charging and discharging strategy, uses a reinforcement learning method to optimize the scheduling strategy to obtain a final energy storage system charging and discharging strategy;

[0050] Real-time monitoring and exception handling module, which implements optimized strategies and performs real-time monitoring, and performs anomaly detection and processing based on monitoring data;

[0051] The system evaluation and feedback module evaluates system performance and provides feedback based on real-time monitoring data to continuously optimize the charging and discharging strategy of the energy storage system.

[0052] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the photovoltaic power generation and energy storage control method based on AI group control technology as described in the first aspect of the present invention is implemented.

[0053] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the photovoltaic power generation and energy storage control method based on AI group control technology as described in the first aspect of the present invention is implemented.

[0054] The beneficial effects of the present invention are as follows: through real-time data collection, multimodal data collection, historical data integration and data preprocessing steps, effective data integration and cleaning are achieved, the quality of the data is ensured, training and prediction are performed through deep learning models, and prediction results of future power generation and future power consumption are generated, thereby improving the prediction accuracy, and the formulation of a preliminary energy storage system charging and discharging strategy is based on the prediction results of future power generation and future power consumption, making the strategy more reasonable and improving energy utilization efficiency, and optimizing the scheduling strategy using reinforcement learning methods, thereby improving the intelligence and dynamic adaptability of the strategy, thereby reducing operating costs and improving economic benefits, and implementing the final energy storage system charging and discharging strategy while performing real-time monitoring, which can promptly detect and handle abnormal situations and ensure the safe operation of the system, and based on real-time monitoring data, system evaluation and feedback are performed, and the energy storage system charging and discharging strategy can be continuously optimized. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0056] Figure 1 This is a flow chart of the photovoltaic power generation energy storage control method based on AI group control technology in Example 1.

[0057] Figure 2 This is a decision diagram for the charging and discharging strategy of the energy storage system in Example 1. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0059] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a photovoltaic power generation energy storage control method based on AI group control technology, comprising the following steps:

[0060] S1. The real-time data collected are the power generation of photovoltaic panels, ambient temperature, and light intensity. The multimodal data set is the market electricity price data, weather forecast data, and social media data. The historical data integrates the historical data of the past year, including the power generation of photovoltaic panels, ambient temperature, and light intensity.

[0061] All collected information is preprocessed. The preprocessing process includes data cleaning and data standardization.

[0062] After preprocessing, the real-time data and historical data are fused, and then further fused with multimodal data. The expression is:

[0063] ;

[0064] ;

[0065] in, Represents the data after the real-time data and historical data are integrated. represents the fusion coefficient, Indicates the normalized value of real-time data. Indicates the sequence number of the feature, Indicates the sequence number of the time point. Indicates the serial number of the day. represents the value of the corresponding feature in the historical data, Represents the data in the multimodal dataset after all data are fused, represents the market electricity price data, Represents weather forecast data, Represents the results of social media data analysis, It represents the importance coefficient of the data after the fusion of real-time data and historical data in the multimodal data set. Represents the importance coefficient of market electricity price data in the multimodal data set, Represents the importance coefficient of weather forecast data in the multimodal dataset, Represents the importance coefficient of social media data in multimodal datasets. Their values ​​are between 0 and 1, and the sum is 1. For example, we can choose =0.4, =0.2, =0.2, =0.2.

[0066] Further explanation: Indicates the data after the real-time data and historical data are integrated In multimodal datasets The importance coefficient in the , which determines the impact of the fusion of real-time data and historical data on the overall prediction; Indicates market electricity price data In multimodal datasets The importance coefficient in the market electricity price data is very important for predicting power generation and consumption, because it is directly related to the cost and benefit of electricity; Indicates weather forecast data In multimodal datasets The importance coefficient in,weather forecast data is crucial for predicting photovoltaic power generation,because the light intensity directly affects the power generation capacity of photovoltaic panels; Represents social media data analysis results in multimodal datasets Social media data may contain public discussions about weather conditions, holidays, or other events, which may affect electricity demand.

[0067] Using the matrix Indicates the history The value of the ith feature of the day, the dth day represents a certain day in the historical data, for example, d=1 represents the first day of the year, d=365 (if it is not a leap year) represents the last day of the year, the ith feature refers to the different types of measured values ​​observed, including the power generation of photovoltaic panels, ambient temperature, and light intensity. The expression is as follows:

[0068] ;

[0069] It is the history The value of the ith feature of day, day d represents a certain day in the historical data, for example, is the power generation on the first day, The temperature on the first day. It is the light intensity of the first day. Each column represents the data of one day, and each row represents a feature.

[0070] S2. Multimodal dataset After normalization and information filtering, we get ;

[0071] Use a deep learning model to train and predict future power generation and consumption, which will include a combination of convolutional neural networks and long short-term memory networks;

[0072] The predicted values ​​of power generation and power consumption are expressed as:

[0073] ;

[0074] ;

[0075] in, represents the predicted value of future power generation, represents the predicted value of future electricity consumption, represents the model for predicting power generation, represents a model for predicting electricity consumption, Denotes the time-integrated variable A small increment of Used to calculate from the current time point Backtracking The cumulative value of the data within a time period. This means that the model takes into account the changing trend of the data during this period to capture the dynamic characteristics of this period. Indicates the current time point. Indicates the length of the time window, which defines the time range for predicting future power generation and power consumption, starting from the current time point The length of time to look back, Denotes the time-integrated variable Data after normalization and information filtering.

[0076] Further explanation: arrive The time range is used as the integration interval to calculate the arrive The cumulative value of power generation and power consumption during this period. Assuming that the current time point is 10:00 a.m., if If it is set to 1 hour, then the integration interval is from 9:00 am to 10:00 am. The model will predict future power generation and power consumption based on the data during this period.

[0077] It is further explained that the combination of convolutional neural networks and long short-term memory networks combines the advantages of these two models, which can process data in spatial and temporal dimensions at the same time, improve the overall performance of the prediction model, and summarize data over a period of time through integration operations. This method can capture the changing trend of data over time and help improve the accuracy and stability of predictions.

[0078] S3. Based on the prediction results, formulate a preliminary energy storage system charging and discharging strategy;

[0079] Based on the prediction results of power generation and power consumption, a preliminary strategy is formulated. This strategy will be based on the difference between power generation and power consumption to achieve the optimal operation of the energy storage system. Its expression is:

[0080] ;

[0081] in, Indicates the initial strategy, represents the policy parameters, Indicates the current time point. Represents the loop variable, represents the nonlinear coefficient, which is used to adjust the shape of the strategy parameters. Indicates that at all time points The actual power generation Indicates that at all time points target power generation;

[0082] This part of the formula is the Sigmoid function, whose output range is between 0 and 1. The output range of the Sigmoid function is converted from [0, 1] to [-0.5, 0.5]. When it is close to 0.5, it indicates that the energy storage system tends to charge, and when it is close to −0.5, it tends to discharge.

[0083] Based on the preliminary strategy, it is adaptively adjusted to obtain the optimized energy storage system charging and discharging strategy, which is expressed as:

[0084] ;

[0085] in, Represents the optimized energy storage system charging and discharging strategy. Updated, represents the number of iterations, Represents the learning rate, which is used to control the amplitude of parameter updates. Represents a nonlinear function, which is used to adjust the shape of the objective function. ( ) represents the cost function, which is used to quantify the quality of the strategy. Usually, a good strategy will lead to a lower cost function value. Indicates passing After iterative optimization, the time point The strategy parameters, Indicates passing strategy parameters.

[0086] The above formula describes the updating process of the preliminary policy, which shows how to update the policy parameters based on the current policy parameters, the gradient of the cost function, and the predicted difference related to the current time point t.

[0087] S4. Based on the optimized energy storage system charging and discharging strategy, the scheduling strategy is optimized using the reinforcement learning method to obtain the final energy storage system charging and discharging strategy, which is expressed as:

[0088] ;

[0089] in, represents the final strategy after reinforcement learning, represents the learning rate of reinforcement learning, Indicates at a point in time Pass by Strategy parameters The rewards received, represents the average value of the reward, Represents a derivative, representing a nonlinear function The rate of change is used to reflect how changes in input variables affect changes in output.

[0090] The above Indicates that at a given point in time and after Strategy parameters The reward obtained by the energy storage system under this condition. This reward value is regarded as a measure of the performance of the strategy in a specific state. The reward is a feedback mechanism of the environment to the behavior of the agent, which is used to guide the energy storage system control strategy to learn how to make the best decision. The reward value can be positive or negative, depending on the performance of the strategy. For example, when the strategy leads to the desired result or state, the energy storage system will receive a positive reward, which encourages the energy storage system to take similar actions in the future; if the strategy leads to an undesirable result or state, the energy storage system will be punished, that is, it will receive a negative reward, which will prompt the energy storage system to avoid such behavior as much as possible in the future;

[0091] Rewards are calculated based on the following aspects:

[0092] Economical: reduce energy costs or increase revenue. For example, for energy storage systems, if a certain charging and discharging operation results in lower energy costs, positive rewards can be given;

[0093] Grid stability: reduce grid fluctuations and increase grid reliability. If the operation of the energy storage system helps smooth the grid load, positive rewards can be given;

[0094] Energy balance: Ensure the balance between energy supply and demand to avoid overcharging and overdischarging. If the operation of the energy storage system better matches energy supply and demand, positive rewards can be given.

[0095] Equipment life: Extend the service life of key components such as batteries. If the operation of the energy storage system avoids excessive charging and discharging, positive rewards can be given.

[0096] S5. Implement the final energy storage system charging and discharging strategy, perform real-time monitoring, and perform abnormality detection and processing based on the real-time monitoring data. The expression is:

[0097] ;

[0098] in, represents the anomaly detection function, Expressed as time Real-time monitoring data, Indicates the number of historical data points used for comparison, Represents an index variable, used to traverse n historical data points, Indicates a point in time, Represents the final strategy after reinforcement learning The calculated expected value;

[0099] Design an exception handling mechanism and preset a response threshold , The value of is 1. When the value of the anomaly detection function exceeds this threshold, it is considered that an anomaly has occurred. The anomaly is divided into three levels and handled accordingly. When the value of the anomaly detection function is between U and 2U, it is level one, indicating a minor anomaly, which is repaired by adjusting the charging and discharging power of the energy storage system; when the value of the anomaly detection function is between 2U and 3U, it is level two, indicating a moderate anomaly, and the maintenance personnel are notified to check; when the value of the anomaly detection function is greater than 3U, it is level three, indicating a high anomaly, and the energy storage system is shut down urgently to prevent further damage.

[0100] S6. Set target scores and conduct system evaluation and feedback based on real-time monitoring data to continuously optimize the energy storage system charging and discharging strategy;

[0101] The system evaluation expression is:

[0102] ;

[0103] ;

[0104] in, represents the system evaluation score, Indicates The target value of the performance indicator is represents the number of performance indicators, Represents the final strategy after reinforcement learning and real-time monitoring data The calculated The actual value of the performance indicator;

[0105] The value range of the system evaluation score is [0,10). The smaller the value, the closer the system performance is to the target performance, and the system performs well. The larger the value, the greater the gap between the system performance and the target performance. The larger the value, the greater the performance deviation.

[0106] Evaluate whether the performance indicators of the energy storage system meet the expected goals, including efficiency, stability, etc.

[0107] The feedback level is determined based on the gap between the system evaluation score and the target score. The actual effect is considered to have a small gap if the difference between the actual effect and the expected effect is within 200, a moderate gap if the difference between the actual effect and the expected effect is within 500, and a large gap if the difference between the actual effect and the expected effect is within 800. If the gap is small, the current strategy is maintained; if the gap is moderate, moderate adjustments are made; if the gap is large, major adjustments are made. For example, the target power generation is 900 kWh, and the actual power generation is 500 kWh, with a difference of 400 kWh. The feedback level is moderate, and moderate adjustments are made at this time, such as adding the angle adjustment strategy of the solar panel to improve the power generation efficiency; maintain and update the strategy parameters, continue to monitor the power generation on the second day, and recalculate the gap. If the gap is reduced, it means that the strategy is effective; otherwise, continue to adjust until the ideal state is reached.

[0108] This embodiment also provides a photovoltaic power generation energy storage control system based on AI group control technology, including:

[0109] The data collection and preprocessing module is responsible for the collection of real-time data, multimodal data, and the integration of historical data. It also performs cleaning, preprocessing, and fusion to obtain a multimodal data set after all data are fused.

[0110] The predictive analysis module uses deep learning models to train the cleaned and preprocessed multimodal data to predict future power generation and consumption;

[0111] The initial strategy formulation module formulates a preliminary energy storage system charging and discharging strategy based on the prediction results, and makes adaptive adjustments to it to obtain the optimized energy storage system charging and discharging strategy;

[0112] A strategy optimization module, based on the optimized energy storage system charging and discharging strategy, uses a reinforcement learning method to optimize the scheduling strategy to obtain a final energy storage system charging and discharging strategy;

[0113] The real-time monitoring and exception handling module implements the final strategy and conducts real-time monitoring, while performing anomaly detection and processing based on monitoring data;

[0114] The system evaluation and feedback module evaluates system performance and provides feedback based on real-time monitoring data to continuously optimize the charging and discharging strategy of the energy storage system.

[0115] This embodiment also provides a computer device, which is suitable for the photovoltaic power generation and energy storage control method based on AI group control technology, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the photovoltaic power generation and energy storage control method based on AI group control technology proposed in the above embodiment.

[0116] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0117] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by the processor, the photovoltaic power generation energy storage control method based on AI group control technology proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, disk or optical disk.

[0118] In summary, the present invention realizes effective data integration and cleaning through real-time data collection, multimodal data collection, historical data integration and data preprocessing steps, ensures data quality, trains and predicts through deep learning models, generates prediction results of future power generation and power consumption, improves prediction accuracy, and formulates preliminary energy storage system charging and discharging strategies. Based on the prediction results of power generation and power consumption, the strategies are more reasonable and energy utilization efficiency is improved. The scheduling strategy is optimized by reinforcement learning methods, which improves the intelligence and dynamic adaptability of the strategies, thereby reducing operating costs and improving economic benefits. The final energy storage system charging and discharging strategy is implemented while real-time monitoring is performed, which can timely discover and handle abnormal situations and ensure the safe operation of the system. System evaluation and feedback are performed based on real-time monitoring data, which can continuously optimize the energy storage system charging and discharging strategy.

[0119] Example 2, referring to Table 1, is the second example of the present invention. To further verify the advancement of the present invention, experimental simulation data of a photovoltaic power generation energy storage control method and system based on AI group control technology are provided.

[0120] In order to verify the effect of the present invention, a one-month experiment was conducted, in which a photovoltaic power generation system was installed on the roof of a commercial building in a certain area, and an energy storage control system based on AI group control technology was used. The system is equipped with a data acquisition and preprocessing module, a prediction analysis module, an initial strategy formulation module, a strategy optimization module, a real-time monitoring and exception handling module, and a system evaluation and feedback module.

[0121] Test preparation content:

[0122] Hardware equipment:

[0123] Photovoltaic panels: 100 pieces, each with a maximum power of 250W.

[0124] Energy storage device: total capacity is 500kWh.

[0125] Sensors: used to monitor ambient temperature, light intensity, real-time power generation, etc.

[0126] Software Configuration:

[0127] Data acquisition and preprocessing module: integrates real-time data collector, multimodal data collector and historical data integrator.

[0128] Predictive analysis module: Contains deep learning model (CNN-LSTM combination).

[0129] Initial strategy formulation module: formulates the initial charging and discharging strategy based on the prediction results.

[0130] Strategy optimization module: uses reinforcement learning algorithm for strategy optimization.

[0131] Real-time monitoring and exception handling module: monitor the system operation status and handle abnormal situations in a timely manner.

[0132] System Evaluation and Feedback Module: Evaluate system performance and provide feedback.

[0133] Data preparation:

[0134] Real-time data: including photovoltaic panel power generation, ambient temperature, and light intensity.

[0135] Multimodal data sets: including market electricity price data, weather forecast data, and social media data analysis results.

[0136] Historical data: historical data within the past year.

[0137] Detailed implementation process:

[0138] Data collection and preprocessing:

[0139] Use sensors to collect data such as power generation, ambient temperature, and light intensity in real time.

[0140] Market electricity prices, weather forecasts, and social media data are collected as multimodal data.

[0141] Clean and preprocess historical data, remove noise and missing values, and perform standardization.

[0142] Fusion of real-time data and historical data, and further fusion with multimodal data.

[0143] Predictive Analytics:

[0144] The fused multimodal data are trained using the CNN-LSTM deep learning model.

[0145] Forecast electricity generation and consumption for the next week.

[0146] Initial strategy development and optimization:

[0147] According to the prediction results, the initial charging and discharging strategy is formulated.

[0148] The initial strategy is optimized through the reinforcement learning algorithm to obtain the optimized charging and discharging strategy.

[0149] Real-time monitoring and exception handling:

[0150] Implement optimized strategies and continuously monitor system operation status.

[0151] Design an anomaly detection mechanism. When the value of the anomaly detection function exceeds the threshold, different levels of exception handling processes are triggered.

[0152] System evaluation and feedback:

[0153] Evaluate system performance weekly and adjust strategies based on the results.

[0154] Continuously optimize the entire control process to achieve the best performance.

[0155] The details are shown in Table 1:

[0156] Table 1 shows the comparison between the actual results and the predicted results.

[0157]

[0158] Section 3: Tabular Data Analysis

[0159] Data Analysis:

[0160] Accuracy of power generation and consumption forecast:

[0161] It can be seen from the table that the predicted power generation and power consumption are very close to the actual values, and the prediction error is within 2%.

[0162] Compared with the traditional method that only relies on historical data for prediction, the present invention improves the prediction accuracy through multimodal data fusion.

[0163] Optimized strategy score:

[0164] The optimized strategy scored between 85 and 90 points, indicating that the system can effectively utilize energy storage devices while maintaining a balance between power generation and consumption.

[0165] This score is much higher than the strategy score based on fixed rules in the prior art (usually between 60 and 70 points), indicating that the strategy of the present invention is more flexible and effective.

[0166] Abnormality level and system evaluation score:

[0167] During the entire experiment, the anomaly level was level 1 (minor anomaly), indicating that the anomaly detection and handling mechanism can effectively identify and handle problems in the system.

[0168] The system evaluation score (0.06-0.1) indicates that the overall performance of the system is good, which is lower than the average score of the prior art (0.15-0.25), reflecting the significant improvement of the performance stability of the present invention.

[0169] Comparison with existing technologies:

[0170] The photovoltaic power generation and energy storage control system of the present invention is significantly superior to the prior art in terms of power generation prediction accuracy, strategy optimization effect and system stability.

[0171] By introducing multimodal data fusion and reinforcement learning methods, the present invention can better cope with complex and changing environmental conditions and achieve more efficient energy management.

[0172] In summary, the present invention proposes a photovoltaic power generation energy storage control method and system based on AI group control technology, which significantly improves the prediction accuracy of the system and the effectiveness of strategy execution through the effective fusion of real-time data and multimodal data, and through the optimization strategy of reinforcement learning algorithm. In addition, the anomaly detection and processing mechanism ensures the stable operation of the system, and the system evaluation score further proves the performance advantage of the present invention. These results show that the present invention is highly innovative and practical, can effectively solve the problems existing in the prior art, and bring positive impact to the field of photovoltaic energy storage.

[0173] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A photovoltaic power generation energy storage control method based on AI group control technology, characterized in that: include: Data collection and preprocessing, including real-time data collection, multimodal data collection, historical data integration and data preprocessing, to obtain multimodal data sets after collection and preprocessing; Based on the multimodal data set, training and prediction are performed through a deep learning model to generate prediction results of future power generation and future power consumption; Based on the prediction results, a preliminary energy storage system charging and discharging strategy is formulated, and adaptively adjusted to obtain an optimized energy storage system charging and discharging strategy; Based on the optimized energy storage system charging and discharging strategy, a reinforcement learning method is used to optimize the scheduling strategy to obtain the final energy storage system charging and discharging strategy; Implementing the final energy storage system charging and discharging strategy, while performing real-time monitoring, and performing abnormality detection and processing based on real-time monitoring data; Based on real-time monitoring data, system evaluation and feedback are carried out to continuously optimize the charging and discharging strategy of the energy storage system; The real-time data collected include the power generation of photovoltaic panels, ambient temperature, and light intensity. The multimodal data collected include market electricity price data, weather forecast data, and social media data. The historical data integrates historical data from the past year, including the power generation of photovoltaic panels, ambient temperature, and light intensity. All collected information is preprocessed. The preprocessing process includes data cleaning and data standardization. After preprocessing, the real-time data and historical data are fused, and then further fused with multimodal data. The expression is: ； ； in, Represents the data after the real-time data and historical data are integrated. represents the fusion coefficient, Indicates the normalized value of real-time data. Indicates the sequence number of the feature, Indicates the sequence number of the time point. Indicates the serial number of the day. represents the value of the corresponding feature in the historical data, Represents the data in the multimodal dataset after all data are fused, represents the market electricity price data, Represents weather forecast data, Represents social media data, Indicates the importance coefficient of the data after the fusion of real-time data and historical data in the multimodal data set, Represents the importance coefficient of market electricity price data in the multimodal data set, Represents the importance coefficient of weather forecast data in the multimodal dataset, Represents the importance coefficient of social media data in multimodal datasets.

2. The photovoltaic power generation energy storage control method based on AI group control technology according to claim 1, characterized in that: Multimodal datasets After normalization and information filtering, we get ; Use a deep learning model to train and predict future power generation and future power consumption. The model includes a combination of convolutional neural networks and long short-term memory networks. The predicted values ​​of future power generation and future power consumption are expressed as: ； ； in, represents the predicted value of future power generation, represents the predicted value of future electricity consumption, represents the model for predicting power generation, represents a model for predicting electricity consumption, Denotes the time-integrated variable A small increment of Indicates the current time point. represents the time window length, Denotes the time-integrated variable Data after normalization and information filtering.

3. The photovoltaic power generation energy storage control method based on AI group control technology according to claim 2 is characterized in that: Based on the prediction results, a preliminary energy storage system charging and discharging strategy is formulated, including: A preliminary strategy is formulated based on the forecast results of future power generation and future power consumption, and its expression is: ； in, Indicates the initial strategy, represents the policy parameters, Indicates the current time point. Represents the loop variable, represents the nonlinear coefficient, Indicates that at all time points The actual power generation Indicates that at all time points target power generation; Based on the preliminary strategy, it is adaptively adjusted to obtain the optimized energy storage system charging and discharging strategy, which is expressed as: ； in, represents the optimized charging and discharging strategy of the energy storage system, represents the number of iterations, represents the learning rate, represents a nonlinear function, represents the cost function, Indicates passing After iterative optimization, the time point The strategy parameters, Indicates passing strategy parameters.

4. The photovoltaic power generation energy storage control method based on AI group control technology as claimed in claim 3, characterized in that: Based on the optimized energy storage system charging and discharging strategy, the scheduling strategy is optimized using the reinforcement learning method to obtain the final energy storage system charging and discharging strategy, which is expressed as: ； in, represents the final strategy after reinforcement learning, represents the learning rate of reinforcement learning, Indicates at a point in time Pass by Strategy parameters The rewards received, represents the average value of the reward, represents a derivative.

5. The photovoltaic power generation energy storage control method based on AI group control technology as claimed in claim 4, characterized in that: The final energy storage system charging and discharging strategy is implemented, real-time monitoring is performed at the same time, and abnormality detection and processing are performed according to the real-time monitoring data. The expression is: ； in, represents the anomaly detection function, Expressed as time Real-time monitoring data, Indicates the number of historical data points used for comparison, Represents an index variable, Indicates the current time point. Represents the final strategy after reinforcement learning The calculated expected value; Design an exception handling mechanism and preset a response threshold , when the value of the anomaly detection function exceeds the response threshold , it is considered that an abnormality has occurred and the abnormality is divided into three levels: level one indicates a slight abnormality, level two indicates a moderate abnormality, and level three indicates a severe abnormality; and corresponding processing is performed according to the abnormality level.

6. The photovoltaic power generation energy storage control method based on AI group control technology as claimed in claim 5, characterized in that: Set target scores and conduct system evaluation and feedback based on real-time monitoring data to continuously optimize the energy storage system charging and discharging strategy; The system evaluation expression is: ； ； in, represents the system evaluation score, Indicates The target value of the performance indicator is represents the number of performance indicators, Represents the final strategy after reinforcement learning and real-time monitoring data The calculated The actual value of the performance indicator; The feedback level is determined based on the gap between the system evaluation score and the target score, and the system updates its strategy based on the feedback level until the expected goal is achieved.

7. A photovoltaic power generation energy storage control system based on AI group control technology, based on the photovoltaic power generation energy storage control method based on AI group control technology according to any one of claims 1 to 6, characterized in that: include: The data collection and preprocessing module is responsible for the collection of real-time data, multimodal data, and the integration of historical data. It also performs cleaning, preprocessing, and fusion to obtain a multimodal data set after all data are fused. The predictive analysis module uses deep learning models to train the cleaned and preprocessed multimodal data sets to predict future power generation and consumption; The initial strategy formulation module formulates a preliminary energy storage system charging and discharging strategy based on the prediction results, and makes adaptive adjustments to it to obtain the optimized energy storage system charging and discharging strategy; A strategy optimization module, based on the optimized energy storage system charging and discharging strategy, uses a reinforcement learning method to optimize the scheduling strategy to obtain a final energy storage system charging and discharging strategy; The real-time monitoring and exception handling module implements the final strategy and conducts real-time monitoring, while performing anomaly detection and processing based on monitoring data; The system evaluation and feedback module evaluates system performance and provides feedback based on real-time monitoring data to continuously optimize the charging and discharging strategy of the energy storage system.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the photovoltaic power generation and energy storage control method based on AI group control technology are implemented as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the photovoltaic power generation and energy storage control method based on AI group control technology described in any one of claims 1 to 6 are implemented.

Images