Photovoltaic energy storage battery group action intelligent control method and system

Abstract

The present invention discloses a method and system for intelligently regulating the action of a photovoltaic energy storage battery pack. The present invention collects the status data of each battery cell in the photovoltaic energy storage battery pack in real time, and uses an intelligent body model to perform status data analysis and target predictive evaluation, so as to achieve regulation and parameter optimization of the single-cell charging and discharging action for each battery cell. The present invention accurately predicts the power supply demand of the battery pack within a predetermined time, and then ensures the best match between energy supply and demand through independent monitoring and control of the battery cells. The present invention dynamically adjusts the charging and discharging strategy of the battery cell, which not only optimizes the energy supply efficiency, but also effectively extends the overall service life of the battery pack. Through refined single-cell management, the performance and reliability of the photovoltaic energy storage battery pack in complex application environments are improved.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic technology, and in particular to a method and system for intelligently controlling the operation of a photovoltaic energy storage battery group. Background Art

[0002] With the rapid development of photovoltaic technology, photovoltaic energy storage systems that use solar energy as the main energy source and combine it with energy sources such as city electricity and generators have been increasingly widely used in various application scenarios. Whether it is household electricity, outdoor camps, or emergency rescue and power supply in remote areas, photovoltaic energy storage systems provide stable and environmentally friendly power support. In home applications, it can be used as the main or backup source of electricity to ensure the stability and reliability of household electricity. In outdoor camps and emergency rescue scenarios, photovoltaic energy storage systems can provide the necessary power support to meet basic electricity needs such as lighting, communications, and medical equipment. In addition, photovoltaic energy storage systems are also widely used in mobile RVs, ship power supply, agricultural irrigation systems and many other fields.

[0003] The photovoltaic energy storage system mainly consists of the following key components. (1) Solar photovoltaic panels, which convert solar energy into direct current and transmit it to the energy storage inverter through cables. (2) Lithium iron phosphate battery pack: It includes several parallel lithium iron phosphate single battery units, which are responsible for receiving and storing the electricity generated by the photovoltaic panels; in addition, it can also receive and store electricity from the mains power grid or backup generators to ensure sufficient power supply during periods of insufficient sunshine or peak electricity consumption. (3) Intelligent control system: It is used to manage the charging and discharging process of the battery pack, manage and distribute the electricity from the photovoltaic panels, and ensure the efficient operation of the system and the long expected life of the battery. (4) Photovoltaic off-grid inverter: It converts the direct current in the lithium iron phosphate battery pack into alternating current, which is suitable for most household and industrial electrical equipment; the inverter's efficient conversion ensures that the power loss is minimized and improves the overall efficiency of the system. (5) Power facilities: The alternating current generated by the photovoltaic off-grid inverter is used to supply power to various power facilities, which can include lighting systems, office equipment, household appliances, agricultural equipment, etc., to meet the power needs in various scenarios.

[0004] In photovoltaic energy storage systems, battery pack management is a key but complex link. Although the existing technology has made significant progress, there are still some shortcomings and defects, mainly concentrated in the following aspects: (1) Charge and discharge balance of single battery cells: The battery pack is composed of multiple single battery cells, which may have differences in capacity, internal resistance and other characteristics; these differences will cause some single battery cells to be overcharged or over-discharged during the charge and discharge process, thereby affecting the performance and life of the entire battery pack; existing battery management systems usually rely on passive balancing or active balancing to balance the charge and discharge state of single battery cells, but these methods may be inefficient or costly. (2) During long-term use, the capacity of the battery will gradually decay and the internal resistance will increase, resulting in a decrease in battery performance. Factors such as temperature and charge and discharge rate will affect the life of the battery. The current battery management system mostly adopts a fixed charge and discharge strategy, which may not be able to adapt to different working environments and usage conditions, resulting in a shortened battery life. Among them, the operating temperature of the battery has a significant impact on its performance and life. High temperature will accelerate battery aging, and low temperature will reduce battery capacity and discharge efficiency. A more intelligent temperature management system is needed that can monitor the battery temperature in real time and keep the battery within the optimal operating temperature range. (3) The photovoltaic energy storage system needs to configure the battery capacity according to the expected power supply demand. If the configuration is improper, it may cause the battery to over-discharge or insufficient energy storage, affecting the stability of power supply. Existing methods mostly rely on static demand forecasting models, which are difficult to cope with dynamically changing power demand. Therefore, it is necessary to develop a dynamic demand forecasting model that combines multiple factors such as weather forecasts and historical power consumption data to adjust the working mode of the battery pack in real time and improve the power supply matching degree. Summary of the invention

[0005] In view of the above problems, the purpose of the present invention is to provide a method and system for intelligent control of the action of a photovoltaic energy storage battery pack. The present invention collects the status data of each battery cell in the photovoltaic energy storage battery pack in real time, and uses an intelligent body model to perform status data analysis and target predictive evaluation, so as to achieve regulation and parameter optimization of the single-cell charging and discharging action for each battery cell. The present invention accurately predicts the power supply demand of the battery pack within a predetermined time, and then ensures the best match between energy supply and demand through independent monitoring and control of the battery cells. The present invention dynamically adjusts the charging and discharging strategy of the battery cell, which not only optimizes the energy supply efficiency, but also effectively extends the overall service life of the battery pack. Through refined single-cell management, the performance and reliability of photovoltaic energy storage battery packs in complex application environments are improved.

[0006] The present invention provides a method for intelligently controlling the operation of a photovoltaic energy storage battery group, which is characterized by comprising the following steps:

[0007] Step S101: Determine a power supply forecast mode according to external factors such as season, estimated time period, and weather forecast, and then combine historical data under the same power supply forecast mode to predict power supply demand within a predetermined time in the future;

[0008] Step S102: using multiple types of sensors, real-time parameters of each battery cell in the photovoltaic energy storage battery group are collected in real time, and a representation of the expected life span is generated for each battery cell;

[0009] Step S103: establishing a power supply demand allocation model for each battery unit according to the predicted power supply demand and the quantitative representation value of the expected battery life of each battery unit in the photovoltaic energy storage battery group;

[0010] Step S104: The power supply demand allocated to each battery cell of the battery pack is used as a state observation quantity, and input into a trained intelligent agent model for realizing optimal management of the battery pack to obtain control action parameters and target evaluation results for each battery cell of the battery pack; and the charging and discharging actions of each battery cell of the battery pack are controlled according to the control action parameters.

[0011] Preferably, in step S101, external factors such as season, estimated time period, and weather forecast are used as inputs of a predictive neural network, and the output of the predictive neural network is a power supply prediction model. Then, combined with historical data belonging to the same power supply prediction model, linear regression is used to predict the photovoltaic energy storage power supply demand within a predetermined time in the future.

[0012] Preferably, the seasonal data is represented by s, which is the seasonal stage converted into the corresponding numerical value or thermal value code; the estimated time period data is represented by t, which is the interval number of each hour from 0:00 to 24:00 every day; the weather forecast data is represented by w, which is an array including temperature, humidity, cloudiness, and air particle information; the input layer of the predictive neural network inputs the total feature value of the combination of seasonal data, estimated time period data and weather forecast data , the hidden layer of the predictive neural network is set to multiple layers, and the activation function used in each layer It is a ReLU or Sigmoid function; the output layer outputs the predicted category quantity representing the power supply prediction mode , the output of the neural network is expressed as ,in is the weight matrix of the neural network, The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture.

[0013] Preferably, historical data points belonging to the same category as the power supply forecasting model are selected from the historical data, and the power supply demand of these historical data points is used as the dependent variable , taking time as the independent variable , then the model form of linear regression prediction is: ,in is the intercept, is the slope, is the error term, obtained through historical data points and The value of can then predict the photovoltaic energy storage power supply demand within a predetermined time in the future.

[0014] Preferably, in step S102, the real-time parameters of each battery cell in the photovoltaic energy storage battery group include: voltage, current, SOC, temperature, current charge amount, current discharge amount, and internal resistance value of each battery cell.

[0015] Preferably, in step S102, for each battery cell, a battery cell model function is used to obtain a quantitative representation of the expected battery life according to the real-time parameters of the battery cell; the battery cell model function is decomposed into several sub-functions of life influencing factors, expressed as: ;in, The quantitative representation value of the expected battery life, function Indicates voltage , Current , SOC, temperature Impact function on battery life expectancy; function Indicates the current charging amount And the current discharge amount Impact function on battery life expectancy; function Indicates internal resistance Impact function on battery life expectancy.

[0016] Preferably, in step S103, the allocation model is expressed as: ,in, For time interval Assign to The power supply requirements of each battery unit; A decision function for allocation to achieve the best match between power supply demand and expected life of battery units; For the prediction time interval Power supply requirements is the expected battery life of each battery cell.

[0017] Preferably, in step S104, the state space of the agent model is represented as , the power supply requirements allocated to each battery cell of the battery pack are used as state observations in the state space ; Control action space , where the charge and discharge state action switching and charge and discharge power parameters of each battery cell in the battery pack are used as the control actions of this space ; The agent model responds to the state observation Decide on control actions The response function is defined as ,in is the network parameter tensor of the deep neural network used by the response function to implement the autonomous learning mechanism; the intelligent agent model evaluates multiple optimization goals such as energy supply and demand balance, battery life, and safety based on state observations, and defines the target evaluation function as ,in The network parameter tensor of the deep neural network for the target evaluation function; through agent training, the response function is obtained And the objective evaluation function , through which the network parameter tensor of the deep neural network and Assignment after training and , implement the response function The control action for the battery pack is output according to the state observation of the battery pack, and the state observation of the battery pack is evaluated by the target evaluation function Ability to obtain continuous optimization results of target evaluation.

[0018] Preferably, the training process of the agent model includes: initializing the state observation of the input agent model , and initialize the response function of the intelligent agent model to determine the control action in response to the state observation , where the network parameter tensor of the deep neural network of the response function is initialized as ; Initialize the target evaluation function of the agent model , the network parameter tensor of the deep neural network of the objective evaluation function is initialized as ; It is scheduled to perform N rounds of agent model training, and in each round n ( ), according to the time step t of this round (here , is the maximum number of time steps in round n), according to the state observation at time step t , and the network parameter tensor of the deep neural network of the response function after training and parameter adjustment at this time step t , calculate the control action of this time step t through the response function , and then observe the control action at this time step t After execution, the state observation at the next time step t+1 is ; Based on the state observations of the current time step t and the next time step t+1 and , the target evaluation is performed through the target evaluation function, and the target evaluation result , , where the network parameter tensor of the deep neural network of the target evaluation function after training and parameter adjustment at time step t is ; Then, calculate the time series difference component of the target evaluation result for ; Then, update the network parameter tensor of the deep neural network of the response function in the agent model at the t+1th time step ,in is a hyperparameter representing the learning rate, and ,express About parameters The gradient of ; and, update the network parameter tensor of the deep neural network of the target evaluation function in the agent model at the t+1th time step ,in is a hyperparameter representing the learning rate, is the objective evaluation function About parameters Gradient of; Thus, through training of all time steps in a total of N rounds, the response function of the intelligent agent model for realizing optimal management of the battery pack is obtained ; And, obtain the target evaluation function of the agent model .

[0019] On the other hand, the present invention also provides a photovoltaic energy storage battery group action intelligent control system, including: a power supply demand prediction unit, a multi-type sensor unit, an expected life unit, a power supply demand allocation unit, and an intelligent body control unit;

[0020] The power demand forecasting unit is used to determine the power supply forecasting mode according to the external factors of season, estimated time period and weather forecast, and then combine the historical data under the same power supply forecasting mode to forecast the power supply demand within the future predetermined time;

[0021] Multiple types of sensor units are used to collect real-time parameters of each battery cell in the photovoltaic energy storage battery group;

[0022] An expected life unit, for obtaining a quantitative representation of the expected life of the battery according to real-time parameters of the battery cell using a battery cell model function;

[0023] A power supply demand allocation unit is used to establish a power supply demand allocation model for each battery unit according to the predicted power supply demand and the quantitative representation value of the expected battery life of each battery unit in the photovoltaic energy storage battery group;

[0024] An intelligent agent control unit is used to take the power supply demand allocated to each battery cell of the battery pack as a state observation, input it into a trained intelligent agent model for realizing optimal management of the battery pack, obtain control action parameters and target evaluation results for each battery cell of the battery pack; and control the charging and discharging actions of each battery cell of the battery pack according to the control action parameters.

[0025] The beneficial effects of the above technical solutions provided by the embodiments of the present invention include at least: achieving optimal battery pack energy management: by dynamically adjusting the charging and discharging strategy through the intelligent body model, the energy distribution and utilization can be optimized according to the real-time energy supply and demand status, thereby improving the energy utilization rate of the entire system; intelligent optimization management helps to avoid excessive charging and discharging of the battery and working in an inappropriate state, which can reduce battery loss and extend the service life of the battery pack. By intelligently monitoring the working status of the battery pack, potential safety risks such as overheating and overcharging can be discovered and handled in a timely manner, thereby ensuring the safety of system operation. As the service life of the battery pack is extended, the maintenance cycle and maintenance cost of the system will be reduced accordingly; the intelligent body model can automatically adjust the charging and discharging strategy according to different environments and usage conditions, so that the system has better adaptability and robustness; optimized energy management helps to reduce energy waste and has less impact on the environment.

[0026] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0027] The technical solution of the present invention is further described in detail below through the drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0029] Figure 1 It is a flow chart of the intelligent control method of photovoltaic energy storage battery group action;

[0030] Figure 2 This is a schematic diagram of the structure of the intelligent control system for photovoltaic energy storage battery group action. DETAILED DESCRIPTION

[0031] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0032] The purpose of the present invention is to provide a method and system for intelligently controlling the action of a photovoltaic energy storage battery pack. The present invention collects the status data of each battery cell in the photovoltaic energy storage battery pack in real time, and uses an intelligent body model to perform status data analysis and target predictive evaluation, so as to achieve regulation and parameter optimization of the single-cell charging and discharging action for each battery cell. The present invention accurately predicts the power supply demand of the battery pack within a predetermined time, and then ensures the best match between energy supply and demand through independent monitoring and control of the battery cells. The present invention dynamically adjusts the charging and discharging strategy of the battery cell, which not only optimizes the energy supply efficiency, but also effectively extends the overall service life of the battery pack. Through refined single-cell management, the performance and reliability of photovoltaic energy storage battery packs in complex application environments are improved.

[0033] Reference Figure 1 As shown, the photovoltaic energy storage battery group action intelligent control method provided by the embodiment of the present invention includes:

[0034] Step S101: Determine a power supply forecasting mode according to external factors such as season, estimated time period, and weather forecast, and then combine historical data belonging to the same power supply forecasting mode to predict power supply demand within a predetermined time in the future.

[0035] Specifically, external factors such as season, estimated time period, and weather forecast are used as inputs of the predictive neural network, and the output of the predictive neural network is a power supply prediction model. Combined with historical data under the same power supply prediction model, linear regression is used to predict the photovoltaic energy storage power supply demand within a predetermined time in the future.

[0036] In this step, the seasonal data is represented by s, which is the conversion of spring, summer, autumn, winter and other seasonal stages into corresponding numerical or thermal values; the estimated time period data is represented by t, which is the interval number of each hour from 0:00 to 24:00 every day; the weather forecast data is represented by w, which is an array including temperature, humidity, cloud cover, and air particle information. The input layer of the predictive neural network inputs the total feature value of the combination of seasonal data, estimated time period data and weather forecast data , the hidden layer of the predictive neural network can be set to multiple layers, each layer uses an activation function , such as ReLU or Sigmoid; the output layer outputs the predicted category quantity representing the power supply prediction mode , the output of the neural network is expressed as ,in is the weight matrix of the neural network, The historical data is used as the training set. The training samples include the input feature quantity and the corresponding category label of the power supply forecast mode. The back propagation algorithm and gradient descent are used to optimize the weight of the neural network during the training process. The trained neural network is used to take the new season, time period and weather forecast data as input to output the category of the power supply forecast mode.

[0037] In this step, according to the power supply prediction mode output by the predictive neural network, combined with the historical data under the same power supply prediction mode, linear regression is used to predict the photovoltaic energy storage power supply demand within the predetermined time in the future. Specifically, historical data points belonging to the same category as the power supply prediction mode are selected from the historical data, and the power supply demand of these historical data points is used as the dependent variable , taking time (the interval number of each hour from 0:00 to 24:00 every day) as the independent variable , then the model form of linear regression prediction is: ,in is the intercept, is the slope, is the error term, obtained through historical data points and The value of can then predict the photovoltaic energy storage power supply demand within a predetermined time in the future.

[0038] Step S102: Using multiple types of sensors, real-time parameters of each battery cell in the photovoltaic energy storage battery group are collected in real time, and a representation of the expected life span is generated for each battery cell.

[0039] Specifically, the real-time parameters of each battery cell in the photovoltaic energy storage battery pack include: voltage, current, SOC, temperature, current charge amount, current discharge amount, and internal resistance value of each battery cell.

[0040] For each battery cell, a battery cell model function is used to obtain a quantitative representation of the expected battery life based on the real-time parameters of the battery cell. The battery cell model function can be obtained by fitting the experimental data of the battery cell sample to describe the behavior and performance of the battery.

[0041] The battery cell model function can be decomposed into several sub-functions of life-influencing factors, expressed as: ;in, The quantitative representation value of the expected battery life, function Indicates voltage , Current , SOC (i.e. the percentage of battery remaining power), temperature Impact function on battery life expectancy; function Indicates the current charging amount And the current discharge amount Impact function on battery life expectancy; function Indicates internal resistance The function of the influencing factors on the expected battery life. The function of the influencing factors can be obtained by fitting the experimental data. For example, these functions can be established using linear regression, polynomial fitting or machine learning algorithms. The final quantitative representation value of the expected battery life It can be calculated by combining these functions and inputting real-time parameters.

[0042] Step S103: Based on the predicted power demand and the quantitative representation of the expected battery life of each battery unit in the photovoltaic energy storage battery group, a power demand allocation model for each battery unit is established.

[0043] Specifically, the allocation model is expressed as: ,in, For time interval Assign to The power supply requirements of each battery unit; A decision function for allocation to achieve the best match between power supply demand and expected life of battery units; For the prediction time interval Power supply requirements is the expected battery life of each battery cell.

[0044] Step S104: The power supply demand allocated to each battery cell of the battery pack is used as a state observation quantity, and input into a trained intelligent agent model for realizing optimal management of the battery pack to obtain control action parameters and target evaluation results for each battery cell of the battery pack; and the charging and discharging actions of each battery cell of the battery pack are controlled according to the control action parameters.

[0045] The intelligent agent model needs to achieve multi-objective optimization management of the battery pack. Therefore, the intelligent agent model needs to automatically optimize the charge and discharge control actions and control action parameters based on autonomous learning, and based on the evaluation of the optimization management goals, continuously update the learning model based on the new data collected, and adjust the control action strategy according to the updated model. This adaptive update capability ensures that the battery pack always operates in the most optimized way, which not only meets the energy supply and demand balance goals, but also extends the battery pack's service life and improves safety.

[0046] To this end, the modeling of the intelligent agent model requires defining the state space, control action space and target evaluation function of the battery pack, where the state space includes the power supply requirements allocated to each battery cell in the battery pack; the control action space includes the charging or discharging state action switching of each battery cell in the battery pack, as well as adjustable control action parameters such as charging or discharging power; the target evaluation function is designed based on the health status of the battery pack, its expected lifespan and factors that meet the balance between energy supply and demand.

[0047] Specifically, the state space of the agent model is represented as , the power supply requirements allocated to each battery cell of the battery pack are used as state observations in the state space ; Control action space , where the charge and discharge state action switching and charge and discharge power parameters of each battery cell in the battery pack are used as the control actions of this space ; The agent model responds to the state observation Decide on control actions The response function is defined as ,in is the network parameter tensor of the deep neural network used by the response function to implement the autonomous learning mechanism; the intelligent agent model evaluates multiple optimization goals such as energy supply and demand balance, battery life, and safety based on state observations, and defines the target evaluation function as ,in The network parameter tensor of the deep neural network that evaluates the function for this objective. The ultimate goal of agent training is to obtain the response function And the objective evaluation function , through which the network parameter tensor of the deep neural network and Assignment after training and , implement the response function The control action for the battery pack is output according to the state observation of the battery pack, and the state observation of the battery pack is evaluated by the target evaluation function Ability to obtain continuous optimization results of target evaluation.

[0048] The training process of the intelligent agent model includes: initializing the state observation of the input intelligent agent model , and initialize the response function of the intelligent agent model to determine the control action in response to the state observation , where the network parameter tensor of the deep neural network of the response function is initialized as ; Initialize the target evaluation function of the agent model , the network parameter tensor of the deep neural network of the objective evaluation function is initialized as . It is planned to perform N rounds of agent model training, and in each round n ( ), according to the time step t of this round (here , is the maximum time step of round n. The maximum time step of each round can be set to the same, or different maximum time steps can be set for each round). According to the state observation at time step t , and the network parameter tensor of the deep neural network of the response function after training and parameter adjustment at this time step t , calculate the control action of this time step t through the response function , and then observe the control action at this time step t After execution, the state observation at the next time step t+1 is . Based on the state observations at the current time step t and the next time step t+1 and , the target evaluation is performed through the target evaluation function, and the target evaluation result , , where the network parameter tensor of the deep neural network of the target evaluation function after training and parameter adjustment at time step t is ; Then, calculate the time series difference component of the target evaluation result for . Then, update the network parameter tensor of the deep neural network of the response function in the agent model at the t+1th time step ,in is a hyperparameter representing the learning rate, and ,express About parameters And, update the network parameter tensor of the deep neural network of the target evaluation function in the agent model at the t+1th time step ,in is a hyperparameter representing the learning rate, is the objective evaluation function About parameters Thus, through training of all time steps in a total of N rounds, the response function of the intelligent agent model for realizing optimal management of battery packs is obtained. ; And, obtain the target evaluation function of the agent model .

[0049] In the actual optimization control of the battery pack, the power supply demand allocated to each battery cell of the battery pack is used as the state observation , input the trained intelligent agent model response function for optimal battery management , obtain the control action parameters for each battery cell of the battery pack , that is, the charge and discharge state action switching and charge and discharge power parameters of each battery cell. According to the control action parameters output by the intelligent agent model, the charge and discharge action of each battery cell of the battery pack is controlled.

[0050] In step S103, in the actual optimization control of the battery pack, the power supply demand allocated to each battery cell of the battery pack is also synchronously used as a state observation quantity. , input the trained target evaluation function , obtain the target evaluation result. Determine whether the target evaluation result is within the threshold range of the battery pack multi-objective continuous optimization. If not, repeat the training process of the above-mentioned intelligent agent model, update the parameter tensor of the deep learning network of the intelligent agent model, and obtain the updated response function and target evaluation function.

[0051] See also Figure 2 The embodiment of the present invention provides an intelligent control system for photovoltaic energy storage battery group operation, including: a power supply demand prediction unit, a multi-type sensor unit, an expected life unit, a power supply demand allocation unit, and an intelligent body control unit.

[0052] The power demand forecasting unit is used to determine the power supply forecasting mode according to the external factors of season, estimated time period and weather forecast, and then combine the historical data under the same power supply forecasting mode to forecast the power supply demand within the future predetermined time;

[0053] Multiple types of sensor units are used to collect real-time parameters of each battery cell in the photovoltaic energy storage battery group;

[0054] An expected life unit, for obtaining a quantitative representation of the expected life of the battery according to real-time parameters of the battery cell using a battery cell model function;

[0055] A power supply demand allocation unit is used to establish a power supply demand allocation model for each battery unit according to the predicted power supply demand and the quantitative representation value of the expected battery life of each battery unit in the photovoltaic energy storage battery group;

[0056] An intelligent agent control unit is used to take the power supply demand allocated to each battery cell of the battery pack as a state observation, input it into a trained intelligent agent model for realizing optimal management of the battery pack, obtain control action parameters and target evaluation results for each battery cell of the battery pack; and control the charging and discharging actions of each battery cell of the battery pack according to the control action parameters.

[0057] The present invention dynamically adjusts the charging and discharging strategy through the intelligent body model, and can optimize the distribution and utilization of energy according to the real-time energy supply and demand status, thereby improving the energy utilization rate of the entire system; intelligent optimization management helps to avoid excessive charging and discharging of the battery and working in an inappropriate state, which can reduce battery loss and extend the service life of the battery pack. By intelligently monitoring the working status of the battery pack, potential safety risks such as overheating and overcharging can be discovered and handled in a timely manner, thereby ensuring the safety of system operation. As the service life of the battery pack is extended, the maintenance cycle and maintenance cost of the system will be reduced accordingly; the intelligent body model can automatically adjust the charging and discharging strategy according to different environments and usage conditions, so that the system has better adaptability and robustness; optimized energy management helps to reduce energy waste and has less impact on the environment

[0058] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for intelligently controlling the operation of a photovoltaic energy storage battery group, characterized in that: The following steps are involved: Step S101: Determine a power supply forecast mode according to external factors such as season, estimated time period, and weather forecast, and then combine historical data under the same power supply forecast mode to predict power supply demand within a predetermined time in the future; Step S102: using multiple types of sensors, real-time parameters of each battery cell in the photovoltaic energy storage battery group are collected in real time, and a representation of the expected life span is generated for each battery cell; Step S103: establishing a power supply demand allocation model for each battery unit according to the predicted power supply demand and the quantitative representation value of the expected battery life of each battery unit in the photovoltaic energy storage battery group; Step S104: The power supply demand allocated to each battery cell of the battery pack is used as a state observation, and input into the trained intelligent agent model for realizing optimal management of the battery pack to obtain the control action parameters and target evaluation results for each battery cell of the battery pack; the charging and discharging actions of each battery cell of the battery pack are controlled according to the control action parameters. In step S104, the state space of the intelligent agent model is expressed as , the power supply requirements allocated to each battery cell of the battery pack are used as state observations in the state space ; Control action space , where the charge and discharge state action switching and charge and discharge power parameters of each battery cell in the battery pack are used as the control actions of this space ; The agent model responds to the state observation Decide on control actions The response function is defined as ,in is the network parameter tensor of the deep neural network used by the response function to implement the autonomous learning mechanism; the intelligent agent model evaluates the target based on the state observation for multiple optimization goals such as energy supply and demand balance, battery life, and safety, and defines the target evaluation function as ,in The network parameter tensor of the deep neural network for the target evaluation function; through agent training, the response function is obtained And the objective evaluation function , through which the network parameter tensor of the deep neural network and Assignment after training and , implement the response function The control action for the battery pack is output according to the state observation of the battery pack, and the state observation of the battery pack is evaluated by the target evaluation function Ability to obtain continuous optimization results of target evaluation.

2. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 1, characterized in that: In step S101, external factors such as season, estimated time period, and weather forecast are used as inputs to a predictive neural network, the output of which is a power supply prediction model. This is combined with historical data under the same power supply prediction model, and linear regression is used to predict the photovoltaic energy storage power supply demand within a predetermined time in the future.

3. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 2, characterized in that: Seasonal data is represented by s, which is the seasonal stage converted into the corresponding numerical or thermal value code; the estimated period data is represented by t, which is the interval number of each hour from 0:00 to 24:00 every day; the weather forecast data is represented by w, which is an array including temperature, humidity, cloud cover, and air particle information; the input layer of the predictive neural network inputs the total feature value of the combination of seasonal data, estimated period data, and weather forecast data , the hidden layer of the predictive neural network is set to multiple layers, and the activation function used in each layer It is a ReLU or Sigmoid function; the output layer outputs the predicted category quantity representing the power supply prediction mode , the output of the neural network is expressed as ,in is the weight matrix of the neural network, The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture. The experimental results show that the proposed method can effectively improve the performance of power supply forecasting and the network architecture.

4. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 3 is characterized in that: Select historical data points that belong to the same category as the power supply forecast model from the historical data, and use the power supply demand of these historical data points as the dependent variable , taking time as the independent variable , then the model form of linear regression prediction is: ,in is the intercept, is the slope, is the error term, obtained through historical data points and The value of can then predict the photovoltaic energy storage power supply demand within a predetermined time in the future.

5. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 1, characterized in that: In step S102, the real-time parameters of each battery cell in the photovoltaic energy storage battery group include: voltage, current, SOC, temperature, current charge amount, current discharge amount, and internal resistance value of each battery cell.

6. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 5, characterized in that: In step S102, for each battery cell, a battery cell model function is used to obtain a quantitative representation of the expected battery life according to the real-time parameters of the battery cell; the battery cell model function is decomposed into several sub-functions of life influencing factors, which are expressed as: ;in, The quantitative representation value of the expected battery life, function Indicates voltage , Current , SOC, temperature Impact function on battery life expectancy; function Indicates the current charging amount And the current discharge amount Impact function on battery life expectancy; function Indicates internal resistance Impact function on battery life expectancy.

7. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 1, characterized in that: In step S103, the allocation model is expressed as: ,in, For time interval Assign to The power supply requirements of each battery unit; A decision function for allocation to achieve the best match between power supply demand and expected life of battery units; For the prediction time interval Power supply requirements is the expected battery life of each battery cell.

8. The method for intelligently controlling the operation of a photovoltaic energy storage battery group according to claim 1, characterized in that: The training process of the intelligent agent model includes: initializing the state observation of the input intelligent agent model , and initialize the response function of the intelligent agent model to determine the control action in response to the state observation , where the network parameter tensor of the deep neural network of the response function is initialized as ; Initialize the target evaluation function of the agent model , the network parameter tensor of the deep neural network of the objective evaluation function is initialized as ; It is scheduled to perform N rounds of agent model training, and in each round n ( ), according to the time step t of this round (here , is the maximum number of time steps in round n), according to the state observation at time step t , and the network parameter tensor of the deep neural network of the response function after training and parameter adjustment at this time step t , calculate the control action of this time step t through the response function , and then observe the control action at this time step t After execution, the state observation at the next time step t+1 is ; Based on the state observations of the current time step t and the next time step t+1 and , the target evaluation function is used to evaluate the target, and the target evaluation result , , where the network parameter tensor of the deep neural network of the target evaluation function after training and parameter adjustment at time step t is ; Then, calculate the time series difference component of the target evaluation result for ; Then, update the network parameter tensor of the deep neural network of the response function in the agent model at the t+1th time step ,in is a hyperparameter representing the learning rate, and ,express About parameters The gradient of ; and, update the network parameter tensor of the deep neural network of the target evaluation function in the agent model at the t+1th time step ,in is a hyperparameter representing the learning rate, is the objective evaluation function About parameters Gradient of; Thus, through training of all time steps in a total of N rounds, the response function of the intelligent agent model for realizing optimal management of the battery pack is obtained ; And, obtain the target evaluation function of the agent model 。 9. A photovoltaic energy storage battery group action intelligent control system, comprising: Power supply demand prediction unit, multi-type sensor unit, life expectancy unit, power supply demand allocation unit, intelligent body control unit; The power demand forecasting unit is used to determine the power supply forecasting mode according to the external factors of season, estimated time period and weather forecast, and then combine the historical data under the same power supply forecasting mode to forecast the power supply demand within the future predetermined time; Multiple types of sensor units are used to collect real-time parameters of each battery cell in the photovoltaic energy storage battery group; An expected life unit, for obtaining a quantitative representation of the expected life of the battery according to real-time parameters of the battery cell using a battery cell model function; A power supply demand allocation unit is used to establish a power supply demand allocation model for each battery unit according to the predicted power supply demand and the quantitative representation value of the expected battery life of each battery unit in the photovoltaic energy storage battery group; The intelligent agent control unit is used to take the power supply demand allocated to each battery cell of the battery pack as the state observation quantity, input the trained intelligent agent model for realizing the optimal management of the battery pack, obtain the control action parameters and target evaluation results for each battery cell of the battery pack; control the charging and discharging action of each battery cell of the battery pack according to the control action parameters, and the state space of the intelligent agent model is expressed as , the power supply requirements allocated to each battery cell of the battery pack are used as state observations in the state space ; Control action space , where the charge and discharge state action switching and charge and discharge power parameters of each battery cell in the battery pack are used as the control actions of this space ; The agent model responds to the state observation Decide on control actions The response function is defined as ,in is the network parameter tensor of the deep neural network used by the response function to implement the autonomous learning mechanism; the intelligent agent model evaluates the target based on the state observation for multiple optimization goals such as energy supply and demand balance, battery life, and safety, and defines the target evaluation function as ,in The network parameter tensor of the deep neural network for the target evaluation function; through agent training, the response function is obtained And the objective evaluation function , through which the network parameter tensor of the deep neural network and Assignment after training and , implement the response function The control action for the battery pack is output according to the state observation of the battery pack, and the state observation of the battery pack is evaluated by the target evaluation function Ability to obtain continuous optimization results of target evaluation.

Images