A method and system for generating an energy storage control strategy based on deep learning

Abstract

The application relates to the technical field of energy storage control, in particular to an energy storage control strategy generation method and system based on deep learning, which comprises receiving an energy storage operation state analysis instruction, collecting data, predicting, regulating and feeding back by using an energy storage control optimization device; an initial energy storage set is constructed based on an energy storage characteristic gradient, a monitored energy storage sequence is obtained by real-time monitoring of an operation state sequence; a monitored energy storage module is extracted, energy density parameters are confirmed, future operation trends are predicted, an operation trend sequence is optimized after abnormal fluctuations are removed, a regulation priority is calculated to obtain identified energy storage data; and the identified energy storage data is summarized to generate a target regulation strategy and a target energy storage module. The application can improve the intelligent degree of energy storage system operation state prediction under a complex power grid environment and reduce human intervention and resource waste.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of energy storage control, and specifically relates to an energy storage control strategy generation method and system based on deep learning. BACKGROUND

[0002] With the wide application of energy storage technology in power systems, the optimization of energy storage control strategies has become a key to improving system efficiency and economy. The role of energy storage systems in power grids is increasingly important, especially in power regulation, peak load shifting, frequency regulation, etc., and its performance directly affects the stability and economy of power systems. However, existing energy storage control methods mostly rely on traditional mathematical models or empirical rules, which are difficult to cope with complex and variable power grid operating environments, especially in terms of dynamic response capability, prediction accuracy and adaptive adjustment.

[0003] After searching, a patent with publication number CN110932296B proposes a control method based on a hierarchical energy storage system and virtual power plant technology, which realizes controllable output of park microgrid grid-connected point power through primary and secondary energy storage systems, and uses peak load shifting mode to ensure the stability of power distribution system. However, this technical solution mainly controls based on preset operating modes, lacks dynamic adaptive adjustment capability for real-time operating states, and is difficult to cope with complex power grid fluctuations. In addition, its control strategy does not fully utilize data-driven methods for optimization, which may result in insufficient control accuracy and response speed, and cannot meet the demand of modern power systems for efficient and intelligent energy storage control.

[0004] Another patent with publication number CN111555372B proposes a variable rate energy storage auxiliary power plant AGC frequency modulation control method, which adjusts the active power change rate of the energy storage system to optimize the charge and discharge behavior, reduces the impact of frequent charge and discharge on battery life, and improves the frequency modulation effect and economy. However, this technical solution relies on fixed energy storage capacity and unit characteristics for control, lacks prediction capability for future operating states, and may limit the frequency modulation performance under complex operating conditions. In addition, its control method does not introduce intelligent algorithms, which makes it difficult to achieve global optimization and maximize long-term benefits, limiting its flexibility and adaptability in practical applications.

[0005] The above problems show that the existing energy storage control method still has obvious deficiencies in dynamic response capability, prediction accuracy and global optimization, and it is difficult to meet the demand of modern power system for efficient and intelligent energy storage control. Therefore, an innovative energy storage control strategy generation method and system are needed to solve the above problems through advanced intelligent means. The present application provides an energy storage control strategy generation method and system based on deep learning, which aims to predict and optimize the operation state of the energy storage system in real time through deep learning algorithm, so as to improve the control accuracy, response speed and adaptive ability, and meet the urgent demand of modern power system for efficient and intelligent energy storage control. SUMMARY

[0006] The present application provides an energy storage control strategy generation method and system based on deep learning, device, electronic equipment and computer readable storage medium.

[0007] The method comprises: receiving an energy storage operation state analysis instruction, and starting an energy storage control optimization device according to the energy storage operation state analysis instruction, wherein the energy storage control optimization device comprises a data acquisition unit, a prediction unit, a regulation and control unit and a feedback unit;

[0008] Based on the pre-constructed energy storage characteristic gradient, an initial energy storage set is obtained, wherein the initial energy storage set comprises a plurality of initial energy storage modules, each initial energy storage module in the plurality of initial energy storage modules has different energy density, and the rated capacity and the charge and discharge efficiency of each initial energy storage module in the plurality of initial energy storage modules are the same;

[0009] Based on the data acquisition unit, an operation state sequence is constructed, and real-time monitoring operation is performed on the initial energy storage set based on the operation state sequence and the data acquisition unit, to obtain a monitored energy storage sequence, wherein the operation state sequence comprises a plurality of operation state values, the monitored energy storage sequence comprises a plurality of monitored energy storage modules, and the operation state values correspond one-to-one to the monitored energy storage modules;

[0010] The monitored energy storage modules are extracted from the monitored energy storage sequence in sequence, and the following operations are performed on the extracted monitored energy storage modules: confirming the energy density parameters in the extracted monitored energy storage modules to obtain energy storage energy density, predicting the future operation trend of the extracted monitored energy storage modules by using the prediction unit to obtain a prediction operation trend sequence, eliminating abnormal fluctuation sets in the prediction operation trend sequence to obtain an optimized operation trend sequence, calculating the regulation and control priority of the extracted monitored energy storage modules based on the optimized operation trend sequence, and obtaining identification energy storage data by using the regulation and control priority, the energy storage energy density, the operation state value and the extracted monitored energy storage module;

[0011] The identification energy storage data is aggregated to obtain an identification energy storage data set corresponding to a monitored energy storage sequence, a target control strategy and a target energy storage module corresponding to the target control strategy are confirmed based on the identification energy storage data set, and a deep learning-based energy storage control strategy is generated based on the target control strategy and the target energy storage module.

[0012] The application can improve the intelligent degree of energy storage system operation state prediction in a complex power grid environment and reduce human intervention and resource waste. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 A flowchart of the deep learning-based energy storage control strategy generation method provided by an embodiment of the application is shown.

[0014] Figure 2 A function module diagram of the deep learning-based energy storage control strategy generation system provided by an embodiment of the application is shown.

[0015] Figure 3 A structural diagram of an electronic device for implementing the deep learning-based energy storage control strategy generation method provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0016] The application provides a deep learning-based energy storage control strategy generation method and system, and the specific implementation is described in detail in combination with the Figure 1 , Figure 2 and Figure 3 in the description of the accompanying drawings. The following will be described from the specific execution steps of the method, the function module composition of the system, and the implementation structure of the electronic device, etc., to ensure the completeness and operability of the technical scheme.

[0017] First, as shown in Figure 1 , the application provides a deep learning-based energy storage control strategy generation method, and the core process of the method includes receiving an energy storage operation state analysis instruction, starting an energy storage control optimization device, obtaining an initial energy storage set, constructing an operation state sequence, real-time monitoring of an energy storage module, predicting future operation trends, eliminating abnormal fluctuations, calculating a control priority, obtaining identification energy storage data, and finally generating a target control strategy. In practical applications, this method can be applied to energy storage system management in a complex power grid environment, such as energy storage scheduling of a new energy power station or load balancing scenarios of a city power grid. The implementation will be described in detail in combination with the specific steps below.

[0018] When the energy storage system receives the energy storage operation state analysis instruction, the energy storage control optimization device is automatically started. The energy storage control optimization device is composed of multiple functional units, including a data acquisition unit, a prediction unit, a regulation and control unit, and a feedback unit. These units work together to complete the generation process of the energy storage control strategy. Among them, the data acquisition unit is responsible for extracting real-time operation data from the energy storage system, the prediction unit is used to predict the future operation trend of the energy storage module based on a deep learning model, the regulation and control unit formulates specific regulation and control strategies according to the prediction results, and the feedback unit is used to evaluate the execution effect of the regulation and control strategy and make dynamic adjustments. In actual operation, the energy storage control optimization device connects with the hardware interface of the energy storage system to obtain the state information of the energy storage module in real time, providing basic data support for subsequent analysis and regulation.

[0019] Next, based on the pre-constructed energy storage characteristic gradient, an initial energy storage set is obtained. The initial energy storage set is the basis for generating the energy storage control strategy, and its composition includes multiple initial energy storage modules. These energy storage modules have different energy densities, but the rated capacity and charge-discharge efficiency remain consistent. This design aims to ensure that the energy storage modules are comparable under the same conditions, while the differences in energy density reflect the performance characteristics of different modules. In practical applications, the construction of the initial energy storage set can be completed through experimental testing or historical data analysis. For example, by conducting charge-discharge cycle tests on multiple energy storage batteries, recording their energy density, charge-discharge efficiency, and other key parameters, the required energy storage modules are selected and composed into the initial energy storage set.

[0020] After obtaining the initial energy storage set, the data acquisition unit begins to construct the operation state sequence. The operation state sequence is composed of multiple operation state values, each corresponding to the real-time operation state of an energy storage module. The operation state values can be obtained by measuring physical quantities such as voltage, current, and temperature, or by calculating comprehensive indicators through algorithms. For example, the operation state value can calculate the health state index of the energy storage module through the weighted average method, thereby reflecting its current working condition. After constructing the operation state sequence, the data acquisition unit uses the sequence to perform real-time monitoring operations on the initial energy storage set, obtaining the monitored energy storage sequence. The monitored energy storage sequence contains multiple monitored energy storage modules, each corresponding to an operation state value in the operation state sequence. This process ensures that the operation state of the energy storage module can be accurately captured and recorded.

[0021] Subsequently, the monitoring energy storage modules are extracted from the monitoring energy storage sequence in turn, and a series of operations are performed on each extracted monitoring energy storage module. First, the energy density parameter of the extracted monitoring energy storage module is confirmed to obtain the energy storage energy density. The confirmation of the energy density parameter can be obtained by querying the technical specifications of the energy storage module or by experimental measurement. Next, the future operation trend of the extracted monitoring energy storage module is predicted using the prediction unit to obtain a prediction operation trend sequence. The generation of the prediction operation trend sequence relies on a deep learning model, which can be trained and predicted according to historical operation data and current operation state values. Specifically, the deep learning model can adopt an architecture such as a long short-term memory network (LSTM) or a convolutional neural network (CNN), and predict the operation trend of the energy storage module in the future period of time by learning the time series data. The formula is expressed as follows: Let the prediction operation trend sequence be , where represents the predicted value of the th time step, and where is an input feature vector, is a model parameter, is a mapping function of the deep learning model.

[0022] After obtaining the prediction operation trend sequence, abnormal fluctuation sets need to be removed to improve the reliability of the prediction results. The removal of abnormal fluctuation sets can be achieved by statistical methods or machine learning algorithms. For example, the standard deviation of the prediction operation trend sequence can be calculated , and a threshold value is set (the mean value , and the constant ), and the data points exceeding the threshold value range are regarded as abnormal fluctuations and are removed. After processing, an optimized operation trend sequence is obtained. The generation formula of the optimized operation trend sequence is , where represents the predicted value of the th time step after removing the abnormal fluctuations. Based on the optimized operation trend sequence, the regulation priority of the extracted monitoring energy storage module is further calculated. The calculation formula of the regulation priority is , where is the energy storage energy density, is a stability index of the optimized operation trend sequence, is the operation state value, , , is a weight coefficient. Through the calculation of the regulation priority, the regulation demand of each monitoring energy storage module can be quantified, thereby providing a basis for subsequent strategy generation.

[0023] After the calculation of the regulation priority is completed, the identification energy storage data is obtained by using the regulation priority, the energy density, the running state value and the extracted monitoring energy storage module. The identification energy storage data is the result of integrating the key information of the monitoring energy storage module, and its content includes but is not limited to the module number, the energy density, the regulation priority, the running state value and the like. The acquisition process of the identification energy storage data can be realized by means of data structuring, for example, storing the above information as a dictionary data in the form of key-value pair. After the identification energy storage data of all monitoring energy storage modules is summarized, the identification energy storage data set corresponding to the monitoring energy storage sequence is obtained. The construction of the identification energy storage data set provides comprehensive data support for the subsequent target regulation strategy generation.

[0024] Based on the identification energy storage data set, the target regulation strategy and the target energy storage module corresponding to the target regulation strategy are further confirmed. The generation process of the target regulation strategy includes the following steps: first, according to the regulation priority sorting in the identification energy storage data set, the energy storage module with the highest regulation demand is selected as the candidate target energy storage module; second, the regulation potential of the candidate target energy storage module is evaluated in combination with the energy density and the running state value; and finally, a specific regulation scheme is formulated by using the regulation unit, such as adjusting the charging and discharging power, switching the working mode and the like. The generation formula of the target regulation strategy is , wherein D is the identification energy storage data set, is a regulation rule parameter, is a regulation strategy generation function. After the target regulation strategy is generated, it is associated with the target energy storage module, and the generation of the energy storage control strategy based on deep learning is completed.

[0025] As shown in Figure 2 , the present application also provides an energy storage control strategy generation system based on deep learning, which comprises a data acquisition unit, a prediction unit, a regulation unit and a feedback unit. The data acquisition unit is responsible for extracting the running state data from the energy storage system, the prediction unit generates the predicted running trend sequence based on the deep learning model, the regulation unit formulates the regulation strategy according to the prediction result, and the feedback unit is used to evaluate the execution effect of the regulation strategy and make dynamic adjustment. The information exchange between the functional modules is realized through the data interface, ensuring the efficient operation of the whole system.

[0026] In addition, as shown in Figure 3 , the present application also provides an electronic device for implementing the energy storage control strategy generation method based on deep learning. The electronic device comprises a processor, a memory and a communication interface. The processor is used to perform various operations in the energy storage control strategy generation method based on deep learning, the memory is used to store program codes and running data, and the communication interface is used to interact with external devices. The hardware structure design of the electronic device fully considers the calculation requirements of the energy storage control strategy generation, ensuring that the system can stably operate in a high-load environment.

[0027] In summary, by introducing deep learning technology, the application realizes the intelligentization and automation of energy storage system operation state prediction, significantly improves the generation efficiency and accuracy of energy storage control strategy. In practical application, the application can be used for energy storage scheduling of new energy power station, load balancing of urban power grid and other scenarios, and provides strong technical support for energy storage management in complex power grid environment.

Claims

1. A method for generating energy storage control strategies based on deep learning, characterized in that, The method includes: Receive energy storage operation status analysis instructions, and start the energy storage control optimization device according to the energy storage operation status analysis instructions. The energy storage control optimization device includes a data acquisition unit, a prediction unit, a control unit and a feedback unit. An initial energy storage set is obtained based on a pre-constructed energy storage characteristic gradient. The initial energy storage set includes multiple initial energy storage modules. Each initial energy storage module is configured with a different energy density, and each initial energy storage module has the same rated capacity and charge / discharge efficiency. An operating status sequence is constructed based on the data acquisition unit. Real-time monitoring operations are performed on the initial energy storage set based on the operating status sequence and the data acquisition unit to obtain a monitoring energy storage sequence. The operating status sequence includes multiple operating status values, and the monitoring energy storage sequence includes multiple monitoring energy storage modules. The operating status values ​​correspond one-to-one with the monitoring energy storage modules. The monitoring energy storage modules are extracted sequentially from the monitoring energy storage sequence, and the following operations are performed on the extracted monitoring energy storage modules: the energy density parameters in the extracted monitoring energy storage modules are identified to obtain the energy storage energy density; the future operating trend of the extracted monitoring energy storage modules is predicted using the prediction unit to obtain the predicted operating trend sequence; the abnormal fluctuation set in the predicted operating trend sequence is removed to obtain the optimized operating trend sequence; the control priority of the extracted monitoring energy storage modules is calculated based on the optimized operating trend sequence; and the identification energy storage data is obtained using the control priority, energy storage energy density, operating status value, and the extracted monitoring energy storage modules. Summarize the identified energy storage data to obtain the identified energy storage dataset corresponding to the monitored energy storage sequence. Based on the identified energy storage dataset, identify the target control strategy and the target energy storage module corresponding to the target control strategy. Based on the target control strategy and the target energy storage module, complete the generation of the energy storage control strategy based on deep learning. The control priority of the monitoring energy storage module extracted based on the optimized operation trend sequence calculation includes: A formula for calculating the control priority is constructed. The control priority is calculated using the formula, energy storage energy density, stability index of the optimized operation trend sequence, and operation status value. The formula for calculating the control priority is P=αE+βR+γS, where α, β, and γ are weighting coefficients, E is the energy storage energy density, R is the stability index of the optimized operation trend sequence, and S is the operation status value.

2. The method for generating energy storage control strategies based on deep learning as described in claim 1, characterized in that, The construction of the operating status sequence based on the data acquisition unit includes: The system acquires real-time operating data of the energy storage system, generates multiple operating status values ​​based on the real-time operating data, and arranges the multiple operating status values ​​in chronological order to obtain an operating status sequence. The real-time operating data includes voltage, current, and temperature.

3. The method for generating energy storage control strategies based on deep learning as described in claim 1, characterized in that, The method of using a prediction unit to predict the future operating trend of the extracted monitoring energy storage module includes: The historical operating data of the extracted monitoring energy storage modules are trained using a deep learning model to generate a predicted operating trend sequence. The deep learning model uses a long short-term memory network or a convolutional neural network.

4. The method for generating energy storage control strategies based on deep learning as described in claim 1, characterized in that, The set of abnormal fluctuations to be removed from the predicted operating trend sequence includes: Calculate the standard deviation and mean of the predicted trend sequence, set a threshold range of mean plus or minus a constant multiple of the standard deviation, remove data points that exceed the threshold range, and obtain the optimized trend sequence.

5. The method for generating energy storage control strategies based on deep learning as described in claim 1, characterized in that, The step of identifying the target control strategy and the corresponding target energy storage module based on the identified energy storage dataset includes: Based on the priority ranking of the energy storage dataset, the energy storage modules with the highest control requirements are selected as candidate target energy storage modules. The control potential of the candidate target energy storage modules is evaluated by combining the energy storage energy density and operating status values, and specific control schemes are formulated using the control unit.

6. An electronic device that implements the deep learning-based energy storage control strategy generation method as described in any one of claims 1 to 5, characterized in that, The electronic device includes a processor, a memory, and a communication interface. The processor is used to execute various operations in the deep learning-based energy storage control strategy generation method. The memory is used to store program code and running data. The communication interface is used to interact with external devices.

Images