CN-121417456-B - Storage battery charge and discharge parameter intelligent optimization method and system integrating reinforcement learning

CN121417456BCN 121417456 BCN121417456 BCN 121417456BCN-121417456-B

Abstract

The application provides an intelligent optimization method and system for storage battery charging and discharging parameters fused with reinforcement learning, and belongs to the field of intelligent charging and discharging. The method comprises the steps of collecting a charge-discharge capacity sequence of a target storage battery pack, obtaining a plurality of historical operation characteristic parameter sets of a plurality of storage battery units, carrying out state evaluation on the target storage battery pack to obtain a reference state coefficient, determining reference charge-discharge parameters, carrying out charge-discharge capacity evaluation, determining allowable charge-discharge parameters of each storage battery unit to obtain a plurality of allowable charge-discharge parameters, carrying out matching degree analysis on the allowable charge-discharge parameters and the reference charge-discharge parameters to obtain a plurality of charge-discharge matching degrees, adjusting the reference charge-discharge parameters to generate adjusted charge-discharge parameters, and configuring the target storage battery pack. The dynamic self-adaptive adjustment of the charge and discharge parameters of the storage battery pack is realized through fusion reinforcement learning, the charge and discharge efficiency is improved, and the service life of the battery is prolonged.

Inventors

YU ZIYONG
Chen Guanshou
Chen Cijian
ZOU JIEXIN
LI GUOGUANG
LUO YONGSHENG
LI BAINIAN
LIANG JINYUAN
CHEN JIAJIAN

Assignees

广东顺畅科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251215

Claims (6)

1. The intelligent optimization method for the charge and discharge parameters of the storage battery fused with reinforcement learning is characterized by comprising the following steps of: The method comprises the steps of connecting a target storage battery pack, wherein the target storage battery pack comprises a plurality of storage battery units connected in series, collecting charge-discharge capacity sequences of the target storage battery pack, and obtaining a plurality of historical operation characteristic parameter sets of the storage battery units; performing state evaluation on the target storage battery pack according to the charge-discharge capacity sequence to obtain a reference state coefficient, and determining a reference charge-discharge parameter according to the reference state coefficient; performing charge and discharge capacity evaluation according to the plurality of historical operation characteristic parameter sets, and determining allowable charge and discharge parameters of each storage battery unit to obtain a plurality of allowable charge and discharge parameters; performing matching degree analysis on the allowable charge and discharge parameters and the reference charge and discharge parameters to obtain a plurality of charge and discharge matching degrees; adjusting the reference charge-discharge parameters based on the charge-discharge matching degrees, generating adjusted charge-discharge parameters, and configuring the target storage battery pack; Performing state evaluation on the target storage battery pack according to the charge-discharge capacity sequence to obtain a reference state coefficient, and determining a reference charge-discharge parameter according to the reference state coefficient, wherein the method comprises the following steps: performing state evaluation on the target storage battery pack according to the charge-discharge capacity sequence to obtain a battery pack aging coefficient; Determining the reference state coefficient based on the battery pack aging coefficient, and determining a reference charge-discharge parameter in a charge-discharge parameter mapping table according to the reference state coefficient; and performing state evaluation on the target storage battery pack according to the charge-discharge capacity sequence to obtain a battery pack aging coefficient, wherein the method comprises the following steps: constructing a storage battery state estimator according to the charge-discharge capacity sequence; inputting the charge-discharge capacity sequence into the storage battery state estimator to obtain a battery aging coefficient; Constructing a battery pack state estimator according to the charge-discharge capacity sequence, comprising: Constructing a charge aging analysis unit and a discharge aging analysis unit, wherein the charge aging analysis unit is generated based on a sample pre-charge battery pack aging coefficient, a sample charge capacity and a sample post-charge battery pack aging coefficient in a training manner, and the discharge aging analysis unit is generated based on a sample pre-discharge battery pack aging coefficient, a sample discharge capacity and a sample post-discharge battery pack aging coefficient in a training manner; Analyzing the charge-discharge capacity sequence to obtain a charge-discharge operation time sequence, wherein the charge-discharge operation time sequence is provided with a plurality of operation nodes, each operation node is provided with a corresponding operation type, and the operation type is charge operation or discharge operation; According to the operation type of each operation node in the charge-discharge operation time sequence, sequentially selecting a charge aging analysis unit or a discharge aging analysis unit for cascade combination to form the storage battery state estimator; Performing charge and discharge capability evaluation according to the plurality of historical operation characteristic parameter sets, determining allowable charge and discharge parameters of each storage battery unit, and obtaining a plurality of allowable charge and discharge parameters, wherein the method comprises the following steps: determining a first storage battery unit from the plurality of storage battery units, and acquiring a first historical operation characteristic parameter set corresponding to the first storage battery unit from the plurality of historical operation characteristic parameter sets; Carrying out data splitting on the first historical operating characteristic parameter set to obtain a first charging operating characteristic parameter set and a first discharging operating characteristic parameter set; Invoking a dual-branch reinforcement learner, wherein the dual-branch reinforcement learner comprises a charging processing branch and a discharging processing branch; Inputting the first charging operation characteristic parameter set into the charging processing branch to obtain a first allowable charging parameter; Inputting the first discharge operation characteristic parameter set into the discharge processing branch to obtain a first allowable discharge parameter; summarizing the first allowable charge parameter and the first allowable discharge parameter to obtain a first allowable charge and discharge parameter of a first storage battery unit; And determining the allowable charge and discharge parameters of the rest storage battery units according to the mode of determining the first allowable charge and discharge parameters of the first storage battery unit, so as to obtain a plurality of allowable charge and discharge parameters.
2. The method of claim 1, wherein collecting a sequence of charge and discharge capacities of a target battery pack and obtaining a plurality of historical operating characteristic parameter sets for the plurality of battery cells comprises: Collecting charge and discharge records of the target storage battery pack, and carrying out serialization processing on the charge and discharge records to obtain a charge and discharge capacity sequence; establishing preset operation characteristics, wherein the preset operation characteristics comprise preset charging characteristics and preset discharging characteristics, the preset charging characteristics comprise historical charging voltage, historical charging current and historical charging temperature, and the preset discharging characteristics comprise historical discharging voltage, historical discharging current and historical discharging temperature; and acquiring historical operation data of a plurality of storage battery units according to the preset charging characteristics and the preset discharging characteristics, and acquiring a plurality of historical operation characteristic parameter sets of the storage battery units.
3. The method of claim 1, wherein the constructing step of the dual-branch reinforcement learner comprises: Setting temperature safety constraints, wherein the temperature safety constraints comprise a charging temperature constraint and a discharging temperature constraint; collecting a plurality of sample charging operation characteristic parameter sets, performing reinforcement learning training on a plurality of sample charging records in each sample charging operation characteristic parameter set based on the charging temperature constraint, and constructing a charging processing branch; collecting a plurality of sample discharge operation characteristic parameter sets, performing reinforcement learning training on a plurality of sample discharge records in each sample discharge operation characteristic parameter set based on the discharge temperature constraint, and constructing a discharge processing branch; and integrating the charge processing branch and the discharge processing branch to obtain the double-branch reinforcement learner.
4. The method of claim 1, wherein performing a matching degree analysis on the plurality of allowable charge and discharge parameters and the reference charge and discharge parameter to obtain a plurality of charge and discharge matching degrees, comprises: vectorizing the reference charge-discharge parameter and the allowable charge-discharge parameters respectively to obtain a reference charge-discharge vector and allowable charge-discharge vectors; Constructing a charge-discharge parameter space, and calculating charge-discharge vector distances of a reference charge-discharge vector and a plurality of allowable charge-discharge vectors in the charge-discharge parameter space to obtain a plurality of charge-discharge vector distances; And determining the charge-discharge matching degree of each storage battery unit based on the plurality of charge-discharge vector distances to obtain a plurality of charge-discharge matching degrees, wherein the charge-discharge vector distances are inversely related to the charge-discharge matching degrees.
5. The method of claim 1, wherein adjusting the reference charge-discharge parameter based on the plurality of charge-discharge matches, generating an adjusted charge-discharge parameter, comprises: Sequencing the plurality of charge-discharge matching degrees, and determining the middle charge-discharge matching degree; extracting corresponding median allowable charge and discharge parameters according to the median charge and discharge matching degree; And carrying out fusion processing on the median allowable charge and discharge parameters and the reference charge and discharge parameters to generate and adjust the charge and discharge parameters.
6. A reinforcement learning-fused intelligent optimization system for battery charge and discharge parameters, for implementing the method of any one of claims 1 to 5, the system comprising: the data acquisition module is used for connecting a target storage battery pack, wherein the target storage battery pack comprises a plurality of storage battery units connected in series, acquiring charge-discharge capacity sequences of the target storage battery pack and acquiring a plurality of historical operation characteristic parameter sets of the storage battery units; The state evaluation module is used for performing state evaluation on the target storage battery pack according to the charge-discharge capacity sequence to obtain a reference state coefficient, and determining a reference charge-discharge parameter according to the reference state coefficient; the capacity evaluation module is used for evaluating the charge and discharge capacity according to the plurality of historical operation characteristic parameter sets, determining the allowable charge and discharge parameters of each storage battery unit and obtaining a plurality of allowable charge and discharge parameters; the matching analysis module is used for carrying out matching degree analysis on the plurality of allowable charge and discharge parameters and the reference charge and discharge parameters to obtain a plurality of charge and discharge matching degrees; And the parameter adjustment module is used for adjusting the reference charge-discharge parameters based on the charge-discharge matching degrees, generating adjustment charge-discharge parameters and configuring the target storage battery pack.

Description

Storage battery charge and discharge parameter intelligent optimization method and system integrating reinforcement learning Technical Field The invention relates to the field of intelligent charge and discharge, in particular to an intelligent optimization method and system for charge and discharge parameters of a storage battery integrating reinforcement learning. Background In the aspect of management of the integrated storage battery, the substation communication power supply system mainly adopts a fixed parameter setting mode, namely, unified parameters such as charge and discharge current, voltage and the like are set according to the nominal parameters of the storage battery, and the parameters are kept unchanged basically in the whole service period. However, in the long-term use process, the charge and discharge capacities of the battery cells in the battery pack may be significantly different due to the influence of factors such as manufacturing process differences, different use environments, and different aging degrees. In this case, the fixed charge and discharge parameters are continuously adopted for management, and dynamic adjustment cannot be performed based on the differentiated state of each unit, so that the problems of low charge and discharge efficiency, increased energy loss, shortened service life and the like are easily caused by overcharge or overdischarge of part of units. Disclosure of Invention Aiming at the technical problems that in the prior art, the charge and discharge parameters of the storage battery are in a fixed setting mode, and cannot be dynamically adjusted based on the differentiated state of each unit, so that the charge and discharge efficiency is low and the service life of the battery is shortened, the invention provides an intelligent optimization method and system for the charge and discharge parameters of the storage battery, which are integrated with reinforcement learning, for solving the problems. The technical scheme for solving the technical problems is as follows: The invention provides an intelligent optimization method for storage battery charge and discharge parameters of fusion reinforcement learning, which comprises the steps of connecting a target storage battery pack, collecting charge and discharge capacity sequences of the target storage battery pack, acquiring a plurality of historical operation characteristic parameter sets of the storage battery pack, carrying out state evaluation on the target storage battery pack according to the charge and discharge capacity sequences to obtain a reference state coefficient, determining the reference charge and discharge parameters according to the reference state coefficient, carrying out charge and discharge capacity evaluation on the historical operation characteristic parameter sets, determining allowable charge and discharge parameters of each storage battery pack to obtain a plurality of allowable charge and discharge parameters, carrying out matching degree analysis on the allowable charge and discharge parameters and the reference charge and discharge parameters to obtain a plurality of charge and discharge matching degrees, adjusting the reference charge and discharge parameters based on the charge and discharge matching degrees to generate an adjusted charge and discharge parameter, and configuring the target storage battery pack. The invention provides an intelligent optimization system for storage battery charge and discharge parameters of fusion reinforcement learning, which comprises a data acquisition module, a state evaluation module, a capability evaluation module and a parameter adjustment module, wherein the data acquisition module is used for connecting a target storage battery pack, the target storage battery pack comprises a plurality of storage battery units connected in series, acquiring charge and discharge capacity sequences of the target storage battery pack, acquiring a plurality of historical operation characteristic parameter sets of the plurality of storage battery units, the state evaluation module is used for carrying out state evaluation on the target storage battery pack according to the charge and discharge capacity sequences, acquiring a reference state coefficient, determining a reference charge and discharge parameter according to the reference state coefficient, the capability evaluation module is used for carrying out charge and discharge capability evaluation according to the plurality of historical operation characteristic parameter sets, determining allowable charge and discharge parameters of each storage battery unit, obtaining a plurality of allowable charge and discharge parameters, the matching analysis module is used for carrying out matching degree analysis on the allowable charge and discharge parameters and the reference charge and discharge parameters, and the parameter adjustment module is used for carrying out adjustment on the reference charge and