A big data computing method and system for battery health assessment

Abstract

The application aims to provide a big data computing method and system for battery health evaluation with wide application range and high evaluation accuracy. The big data computing method for battery health evaluation comprises the following steps: obtaining original data of a matching model of a to-be-tested battery through laboratory test, collecting real-time data of the to-be-tested battery during charging and discharging when the evaluation is performed, constructing a neural network multi-layer computing model by using the obtained original data and real-time data, performing big data analysis on the to-be-tested battery according to the computing model, and finally outputting a battery health evaluation result. The application is applied to the technical field of battery monitoring technology.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of battery monitoring technology, and particularly relates to a big data calculation method and system for battery health degree evaluation. BACKGROUND

[0002] The equipment health degree is a comprehensive evaluation index of the running state, aging degree and fault probability of the equipment. The running state of the equipment is mainly evaluated by obtaining the running parameters of the equipment. However, in actual engineering, the equipment monitoring conditions, environmental factors and the like have an influence, and few technical parameters can directly reflect the health state of the equipment, so it is difficult to select the dimension of the parameters. For electrochemical energy storage, due to the difference of the battery material, the charging and discharging performance, the operation condition and the environmental condition, it is difficult to evaluate the health state of the battery.

[0003] At the same time, due to the continuous development of the new energy industry at the present stage, the operation scale of the energy storage battery is increased. With the increase of the use scale of the energy storage battery, more technical problems are exposed. In the use process, the battery combustion, explosion and the like frequently occur, so the continuous monitoring of the running state of the battery and the accuracy of the health degree evaluation are increasingly important.

[0004] The existing health degree evaluation of the energy storage battery mainly includes the following two ways: 1) a health degree evaluation method based on the SOH value (state of health value) of the battery. The implementation of the evaluation method is based on the ratio of the capacity discharged from the full state of the battery to the cut-off voltage at a certain rate to the corresponding nominal capacity under standard conditions. However, the method based on the SOH value is only suitable for the health state of a single battery component or the battery pack with consistent initial state of the battery. Since the energy storage equipment is operated for a period of time or is a retired battery gradient utilization, the initial health states of the battery monomers in the battery pack are inconsistent, so the effect of evaluating the health degree of the battery in the battery pack based on the SOH value is limited. In addition, the SOH value evaluation is not suitable for the case of health degree decline due to sudden abnormality. 2) a health degree evaluation method based on alarm data. In the evaluation method, the alarm is used as an intuitive representation of the equipment state, the alarm data is used as a data set, the related features of the running state of the energy storage battery system are mined, and a health degree evaluation model is established, which can well avoid the problems caused by the difference of the basic parameters of the equipment state. However, the occurrence of the alarm indicates that the battery system is in an obvious abnormal state, and the effect of the device for small abnormalities is not obvious. In addition, the method based on the alarm information is only suitable for the equipment with remote signal quantity, and is not suitable for the equipment without remote signal quantity. SUMMARY

[0005] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art and provide a big data calculation method and system for battery health evaluation with wide application range and high evaluation accuracy.

[0006] The technical solution adopted by the present application is that the big data calculation method for battery health evaluation comprises the following steps:

[0007] Step S1. Obtain the original data of the matched model of the battery to be tested through laboratory test;

[0008] Step S2. Collect real-time data of the battery to be tested during charging and discharging;

[0009] Step S3. Construct a neural network multi-layer calculation model according to the data obtained in steps S1 and S2;

[0010] Step S4. Perform big data analysis of the battery to be tested according to the calculation model;

[0011] Step S5. Output the battery health evaluation result.

[0012] As can be seen from the above scheme, the original data and the real-time running data of the battery to be tested during actual work are obtained through steps S1 and S2, the preliminary operation of each performance of the battery is performed according to the data, and the neural network multi-layer calculation model is constructed through the operation result, the comprehensive evaluation of each operation result is performed, and the more accurate evaluation result is obtained. At the same time, the original data and the real-time running data are used to perform comprehensive evaluation, which can adapt to different types of battery products, and the application range is more comprehensive. By constructing a multi-dimensional fusion health state evaluation system, the battery aging characteristics can be more comprehensively reflected, thereby promoting the precise matching and efficient application of energy storage batteries.

[0013] One preferred scheme is that step S1 comprises the following specific steps: performing original charging and discharging test on the battery sample under the laboratory condition of the battery, obtaining the charging and discharging voltage and current data of the battery sample under various test environments, and classifying and entering the data according to the battery type.

[0014] One preferred scheme is that step 2 comprises the following specific steps: in the battery normal working condition running environment, collecting the charging and discharging voltage, charging and discharging current, direct current power and temperature data of the battery through the BMS battery management system.

[0015] In a preferred embodiment, the neural network multi-layer calculation model in step 3 comprises an input layer, a hidden layer and an output layer, the input layer takes the three operation data of the battery to be evaluated, i.e. the charge-discharge voltage U, the charge-discharge current I and the direct current power P obtained in step 2 as input nodes, the hidden layer comprises first-level nodes and second-level nodes, the first-level nodes take the temperature T, the humidity RH and the air pressure as classification nodes, the original data matched with the current classification nodes are retrieved as reference benchmarks according to the operation data, the second-level nodes comprise a battery capacity attenuation evaluation function, a power evaluation function, an internal resistance evaluation function and a cycle number evaluation function; the output layer comprises an excitation function, the excitation function performs comprehensive operation on the function results according to the connection weight and the excitation value of each second-level node.

[0016] The evaluation system for performing the big data calculation method comprises a Pack module arranged on each group of batteries to be monitored, a BMS battery management system connected with the batteries to be monitored, a pre-communication module in communication connection with the BMS battery management system and a server, the Pack module collects the charge-discharge operation data of the batteries to be monitored, the BMS battery management system is connected with the charge-discharge module and controls the charge-discharge module according to the operation data, the BMS battery management system communicates with the server through the pre-communication module and feeds back the operation data to the server, the server performs data reduction analysis and stores in the system real-time data, the neural network multi-layer calculation model communicates with the storage data of the server and retrieves the system real-time data for evaluation processing, and the evaluation result of the neural network multi-layer calculation model is stored in the server.

[0017] In a preferred embodiment, the evaluation system further comprises an API application interface, an external access end communicates with the server through the API application interface and obtains the evaluation result. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the system block diagram of the present application. DETAILED DESCRIPTION

[0019] As Figure 1 shown, in the present embodiment, the big data calculation method for battery health degree evaluation comprises the following steps:

[0020] Step S1. Perform original charge-discharge tests on the battery samples to be monitored under a plurality of sets of set environments in a battery laboratory condition to obtain the capacity, coulomb efficiency, overpotential, rate characteristic, cycle characteristic, high-low temperature characteristic, voltage curve characteristic, and other characteristic data of the materials of the battery samples, and then comprehensively obtain original data 1 of the battery samples, obtain the charge-discharge voltage, charge-discharge current and battery internal resistance of the battery samples under various test environments, and classify and record the data according to the type of the battery; wherein the test modes of the charge-discharge test include constant current charging mode, constant voltage charging mode, constant current discharging mode, constant resistance discharging mode, hybrid charge-discharge mode and step charge-discharge mode, and at the same time, a plurality of environmental conditions are added to each charge-discharge test for data collection, including temperature, humidity and air pressure; the original data 1 is collected through the above test mode, and then the corresponding charge-discharge voltage, charge-discharge current and battery internal resistance and other parameter data of the battery sample under each set condition are obtained, and the average performance characteristic data of all use scenarios are comprehensively obtained, so as to ensure the accuracy of the analysis result when evaluating, and at the same time, the health degree is evaluated by matching the parameters of the current battery operating environment;

[0021] Step S2. In the battery normal working condition operating environment, the BMS battery management system 2 collects the operating data detected by the sensors on the battery Pack module, and then obtains the real-time charge-discharge voltage, charge-discharge current, direct current power and temperature data of the battery, and uploads the battery operating data to the front communication module 3 of the energy management platform through the CAN bus or TCP / IP bus. The front communication module 3 classifies and stores the data in the system real-time database of the server 4 after analyzing the data;

[0022] Step S3. Construct a neural network multi-layer calculation model according to the original data 1 and the operating data obtained in steps S1 and S2, wherein the neural network multi-layer calculation model includes an input layer, a hidden layer and an output layer, the input layer takes the charge-discharge voltage U, charge-discharge current I and direct current power P of the battery to be evaluated obtained in step 2 as input nodes; the hidden layer includes first-level nodes and second-level nodes, the first-level nodes take temperature T, humidity RH, air pressure, battery operating temperature T, charge-discharge rate C, discharge depth DOD, cycle interval and charge-discharge cutoff voltage U as classification nodes, and the original data 1 matched with the current classification node is retrieved as a reference benchmark according to the operating data, the second-level nodes include a battery capacity attenuation evaluation function, a power evaluation function, an internal resistance evaluation function and a cycle number evaluation function; the output layer includes an excitation function, the excitation function performs comprehensive operation on the function result according to the connection weight and excitation value of each second-level node, and the excitation function is a root mean square error (RMSE) function:

[0023] ;

[0024] wherein, represents the actual value, represents the model predicted value, and n represents the number of data samples;

[0025] Specifically, the calculation process of the RMSE function is as follows: 1) for each sample, calculate the difference between the predicted value and the actual value; 2) square the difference for all samples; 3) average all squared differences; 4) take the square root of the average to get the RMSE value;

[0026] The modification and adjustment of the hidden layer node function are performed by replacing different hidden layer node functions to compare their effects on the model performance, adjusting the parameters of the hidden layer node function, such as the slope parameter of the Sigmoid function or the threshold parameter of the ReLU function, to optimize the model performance, and increasing or decreasing the number of hidden layer nodes to explore different model structures and performance. The selection and adjustment of the hidden layer node function need to be repeatedly tested, modified and adjusted to find the best model structure and performance;

[0027] Step S4. Perform big data analysis on the battery to be tested according to the calculation model;

[0028] Step S5. Output the battery health degree evaluation result.

[0029] wherein, the battery capacity attenuation evaluation function in step S3 is: ; wherein, Caged is the current capacity of the battery, which is obtained by the BMS battery management system 2; Crated is the rated capacity of the battery, which is obtained by battery laboratory test;

[0030] The power evaluation function is: ; wherein, Qaged-max is the maximum discharge power of the current battery, which is obtained by the BMS battery management system 2; Qnew-max is the maximum discharge power of the new battery, which is obtained by battery laboratory test;

[0031] The internal resistance evaluation function is: ; wherein, Rc is the internal resistance of the current battery, REOL is the internal resistance at the end of the battery life, which is obtained by the BMS battery management system 2; Rnew is the internal resistance of the new battery, REOL and Rnew are obtained by battery laboratory test;

[0032] The cycle number evaluation function is: ; wherein, Cntremain is the remaining cycle number of the battery, which is obtained by the BMS battery management system 2; Cnttota is the total cycle number of the battery, which is obtained by battery laboratory test;

[0033] In the above formula, SOH is the evaluation result value.

[0034] The big data analysis in step S4 includes three evaluation algorithms, namely battery health evaluation, battery safety evaluation, and capacity attenuation evaluation. The evaluation algorithm adopts a root mean square (RMSE) error function. The calculation model first preprocesses and extracts features from the input battery data, and then divides the data into a training set and a test set. Next, a random forest regression model is used for training, and prediction is performed on the test set. Finally, the root mean square error between the predicted result and the actual result is calculated and output as the evaluation result.

[0035] From the entire evaluation module, by reading the charge and discharge history data of the battery operation, through big data intelligent analysis technology, health degree analysis and evaluation technology, capacity attenuation monitoring and evaluation technology, and battery safety performance analysis and evaluation technology, the health degree of the battery can be quickly evaluated, and the evaluation result is finally given, including detailed evaluation results of battery performance and performance classification.

[0036] The evaluation system for performing the big data calculation method for battery health evaluation includes a Pack module arranged on each group of batteries to be monitored, a BMS battery management system 2 connected with the batteries to be monitored, a front communication module 3 in communication connection with the BMS battery management system 2, a server 4, and an API application interface 5. The Pack module collects the charge and discharge operation data of the batteries to be monitored. The BMS battery management system 2 is connected with the charge and discharge module and controls the charge and discharge module according to the operation data. The BMS battery management system 2 communicates with the server 4 through the front communication module 3 and feeds back the operation data to the server 4. The server 4 performs data specification analysis and classifies and stores the system real-time data. The neural network multi-layer calculation model communicates with the storage data of the server 4 and calls the system real-time data for evaluation processing. The evaluation result of the neural network multi-layer calculation model is stored in the server 4. An external access terminal communicates with the server 4 through the API application interface 5 and obtains the evaluation result.

[0037] The API application interface 5 includes remote procedure call RPC, standard query language SQL, file transfer, and information delivery. Remote procedure call RPC is a process or task that realizes inter-process communication by acting on a shared data buffer; standard query language SQL is a standard query language for accessing data, which realizes data sharing between application programs through a database; file transfer is a method of realizing data sharing between application programs by sending formatted files; information delivery refers to small formatted information between loosely coupled or tightly coupled application programs, which realizes data sharing through direct communication between programs.

[0038] Although the embodiments of the present application are described in the practical schemes, they do not constitute the limitation to the meaning of the present application, and the modification to the embodiments thereof and the combination with other schemes according to the present specification are obvious to the person skilled in the art.

Claims

1. A big data computing method for battery state of health estimation, characterized in that, It comprises the following steps: Step S1. Obtain the original data of the battery to be tested by laboratory test; Step S2. Collect real-time data of the battery to be tested during charging and discharging; Step S3. Construct a neural network multi-layer calculation model according to the data obtained in steps S1 and S2; Step S4. Perform big data analysis on the battery to be tested according to the calculation model; Step S5. Output the battery health assessment result; The neural network multi-layer calculation model in step 3 includes an input layer, a hidden layer and an output layer, the input layer takes the charging and discharging voltage U, charging and discharging current I and direct current power P of the battery to be evaluated obtained in step 2 as input nodes, the hidden layer includes first level nodes and second level nodes, the first level nodes take temperature T, humidity RH and air pressure as classification nodes, and the original data matched with the current classification node is retrieved as a reference benchmark according to the running data, the second level nodes include battery capacity attenuation evaluation function, power evaluation function, internal resistance evaluation function and cycle number evaluation function; The output layer includes an excitation function, which performs comprehensive operation on the function result according to the connection weight and excitation value of each second level node. 2.The big data computing method for battery state of health estimation according to claim 1, wherein, Step S1 includes the following specific steps: original charging and discharging test of battery samples under laboratory conditions, obtaining charging and discharging voltage and current data of battery samples under various test environments, and classifying and recording data according to battery type. 3.The big data computing method for battery state of health estimation according to claim 1, wherein, Step 2 includes the following specific steps: in the battery normal working condition environment, the BMS battery management system collects the charging and discharging voltage, current, direct current power and temperature data of the battery.

4. An evaluation system for performing the big data computing method for battery health evaluation according to any one of claims 1 to 3, characterized by, It includes a Pack module arranged on each group of batteries to be monitored, a BMS battery management system connected with the batteries to be monitored, a pre-communication module in communication connection with the BMS battery management system, and a server, the Pack module collects the charging and discharging operation data of the battery to be monitored, the BMS battery management system is connected with the charging and discharging module and controls the charging and discharging module according to the operation data, the BMS battery management system communicates with the server through the pre-communication module and feeds back the operation data to the server, the server performs data specification analysis and classifies and stores in the system real-time data, the neural network multi-layer calculation model communicates with the storage data of the server and retrieves the system real-time data for evaluation processing, and the evaluation result of the neural network multi-layer calculation model is stored in the server.

5. The evaluation system of claim 4, wherein, It also includes an API application interface, an external access end communicates with the server through the API application interface and obtains the evaluation result.

Images