A lithium battery safety operation and maintenance management system and a lithium battery health state evaluation method

Abstract

The application discloses a lithium battery safe operation and maintenance management system and a lithium battery health state evaluation method, and overcomes the problem that in the prior art, when detecting lithium battery faults, the safety hazards of the lithium battery are not considered, the detection result is inaccurate, and when predicting the health state of the lithium battery, the correlation between different characteristics is ignored, resulting in low accuracy of the predicted result. The system comprises a lithium battery series-parallel network comprising a plurality of battery groups, realizes automatic disconnection when a certain battery group has a safety hazard, and replaces the battery group with a backup battery group; a lithium battery controller collects lithium battery operation parameters of the lithium battery series-parallel network and uploads the collected lithium battery operation parameters to a lithium battery intelligent operation and maintenance management unit; and the lithium battery intelligent operation and maintenance management unit completes online operation monitoring and health state evaluation of the lithium battery series-parallel network. The lithium battery health state prediction accuracy is improved, and safe and economic operation of a lithium battery energy storage system is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lithium battery safe operation and maintenance, and particularly relates to a lithium battery safe operation and maintenance management system and a lithium battery health state evaluation method. BACKGROUND

[0002] Lithium ion battery energy storage systems have the advantages of high energy efficiency and fast response speed, and have been widely applied in the fields of smart energy network construction, terminal energy electrification and large-scale renewable energy access. With the increasing size of energy storage systems, it has become a trend to realize intelligent and safe operation and maintenance of lithium battery energy storage systems. However, due to the high requirements of lithium batteries on the use environment and conditions, safety accidents may occur sometimes, and with the increase of use times, lithium batteries may have safety hazards or obvious degradation. Therefore, it is necessary to monitor the online safe operation status of lithium batteries in real time, track and evaluate the health state (SOH) and remaining useful life (RUL) in real time, and quickly switch to backup batteries when there is performance degradation or safety hazards in the lithium battery pack, so as to ensure the safe and normal operation of the system.

[0003] At present, the research on the safe operation of lithium batteries mainly includes online fault detection and remaining life prediction. Online fault detection mainly monitors the battery charging voltage and temperature, and when the voltage and temperature exceed a certain threshold range, fault alarm and backup battery switching are performed, and the remaining capacity (SOC) is calculated through the battery current. However, this method does not consider the potential health hazards of the battery.

[0004] The health state prediction of lithium batteries mainly includes model-based methods, however, the models in these methods mostly directly stack LSTM networks or CNNs or simply concatenate different features extracted, on the one hand, ignoring the correlation between different features, resulting in low accuracy of the prediction results, on the other hand, not fully utilizing the characteristics of different network structure layers.

[0005] Moreover, single lithium battery cannot meet the requirements of actual application scenarios, and in actual use, multiple lithium batteries are usually connected in series and parallel in a fixed manner to form a lithium battery pack or a lithium battery network. However, due to the safety hazards of lithium batteries caused by long-term overcharging, overdischarging and long-term high-temperature operation, once a single lithium battery has performance degradation, the fixed series and parallel connection method can only stop the entire lithium battery network from running, and cannot realize the safe and economic operation of the lithium battery energy storage system. SUMMARY

[0006] The purpose of the present application is to overcome the problems in the prior art that the safety hazards of lithium batteries are not considered when detecting lithium battery faults, the detection results are inaccurate, and the correlation between different characteristics is ignored when predicting the health state of lithium batteries, resulting in low accuracy of the predicted results. The present application provides a lithium battery safe operation and maintenance management system and a lithium battery health state evaluation method. The GRU is introduced into time series analysis, the time series of different health factors are considered comprehensively, the CNN network is used to process the GRU feature map after splicing and fusion, so that the network can learn the correlation between the health factors, and the prediction accuracy of the health state of the lithium battery is improved. At the same time, the lithium battery pack is used to find the safety hazards of the lithium battery pack in real time, and the lithium battery can be switched when a fault occurs, so that the whole network can still operate normally, and the safe and economic operation of the lithium battery energy storage system is realized.

[0007] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0008] A lithium battery safe operation and maintenance management system comprises:

[0009] A lithium battery series-parallel network comprises a plurality of battery packs. Under the control of a lithium battery intelligent operation and maintenance unit, when a safety hazard exists in a certain battery pack, the battery pack can be automatically cut off and replaced by a backup battery pack.

[0010] A lithium battery controller collects lithium battery operation parameters of the lithium battery series-parallel network and uploads the collected lithium battery operation parameters to the lithium battery intelligent operation and maintenance unit.

[0011] The lithium battery intelligent operation and maintenance unit completes online operation monitoring and health state evaluation of the lithium battery series-parallel network according to the lithium battery operation parameters collected by the lithium battery controller.

[0012] A lithium battery pack controller is designed to collect lithium battery online operation parameters, find safety hazards of the lithium battery pack in real time, and find sudden safety faults in time. The designed lithium battery network structure can flexibly switch backup battery packs. When there is a sudden safety accident or the performance of the lithium battery deteriorates seriously, the backup battery pack can be dynamically and quickly switched, so that the whole lithium battery network can still operate, not only preventing problems from occurring, but also providing a new way for the economic operation of the lithium battery energy storage system, and realizing the safe and economic operation of the lithium battery energy storage system.

[0013] As a preferred embodiment, the lithium battery series-parallel network comprises M rows and N columns of lithium battery packs, each lithium battery pack is composed of a plurality of single batteries connected in series, the lithium battery packs in each column are connected in parallel, and the last column of lithium battery packs serves as a backup column; the lithium battery packs in each row are connected in series; the last row of each backup battery pack is a backup battery pack of the column of battery packs; each battery pack and each column of battery packs are provided with a switching switch.

[0014] Each battery pack and each column of battery packs is configured with a switching switch, in normal operation, the lithium battery network has M-1 rows and N-1 columns, when a certain battery pack has a safety hazard, it is automatically cut off, and a backup battery pack in the column is used to replace it; when the batteries in a column need to be maintained, the column is cut off and switched to a backup column.

[0015] As preferred, the lithium battery controller comprises:

[0016] The battery pack monitoring module collects the charging voltage and current of the lithium battery in the lithium battery series-parallel network, and transmits the collected voltage and current data to the control module;

[0017] The temperature acquisition module collects the temperature of the lithium battery in the lithium battery series-parallel network, and transmits the collected temperature data to the control module;

[0018] The control module starts the heat dissipation module of the lithium battery controller to cool down when the temperature of the lithium battery exceeds a threshold value.

[0019] The control module is used to realize the operation and maintenance monitoring of the lithium battery series-parallel network, the lithium battery charging voltage and current collected by the battery pack monitoring module and the lithium battery temperature collected by the temperature acquisition module are all transmitted to the control module, and the heat dissipation module is controlled by the control module. The lithium battery controller transmits the collected lithium battery operation parameters (charging voltage, charging current and battery surface temperature) to the lithium battery intelligent operation and maintenance management unit of the upper computer through a communication interface (such as a CAN interface).

[0020] As preferred, the lithium battery intelligent operation and maintenance management unit comprises:

[0021] The online operation monitoring module judges whether the single lithium battery has a safety hazard according to the operation parameters of the lithium battery, and disconnects the lithium battery pack with a safety hazard and switches to a backup battery pack;

[0022] The lithium battery health state evaluation module estimates the capacity of the lithium battery according to the operation parameters of the lithium battery, and predicts the state of health SOH and the remaining useful life RUL of the lithium battery, disconnects the lithium battery pack with serious performance degradation, and switches to a backup battery pack.

[0023] The present application provides a kind of efficient and practical lithium battery network safety operation and maintenance system, can find lithium battery pack safety hazard in real time, and the lithium battery network structure designed can flexibly switch backup battery pack.

[0024] A lithium battery health state evaluation method comprises the following steps:

[0025] S1: construct a lithium battery health state evaluation model based on GRU-CNN;

[0026] S2: taking the health factor of the lithium battery as an input time sequence of the evaluation model, processing the input time sequence in parallel by using a gated recurrent module GRU of the evaluation model, extracting time sequence features, predicting the health factor at the next moment, and mapping the health factor features extracted by the GRU to a higher-dimensional feature space by using a CNN;

[0027] S3: predicting the health state SOH of the lithium battery and the remaining useful life RUL of the lithium battery by using the lithium battery health state evaluation model, and realizing the lithium battery health state evaluation.

[0028] Firstly, the lithium battery health state evaluation model based on GRU-CNN is constructed, then the health factors of the lithium battery including the charging current, the charging voltage and the battery temperature of the lithium battery are used as the input features of the model, and finally the degradation condition of the lithium battery (the health state SOH and the remaining useful life RUL of the lithium battery) is predicted in advance by the lithium battery health state evaluation model.

[0029] The lithium battery health state evaluation model takes the health factor as the input time sequence of the model, and the gated recurrent unit GRU processes the time sequence in parallel, extracts the time sequence features, and then predicts the health factor at the next moment. The CNN is used to map the health factor features extracted by the GRU to a higher-dimensional feature space. The GRU is introduced into the time sequence analysis, the time sequences of different health factors are comprehensively considered, the CNN network is used to process the GRU feature map after splicing and fusion, so that the network can learn the correlation between the health factors. The prediction error is smaller, and the accuracy of the lithium battery health state evaluation and prediction is improved.

[0030] As a preferred, in the step S3, the lithium battery health state SOH is predicted by using the lithium battery health state evaluation model, which further comprises:

[0031] A1: obtaining the lithium battery monitoring data, and performing z-score standardization preprocessing on the original lithium battery monitoring data to unify the dimension and obtain the lithium battery health factor;

[0032] A2: extracting the time sequence feature information of the lithium battery health factor by using the gated recurrent module GRU, and forming a new indirect health factor;

[0033] A3: splicing the extracted indirect health factor as the input of the CNN network;

[0034] A4: outputting the predicted battery capacity by using a full connection network, calculating the remaining capacity of the lithium battery predicted by the evaluation model by using the ratio of the predicted battery capacity and the initial capacity, and obtaining the predicted value of the lithium battery health state SOH.

[0035] The lithium battery health state evaluation model of the GRU-CNN provided in the application firstly uses mutually parallel GRU modules to extract time sequence information of current, voltage and temperature respectively, avoids manual feature extraction, and can improve the robustness of the system. The three extraction branches run in parallel and do not interfere with each other, and are finally spliced together through feature fusion technology to form higher-dimensional features. The CNN network can make up for the lack of spatial features of GRU. Since there is an inherent relationship between different monitoring data during the charging and discharging process of the lithium battery, pure GRU cannot extract information across features, so the CNN network is used to make up for the missing spatial features. The three direct health factors in the charging and discharging cycle of the lithium battery, namely the charging current, the charging voltage and the temperature, are comprehensively considered, and the time sequence of the three monitoring data is used as the input of the GRU (Gated Recurrent Unit) network to extract the indirect health factors, and then the indirect health factors are fitted with the SOH through the CNN (Convolutional Neural Network) network.

[0036] Preferably, in step A1, the time sequence of the lithium battery monitoring data is linearly changed by using deviation standardization to fall the numerical result in the [0, 1] interval; the deviation standardization calculation formula is:

[0037]

[0038] wherein x t is the monitoring data (current, voltage or temperature) at time t, x max is the maximum value of the monitoring data, and x min is the minimum value of the monitoring data.

[0039] The original lithium battery monitoring data is preprocessed by z-score standardization to unify the dimension.

[0040] Preferably, in the extraction of time sequence feature information, a fixed length sliding window is used to extract a fixed length vector in sequence; each GRU (Gated Recurrent Unit) module contains two GRU units, and each GRU unit upgrades the received time sequence vector.

[0041] GRU (Gate Recurrent Unit) is a variant of LSTM (Long Short-Term Memory) and is a recurrent neural network model widely used in time sequence processing. Under the condition that all hyperparameters are optimized, GRU can obtain performance comparable to LSTM with fewer parameters, and the structure of GRU is simpler. Compared with LSTM, GRU has better prediction effect when the training sample is small. The application combines GRU and CNN to construct a lithium battery health state evaluation model.

[0042] As preferred, the predicting the remaining useful life RUL of the lithium battery by using the lithium battery health state evaluation model further comprises: calculating the number of charge-discharge cycles required for the lithium battery to decay from the current remaining capacity to the failure threshold after a certain number of charge-discharge cycles by using the predicted remaining capacity of the lithium battery, and predicting the remaining useful life RUL of the lithium battery.

[0043] The GRU-CNN-based lithium battery health state evaluation model of the application can well model the relationship between the time sequence characteristics of the health factors and the capacity change of the lithium battery, and improve the prediction accuracy.

[0044] Therefore, the application has the following beneficial effects: 1. The GRU is introduced into the time series analysis, the time series of different health factors is comprehensively considered, the CNN network is used to process the spliced and fused GRU feature map, so that the network can learn the correlation between the health factors, and the lithium battery health state prediction accuracy is improved; 2. The lithium battery pack is adopted, and the safety hidden danger of the lithium battery pack can be found in real time, the lithium battery can be switched when a fault occurs, the whole network can still operate normally, and the safe and economic operation of the lithium battery energy storage system is realized; 3. The safe operation of the lithium battery is realized in the safety line monitoring and the accurate and rapid evaluation of the health state, which provides technical support for the reliable, stable and safe operation of the lithium battery energy storage system. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a system structure schematic diagram of the lithium battery safe operation and maintenance management system in the application.

[0046] Figure 2 is a step flow chart of the lithium battery health state evaluation method in the application.

[0047] Figure 3 is a structure schematic diagram of the GRU-CNN-based lithium battery health state evaluation model in the application.

[0048] Figure 4 is a MAE value view of the B0005 battery pack in the embodiment with different training sample numbers.

[0049] Figure 5 is the RUL prediction result of the B0005 battery pack in the embodiment.

[0050] Figure 6 is the RUL prediction result of the B0006 battery pack in the embodiment.

[0051] In the figure: 1, lithium battery series-parallel network; 2, lithium battery controller; 3, lithium battery intelligent operation and maintenance management unit; 4, lithium battery pack; 5, temperature acquisition module; 6, battery pack monitoring module; 7, heat dissipation module; 8, control module; 9, online operation monitoring module; 10, lithium battery health state evaluation module. DETAILED DESCRIPTION

[0052] The application will be further described in detail below in combination with the accompanying drawings and specific embodiments:

[0053] Embodiment one:

[0054] This embodiment is a lithium battery safe operation and maintenance management system, as shown in the figure, comprising: a lithium battery series-parallel network 1, a lithium battery controller 2 and a lithium battery intelligent operation and maintenance management unit 3, the lithium battery series-parallel network is connected with the lithium battery controller, the lithium battery controller is connected with the lithium battery intelligent operation and maintenance management unit. Figure 1

[0055] The lithium battery series-parallel network comprises a plurality of battery packs, under the control of the lithium battery intelligent operation and maintenance management unit, when a certain battery pack has a safety hidden danger, it is automatically cut off and replaced by a standby battery pack; the lithium battery controller collects lithium battery operation parameters of the lithium battery series-parallel network and uploads the collected lithium battery operation parameters to the lithium battery intelligent operation and maintenance management unit; the lithium battery intelligent operation and maintenance management unit completes online operation monitoring and health state evaluation of the lithium battery series-parallel network according to the lithium battery operation parameters collected by the lithium battery controller.

[0056] The lithium battery pack controller designed in the application collects lithium battery online operation parameters, can find safety hidden dangers of lithium battery packs in real time and find sudden safety failures in time; the designed lithium battery network structure can flexibly switch standby battery packs, when there is a sudden safety accident or lithium battery performance degradation is serious, can dynamically and quickly switch standby battery packs, so that the whole lithium battery network can still operate, not only prevents troubles from occurring, but also provides a new way for economic operation of the lithium battery energy storage system, realizes safe and economic operation of the lithium battery energy storage system.

[0057] Specifically:

[0058] 1, lithium battery series-parallel network.

[0059] The lithium battery series-parallel network comprises M rows and N columns of lithium battery packs, each lithium battery pack is composed of a plurality of single batteries connected in series (in this embodiment, each lithium battery pack is composed of 12 single batteries connected in series); the lithium battery packs in each column are connected in parallel, and the last column of lithium battery packs is used as a standby column; the lithium battery packs in each row are connected in series; the last row of each column of standby battery packs is a standby battery pack of the column of battery packs; each battery pack and each column of battery packs are provided with a switching switch. ​

[0060] Normal operation, lithium battery network has M-1 row, N-1 column, when a certain battery pack exists security risks, will be automatically cut off, and use the spare battery pack in the column to replace; when a column of battery needs to be maintained, the column will be cut off and switched to the standby column.

[0061] 2. Lithium battery controller.

[0062] The lithium battery controller comprises a battery pack monitoring module 6, a temperature acquisition module 5, a control module 8 and a heat dissipation module 7, the battery pack monitoring module and the temperature acquisition module are connected with the lithium battery pack, the control module is connected with the battery pack monitoring module, the temperature acquisition module and the heat dissipation module, and the heat dissipation module is connected with the lithium battery pack.

[0063] The battery pack monitoring module is used for collecting the charging voltage and current of the lithium battery in the lithium battery series-parallel network and transmitting the collected voltage and current data to the control module, the battery pack monitoring module in the embodiment adopts an LTC6804 chip and a peripheral circuit connected with the chip; the temperature acquisition module is used for collecting the temperature of the lithium battery in the lithium battery series-parallel network and transmitting the collected temperature data to the control module; the control module starts the heat dissipation module of the lithium battery controller to cool down when the temperature of the lithium battery exceeds a threshold value, and the control module in the embodiment adopts a single-chip microcomputer.

[0064] The lithium battery controller transmits the collected lithium battery operation parameters to the lithium battery intelligent operation and maintenance management system software of the upper computer through a communication interface; the lithium battery intelligent operation and maintenance management system software serves as a lithium battery intelligent operation and maintenance management unit.

[0065] 3. Lithium battery intelligent operation and maintenance management unit.

[0066] The lithium battery intelligent operation and maintenance management unit comprises an online operation monitoring module 9 and a lithium battery health state evaluation module 10, the online operation monitoring module is connected with the lithium battery controller and the lithium battery series-parallel network respectively, and the lithium battery health state evaluation module is connected with the lithium battery controller and the lithium battery series-parallel network respectively.

[0067] When working, the online operation monitoring module judges whether a single lithium battery exists a security risk according to the operation parameters of the lithium battery, and disconnects the lithium battery pack existing the security risk and switches to a standby battery pack; the lithium battery health state evaluation module estimates the capacity of the lithium battery according to the operation parameters of the lithium battery, and predicts the state of health SOH and the remaining useful life RUL of the lithium battery, disconnects the lithium battery pack with serious performance degradation and switches to a standby battery pack.

[0068] The application can monitor the voltage and surface temperature of the lithium battery in real time, find the safety hidden danger of the lithium battery in time, and the lithium battery pack safety operation controller can quickly switch the standby battery pack when the lithium battery pack has performance degradation or safety hidden danger, so that the safe operation of the lithium battery is realized, the safety line monitoring and the accurate and rapid evaluation of the health state are realized, and the technical support is provided for the reliable, stable and safe operation of the lithium battery energy storage system.

[0069] The high-efficiency and practical lithium battery network safety operation system provided by the application can not only monitor the online safety operation state of the lithium battery in real time, but also accurately evaluate the health state of the lithium battery in time, and automatically switch the standby battery when the safety hidden danger is found, so that the safe and normal operation of the system is ensured.

[0070] The embodiment also provides a lithium battery health state evaluation method, as shown in the following formula (1) : Figure 2 The method comprises the following steps: first, constructing a GRU-CNN-based lithium battery health state evaluation model; second, taking the health factors of the lithium battery as the input time sequence of the evaluation model, using the gate recurrent module GRU of the evaluation model to process the input time sequence in parallel, extracting the time sequence features, predicting the health factors at the next moment, and using CNN to map the health factor features extracted by the GRU to a higher-dimensional feature space; third, using the lithium battery health state evaluation model to predict the lithium battery health state SOH and the lithium battery remaining useful life RUL, and realizing the lithium battery health state evaluation.

[0071] The application first constructs a GRU-CNN-based lithium battery health state evaluation model, then uses the health factors of the lithium battery (including the charging current, charging voltage and battery temperature of the lithium battery) as the input features of the model, and finally uses the lithium battery health state evaluation model to predict the degradation state (the health state SOH and the remaining useful life RUL of the lithium battery) of the lithium battery in advance. The GRU is introduced into the time sequence analysis, the time sequences of different health factors are comprehensively considered, the CNN network is used to process the spliced and fused GRU feature map, so that the network can learn the correlation between the health factors. The prediction error is smaller, and the accuracy of the lithium battery health state evaluation and prediction is improved.

[0072] The lithium battery health state evaluation method of the application will be further described below.

[0073] First step: constructing a GRU-CNN-based lithium battery health state evaluation model.

[0074] The structure of the GRU-CNN-based lithium battery health state evaluation model provided by the application is as shown in the following formula (2) : Figure 3As shown, the embodiment selects three health factors, charging voltage, current and temperature, as input time series of the model, three GRUs respectively and in parallel process the three time series and extract time sequence features, and then predict the current, voltage and temperature values at the next moment.

[0075] Second step: Take the health factors of the lithium battery as the input time series of the evaluation model, use the GRU of the evaluation model to process the input time series in parallel, extract time sequence features, and predict the health factors at the next moment. The CNN is used to map the health factor features extracted by the GRU to a higher dimensional feature space.

[0076] Considering that GRU is not good at inducing these prediction data and fitting the relationship between them and the capacity. The embodiment uses CNN to map the health factor features extracted by the GRU to a higher dimensional feature space. Since CNN can capture the local spatial features of the spliced feature map, and GRU can extract the time sequence features of the global sequence, the GRU-CNN based lithium battery health state evaluation model can well model the relationship between the time sequence features of the health factors and the capacity change of the lithium battery, and improve the prediction accuracy.

[0077] Third step: Use the lithium battery health state evaluation model to predict the lithium battery health state SOH and the remaining useful life RUL of the lithium battery, and realize the lithium battery health state evaluation.

[0078] 1. Use the lithium battery health state evaluation model to predict the lithium battery health state SOH.

[0079] A1: Obtain the lithium battery monitoring data, and perform z-score standardization preprocessing on the original lithium battery monitoring data to unify the dimension and obtain the lithium battery health factors, thereby improving the comparability of the data;

[0080] For the time series of the lithium battery monitoring data, linear change is performed by using the deviation standardization to make the numerical results fall within the [0, 1] interval. The deviation standardization calculation formula is:

[0081]

[0082] Wherein, x t is the monitoring data (current, voltage or temperature) at t moment, x max is the maximum value of the monitoring data, and x min is the minimum value of the monitoring data.

[0083] A2: Extract the time series feature information of the three health factors of the lithium battery by using three parallel gated recurrent module GRUs, and form a new indirect health factor; in this embodiment, when extracting the time series feature information, a fixed-length sliding window is used to extract a fixed-length vector in turn; each gated recurrent module GRU includes two GRU units, and each GRU unit upgrades the received time series vector.

[0084] A3: The extracted indirect health factor is spliced as the input of the CNN network, so that the network can more sensitively capture more in-depth local features, and can use the spatial features of the spliced feature map to improve the prediction accuracy of the network.

[0085] A4: The predicted battery capacity is output by a fully connected network, and the predicted lithium battery remaining capacity is calculated by using the ratio of the predicted battery capacity to the initial capacity, to obtain the predicted value of the lithium battery state of health SOH.

[0086] The state of health SOH of the lithium battery is mainly represented by physical quantities such as the internal resistance, capacity or peak power of the lithium battery, and is used to judge the performance degradation degree of the lithium ion battery. The capacity of the lithium ion battery has a good continuous degradation trend, and the capacity is an important parameter that can directly represent the current storage capacity of the lithium ion battery, and is widely used as an evaluation index of the state of health of the lithium ion battery. Generally, the predicted value of the SOH of the lithium battery is calculated by the ratio of the predicted battery capacity to the initial capacity by the state of health evaluation model.

[0087] In this embodiment, firstly, the time series of 3 health factors: charging voltage, current and temperature are selected and input into the lithium battery state of health evaluation model of GRU-CNN, and then the battery capacity predicted by the model is converted into the state of health SOH value of the lithium battery. Taking the SOH value as the true value index of the prediction can reduce the influence of different cycle modes and different remaining life of the lithium battery, and further improve the accuracy of the remaining life prediction.

[0088] 2, The lithium battery remaining useful life RUL is predicted by using the lithium battery state of health evaluation model.

[0089] The remaining useful life RUL of the lithium battery refers to the number of charge and discharge cycles required for the lithium battery to decay from the current remaining capacity to the failure threshold after a certain number of charge and discharge cycles.

[0090] Based on the relationship between SOH and RUL, in this embodiment, the SOH data in the original cycle data is converted into RUL data and re-used as the label value for network training, and the RUL is predicted.

[0091] Using the predicted lithium battery remaining capacity, the number of charge and discharge cycles required for the lithium battery to decay from the current remaining capacity to the failure threshold after a certain number of charge and discharge cycles is calculated, and the remaining service life RUL of the lithium battery is predicted.

[0092] The lithium battery health state evaluation method of the present application is analyzed below through specific examples:

[0093] This embodiment uses the NASA 18650 type lithium battery data set, which includes a set of four lithium ion batteries (B0005, B0006, B0007, B0018) running at room temperature under three different operating modes of charging, discharging and impedance. The charging mode is carried out at a constant current of 1.5A until the battery voltage reaches 4.2V, and then continues to charge in constant voltage mode until the charging current decreases to 20mA. The discharge mode is carried out at a constant current of 2A until the voltage of the four batteries decreases to 2.7V, 2.5V, 2.2V and 2.5V, respectively. The repeated charging and discharging cycles cause the battery to accelerate aging, which is reflected in the decay of the battery capacity. When the battery reaches the end of life (EOL) standard, i.e. the capacity decays to 70% of the rated capacity, the test stops.

[0094] This embodiment selects the mean absolute error (MAE) of SOH value as an index to evaluate and train the model:

[0095]

[0096]

[0097] where m is the total number of monitoring points, y i represents the true SOH value at time i, and represents the SOH prediction value at time i.

[0098] For the time series of lithium battery monitoring data (current, voltage and temperature), linear change is performed using deviation standardization to bring the numerical results within the [0, 1] interval, in order to improve the calculation speed of the model. The deviation standardization calculation formula is:

[0099]

[0100] where x t is the monitoring data (current, voltage or temperature) at time t, x max is the maximum value of the monitoring data, and x min is the minimum value of the monitoring data.

[0101] After the data preprocessing, the time sliding window with a length of 10 and a step of 1 is adopted in the embodiment, the training sample with a fixed length of 10 is sequentially intercepted as the input of the GRU-CNN evaluation model, and the state quantity and hidden state at the current time are predicted.

[0102] In order to verify the accuracy of the GRU-CNN model in predicting the SOH of the lithium ion battery, since the LSTM and the BP network are commonly used for predicting the health state of the lithium battery, the ConvLSTM network and the BP network are selected as the comparison in the embodiment.

[0103] The embodiment takes the B0005 battery pack as an example to analyze the influence of the number of training samples on the prediction performance of the model: the prediction performance of the model is evaluated when the number of training samples in the training set is 80, 100 and 120 respectively, and the corresponding MAE index is shown in Figure 4

[0104] When the training sample is 100, the average prediction error MAE of the lithium battery health state evaluation model based on the GRU-CNN of the application is the smallest, and when the number of training samples increases from 100 to 120, the performance of the GRU-CNN model decreases slightly with the increase of the number of training samples. Therefore, the number of training samples is selected as 100 in the embodiment.

[0105] Since the abnormal value in the prediction of the B0005 battery pack usually makes the model unable to be well fitted, the embodiment takes the B0005 as an example, sets the step to 100, and evaluates the MAE and RMSE of the three models, and the results are shown in the following table. It can be seen that the MAE and RMSE of the GRU-CNN model of the application are obviously smaller than those of the other two models, which also indicates that the prediction accuracy is higher.

[0106] Comparison of prediction results of B0005 battery pack by different methods under the condition that the step is 100:

[0107] Method MAE RMSE BP 0.0723％ 2.69％ ConvLSTM 0.0429％ 2.07％ GRU-CNN 0.0135％ 1.16％

[0108] The embodiment takes the B0005 and B0006 battery packs as examples to analyze the prediction accuracy of the remaining RUL of the lithium battery with the increase of the charging cycle number, and the results are shown in Figure 5 and Figure 6 It can be seen from Figure 5 and Figure 6 that with the increase of the cycle number of the lithium battery, the prediction curve is more consistent with the true RUL decline trend, and presents a stronger linear relationship. After the cycle number is 130, the remaining life RUL (i.e. the remaining charging number) of the battery is basically reduced to within 2 times.

[0109] ​The following table shows the actual RUL and predicted RUL values of the B0005 battery pack at 81, 101, 131 and 161 charge cycles, respectively, and calculates the average error. It can be seen that as the number of charge cycles increases, the RUL prediction error gradually decreases, and the average prediction error is less than 2%.

[0110] GRU-CNN predicts the remaining useful life of the B0005 battery pack:

[0111]

[0112] The present application can monitor the voltage and surface temperature of the lithium battery in real time, find the safety hazards of the lithium battery in time, and propose a lithium battery health state online evaluation model based on GRU-CNN. The working voltage, current and environmental temperature of the lithium battery are used as input features of the model to predict the degradation condition of the lithium battery in advance. The GRU is introduced into the time series analysis, the time series of different health factors are comprehensively considered, and the CNN network is used to process the spliced and fused GRU feature map, so that the network can learn the correlation between the health factors. The lithium battery health state evaluation model proposed in the present application has smaller prediction error for SOH lithium battery, and improves the accuracy of lithium battery health state evaluation and prediction.

[0113] The above-described embodiments are only a preferred scheme of the present application, and do not limit the present application in any form. There are other variants and modifications without exceeding the technical scheme recited in the claims.

Claims

1. A lithium battery safety operation and maintenance management system, characterized in that, Considering time series data of different health factors, a CNN network is used to process the concatenated and fused GRU feature maps to learn the correlations between the various health factors, including charging current, charging voltage, and battery temperature. A lithium battery series-parallel network, comprising several battery packs, under the control of a lithium battery intelligent operation and maintenance management unit, automatically disconnects when a battery pack has a safety hazard and replaces it with a backup battery pack. The lithium battery controller collects the operating parameters of the lithium battery series and parallel network and uploads the collected operating parameters to the lithium battery intelligent operation and maintenance management unit. The intelligent operation and maintenance management unit for lithium batteries completes online operation monitoring and health status assessment of the lithium battery series-parallel network based on the lithium battery operating parameters collected by the lithium battery controller. The health factors are input into the GRU-CNN lithium battery health status assessment model. The model uses the GRU parallel processing of the input time series and extracts time series features to predict the health factors at the next moment. The CNN is used to map the health factor features extracted by the GRU to a higher-dimensional feature space. The battery capacity predicted by the model is converted into the lithium battery health status SOH value as the true value indicator of the prediction.

2. The lithium battery safety operation and maintenance management system according to claim 1, characterized in that, The lithium battery series-parallel network includes M rows and N columns of lithium battery packs, each lithium battery pack consisting of multiple individual cells connected in series; the lithium battery packs in each column are connected in parallel, with the last column of lithium battery packs serving as a spare column; the lithium battery packs in each row are connected in series; the last row of each spare battery pack is the spare battery pack for that column; each battery pack and each column of battery packs is equipped with a switching switch.

3. A lithium battery safety operation and maintenance management system according to claim 1 or 2, characterized in that, The lithium battery controller includes: The battery pack monitoring module collects the charging voltage and current of the lithium batteries in the series-parallel network and transmits the collected voltage and current data to the control module. The temperature acquisition module collects the temperature of the lithium batteries in the lithium battery series-parallel network and transmits the collected temperature data to the control module. The control module activates the heat dissipation module of the lithium battery controller to cool down the battery when the temperature exceeds a threshold.

4. A lithium battery safety operation and maintenance management system according to claim 1 or 2, characterized in that, The intelligent operation and maintenance management unit for lithium batteries includes: The online monitoring module determines whether there are any safety hazards in individual lithium batteries based on the operating parameters of the lithium batteries, and disconnects lithium battery packs with safety hazards and switches to backup battery packs. The lithium battery health status assessment module estimates the capacity of the lithium battery based on its operating parameters, and predicts the state of health (SOH) and remaining service life (RUL) of the lithium battery. It also disconnects severely degraded lithium battery packs and switches to backup battery packs.

5. A method for assessing the health status of a lithium battery, applied to a lithium battery safety operation and maintenance management system as described in any one of claims 1-4, characterized in that, Includes the following steps: S1: Construct a lithium battery health status assessment model based on GRU-CNN; S2: The health factors of lithium batteries are used as the input time series of the evaluation model. The Gated Recurrent Unit (GRU) module of the evaluation model is used to process the input time series in parallel and extract the time series features to predict the health factors at the next time step. CNN is used to map the health factor features extracted by GRU to a higher-dimensional feature space. S3: Utilize the lithium battery health status assessment model to predict the state of health (SOH) and remaining service life (RUL) of the lithium battery, thereby achieving lithium battery health status assessment.

6. The method for assessing the health status of a lithium battery according to claim 5, characterized in that, In step S3, predicting the state of health (SOH) of the lithium battery using the lithium battery health status assessment model further includes: A1: Obtain lithium battery monitoring data, and perform z-score standardization preprocessing on the raw lithium battery monitoring data to unify the dimensions and obtain the lithium battery health factor; A2: The time-series feature information of lithium battery health factors is extracted using the gated loop module (GRU), and new indirect health factors are formed. A3: The extracted indirect health factors are concatenated into features and used as input to the CNN network; A4: The predicted battery capacity is output through a fully connected network. The remaining capacity of the lithium battery predicted by the evaluation model is calculated by using the ratio of the predicted battery capacity to the initial capacity, and the predicted value of the state of health (SOH) of the lithium battery is obtained.

7. The method for assessing the health status of a lithium battery according to claim 6, characterized in that, In step A1, for the time series of lithium battery monitoring data, deviation standardization is applied to perform linear transformation, ensuring that the numerical results fall within the interval [0, 1]. The deviation standardization calculation formula is: ， Where, x t Let x be the monitoring data at time t. max x represents the maximum value of this monitoring data. min This is the minimum value of the monitoring data.

8. A method for assessing the health status of a lithium battery according to claim 5, 6, or 7, characterized in that, When extracting time series feature information, a fixed-length sliding window is used to extract fixed-length vectors sequentially; each gated recurrent module (GRU) contains two GRU units, and each GRU unit increases the dimensionality of the received time series vector.

9. A method for assessing the health status of a lithium battery according to claim 5, 6, or 7, characterized in that, The method of using a lithium battery health status assessment model to predict the remaining lifespan (RUL) of a lithium battery further includes: using the predicted remaining capacity of the lithium battery, calculating the number of charge-discharge cycles required for the lithium battery to decay from its current remaining capacity to the failure threshold after a certain number of charge-discharge cycles, and predicting the remaining lifespan (RUL) of the lithium battery.

Images