Method and system for calibrating SOC (State of Charge) of user side energy storage battery

Abstract

The invention discloses a user side energy storage battery SOC calibration method and system. The method comprises the steps of obtaining real-time operation data of a user side energy storage battery in a fixed time period for preprocessing; extracting the processed real-time operation data according to a preset battery judgment rule; if the extracted battery characteristic fragment data is dynamic process characteristic fragment data, obtaining an SOC calibration value of the energy storage battery according to the dynamic process characteristic fragment data and a preset deep neural network model; and if the battery characteristic fragment data is stable process characteristic fragment data, calculating the SOC calibration value of the energy storage battery by adopting the corrected OCV-SOC relation curve and a preset deep neural network model according to the stable process characteristic fragment data. The method can distinguish the standing segment from the non-standing segment, employs different modes to complete the calibration, adapts to the complex dynamic characteristics of the discharge process, improves the SOC initial value calibration precision, and meets the high-precision evaluation demands.

Description

Technical Field

[0001] This invention relates to the field of energy storage battery technology, and in particular to a method and system for calibrating the state of charge (SOC) of a user-side energy storage battery. Background Technology

[0002] Accurate estimation of the State of Charge (SOC) of energy storage batteries is a core factor determining the adjustability of energy storage systems. In user-side energy storage scenarios, the capacity of energy storage power stations expands with demand, and the number of batteries increases accordingly. This makes it more difficult to assess the consistency and accuracy of the SOC of each individual battery. The initial SOC calibration is the basis for subsequent multi-stage SOC estimation, and its accuracy directly affects the reliability of the entire assessment process.

[0003] Since there is a relatively stable mapping relationship between OCV (Open Circuit Voltage) and SOC, the current industry commonly uses the OCV-SOC curve as the initial SOC calibration method. User-side energy storage batteries will have a certain period of rest during use, which is beneficial for obtaining accurate SOC values. In addition, user-side energy storage has obvious operating conditions, that is, when the grid electricity consumption is low at night, the battery will receive charging current and enter a constant power charging state, and the charging process is usually relatively stable.

[0004] However, in practical applications of energy storage on the user side, due to factors such as electricity demand and dispatch strategies, the battery resting time is often insufficient. At this time, the measured terminal voltage is not the true OCV. Calculation through OCV-SOC curve will lead to a significant increase in SOC calibration error. Moreover, the discharge process of energy storage batteries is complex. Traditional SOC calibration methods are more suitable for stable resting or simple charge and discharge conditions, and are difficult to adapt to the complex dynamic characteristics of the discharge process. As a result, under non-resting and unstable charging conditions, the accuracy of the initial SOC calibration value drops significantly, which cannot meet the requirements of high-precision evaluation. Summary of the Invention

[0005] This invention provides a user-side energy storage battery SOC calibration method and system. When a static segment is detected, high-precision calibration is achieved by relying on the modified OCV-SOC curve and the preset deep network model. When there is no static segment, calibration is completed using stable charging data through the preset deep network model. This method adapts to the complex dynamic characteristics of the discharge process and improves the accuracy of the initial SOC value calibration, thus meeting the requirements for high-precision evaluation.

[0006] To achieve the above objectives, embodiments of the present invention provide a user-side energy storage battery SOC calibration method, including: Acquire real-time operating data of user-side energy storage batteries within a fixed time period, and preprocess the real-time operating data; Based on the preset battery determination rules, the battery feature segments are extracted from the processed real-time operating data to obtain the battery feature segment data of the energy storage battery. If the battery feature segment data is dynamic process feature segment data, then the SOC calibration value of the energy storage battery is obtained based on the dynamic process feature segment data and the preset deep neural network model. If the battery feature segment data is stable process feature segment data, then based on the stable process feature segment data, the SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve and a preset deep neural network model.

[0007] As an improvement to the above solution, if the preset battery determination rules include battery operation event determination rules and battery state determination rules, The step of extracting battery feature segments from the processed real-time operating data according to preset battery determination rules to obtain battery feature segment data of the energy storage battery includes: If the processed real-time operating data is determined to be after the charging end event has occurred according to the battery operation event determination rules, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rules to obtain the stable process feature segment data of the energy storage battery. If the processed real-time operating data is determined to be in a low-rate charge / discharge event process according to the battery operation event determination rule, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rule to obtain the dynamic process feature segment data of the energy storage battery.

[0008] As an improvement to the above scheme, if the battery feature segment data is stable process feature segment data, then based on the stable process feature segment data, the SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve and a preset deep neural network model, including: If the battery feature segment data is stable process feature segment data, then the first SOC calibration value of the energy storage battery is obtained based on the stable process feature segment data and the preset deep neural network model. Based on the stable process characteristic segment data, the second SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve; The SOC calibration value of the energy storage battery is determined based on the first SOC calibration value, the second SOC calibration value, and the preset SOC threshold.

[0009] As an improvement to the above scheme, after obtaining the SOC calibration value of the energy storage battery, the method further includes: If the SOC calibration value meets the preset physical constraints, then the SOC calibration value is taken as the final SOC calibration value. If the SOC calibration value does not meet the preset physical constraints, the SOC calibration value is corrected, and the corrected SOC calibration value is used as the final SOC calibration value.

[0010] As an improvement to the above solution, the method for obtaining the preset deep neural network model includes: Obtain historical operating data of user-side energy storage batteries and preprocess the historical operating data; Based on the preset battery determination rules, the processed historical operating data is processed to extract battery feature segments, thereby obtaining dynamic process feature segment sample data of the energy storage battery. Correlation analysis was performed on the sample data of the dynamic process feature segments to obtain the training dataset; The deep neural network model is iteratively trained using the training dataset through sliding window sampling until the model accuracy reaches a preset accuracy threshold, thus obtaining the preset deep neural network model.

[0011] As an improvement to the above scheme, the structure of the preset deep neural network model includes: The system consists of a first convolutional layer, a max pooling layer, a second convolutional layer, a GRU layer, and a fully connected layer.

[0012] As an improvement to the above scheme, the method for obtaining the modified OCV-SOC relationship curve includes: Obtain historical operating data of user-side energy storage batteries and preprocess the historical operating data; Based on the preset battery determination rules, the battery feature segments are extracted from the processed historical operating data to obtain the stable process feature segment sample data of the energy storage battery. The OCV-SOC relationship curve was obtained by segmented static placement method, and the battery temperature of the individual cells in the energy storage battery was collected. Based on the battery temperature and the sample data of the characteristic segments of the stabilization process, the current-temperature relationship curve of the energy storage battery is reproduced; Based on the current-temperature relationship curve, the relationship between OCV and SOC at different temperatures is linearly fitted, and the temperature compensation coefficient is calculated. The temperature deviation of the OCV-SOC relationship curve is corrected according to the temperature compensation coefficient to obtain the corrected OCV-SOC relationship curve.

[0013] As an improvement to the above scheme, the battery state determination rule includes current stability criteria and voltage stability criteria; The battery operation event determination rules include charging end event criteria and low-rate charge / discharge event criteria.

[0014] To achieve the above objectives, embodiments of the present invention provide a user-side energy storage battery SOC calibration system, comprising: The real-time data acquisition module is used to acquire real-time operating data of user-side energy storage batteries within a fixed time period and to preprocess the real-time operating data. The feature segment extraction module is used to extract battery feature segments from the processed real-time operating data according to preset battery determination rules, so as to obtain the battery feature segment data of the energy storage battery. The initial calibration value acquisition module is used to obtain the SOC calibration value of the energy storage battery based on the dynamic process feature segment data and the preset deep neural network model if the battery feature segment data is dynamic process feature segment data. The initial calibration value calculation module is used to calculate the SOC calibration value of the energy storage battery based on the stable process characteristic segment data, using the corrected OCV-SOC relationship curve and a preset deep neural network model, if the battery characteristic segment data is stable process characteristic segment data.

[0015] As an improvement to the above solution, if the preset battery determination rules include battery operation event determination rules and battery state determination rules, The feature fragment extraction module is used for: If the processed real-time operating data is determined to be after the charging end event has occurred according to the battery operation event determination rules, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rules to obtain the stable process feature segment data of the energy storage battery. If the processed real-time operating data is determined to be in a low-rate charge / discharge event process according to the battery operation event determination rule, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rule to obtain the dynamic process feature segment data of the energy storage battery.

[0016] Compared with existing technologies, the present invention discloses a user-side energy storage battery SOC calibration method and system. This method acquires real-time operating data of the user-side energy storage battery within a fixed time period and preprocesses the real-time operating data. Based on preset battery determination rules, it extracts battery feature segments from the processed real-time operating data to obtain battery feature segment data. If the battery feature segment data is dynamic process feature segment data, the SOC calibration value of the energy storage battery is obtained based on the dynamic process feature segment data and a preset deep neural network model. If the battery feature segment data is stable process feature segment data, the SOC calibration value of the energy storage battery is calculated based on the stable process feature segment data, using a modified OCV-SOC relationship curve and a preset deep neural network model. When a static segment is detected, high-precision calibration is achieved using the modified OCV-SOC curve and the preset deep network model. When no static segment is detected, calibration is completed using stable charging data through the preset deep network model. This adapts to the complex dynamic characteristics of the discharge process and improves the accuracy of the initial SOC calibration, meeting the requirements for high-precision evaluation. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a user-side energy storage battery SOC calibration method provided in an embodiment of the present invention; Figure 2 This is a power-time curve under actual operating conditions provided by an embodiment of the present invention; Figure 3 This is an OCV-SOC relationship curve diagram provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a user-side energy storage battery SOC calibration system provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that the terms "comprising" and "specific" in this invention, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0020] Please see Figure 1 , Figure 1 This is a flowchart illustrating a user-side energy storage battery SOC calibration method provided in an embodiment of the present invention. The user-side energy storage battery SOC calibration method includes: S1, acquire real-time operating data of user-side energy storage batteries within a fixed time period, and preprocess the real-time operating data; S2, according to the preset battery determination rules, the battery feature segments are extracted from the processed real-time operating data to obtain the battery feature segment data of the energy storage battery; S3, if the battery feature segment data is dynamic process feature segment data, then the SOC calibration value of the energy storage battery is obtained based on the dynamic process feature segment data and the preset deep neural network model. S4. If the battery feature segment data is stable process feature segment data, then based on the stable process feature segment data, the SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve and the preset deep neural network model.

[0021] For example, the real-time operating data (e.g., voltage, current, and power) is preprocessed using linear interpolation and a moving average filtering algorithm. For instance, missing data points are processed using linear interpolation, and outliers significantly exceeding physically reasonable ranges are removed based on the 3σ principle or set thresholds (maximum and minimum current, maximum and minimum voltage). The moving average filtering algorithm smooths the real-time operating data such as battery voltage and current, suppressing high-frequency noise. Battery feature segments are extracted from the processed real-time operating data according to preset battery determination rules to obtain battery feature segment data for the energy storage battery. If the battery feature segment data is dynamic process feature segment data, there are no static event segments. Based on the dynamic process feature segment data and a preset deep neural network model, the SOC calibration value of the energy storage battery is obtained. If the battery feature segment data is stable process feature segment data, there are static event segments. Based on the stable process feature segment data, the SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve and a preset deep neural network model. In this embodiment of the invention, when a stationary segment is detected, high-precision calibration is achieved by relying on the modified OCV-SOC curve and the preset deep network model. When there is no stationary segment, calibration is completed using stable charging data through the preset deep network model. This adapts to the complex dynamic characteristics of the discharge process and improves the accuracy of the initial SOC value calibration, thus meeting the requirements for high-precision evaluation.

[0022] Specifically, if the preset battery determination rules include battery operation event determination rules and battery status determination rules, Then step S2 includes: S21, if it is determined according to the battery operation event determination rule that the processed real-time operation data is after the charging end event has occurred, then the battery feature segment is extracted from the processed real-time operation data according to the battery state determination rule to obtain the stable process feature segment data of the energy storage battery. S22, if the processed real-time operating data is determined to be in a low-rate charge / discharge event process according to the battery operation event determination rule, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rule to obtain the dynamic process feature segment data of the energy storage battery.

[0023] For example, the charging and discharging process of a user-side energy storage system exhibits a certain temporal regularity, but the power fluctuations in each segment of the charging and discharging curve are significant. Selecting characteristic segments for operating condition analysis, such as... Figure 2 As shown, Figure 2 This invention provides an actual operating condition power-time curve. Based on the actual power-time curve, and using the judgment rules for low-rate constant current or constant voltage charging processes, dynamic process feature segment data is extracted, such as... Figure 2 The portion of SOC spanning 90%-100% within the middle frame; and using the rule for determining the resting time after charging completion, extracting feature fragment data of the stable process.

[0024] Specifically, step S4 includes: S41, if the battery feature segment data is stable process feature segment data, then the first SOC calibration value of the energy storage battery is obtained based on the stable process feature segment data and the preset deep neural network model. S42, based on the stable process characteristic segment data, the second SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve; S43, determine the SOC calibration value of the energy storage battery based on the first SOC calibration value, the second SOC calibration value, and the preset SOC threshold.

[0025] For example, when a static event segment is detected, the current voltage and temperature are read, and the theoretical OCV value (second SOC calibration value) is calculated using the corrected OCV-SOC relationship curve. This value is then compared with the SOCnet (first SOC calibration value) output by the preset deep network model. delta_SOC = |SOCnet-OCV -1 (SOC,T)|, In the formula, delta_SOC is the absolute value of the SOC difference; OCV-1(SOC,T) is the SOC calibration value of the energy storage battery calculated through the corrected OCV-SOC relationship curve, where OCV-1 represents the inverse function of the OCV-SOC relationship, that is, the state of charge (SOC) is inversely derived from the open-circuit voltage (OCV) and temperature (T); if delta_SOC is greater than the preset SOC threshold SOC_threshold, calibration is triggered to obtain the calibration value SOC. st = OCV -1 (SOC,T) and simultaneously calibrated SOC st The parameters of the preset deep network model are used as the true value to correct the back-end. If no suitable static segment is detected, after charging is completed in a specific time period each day, the power and terminal voltage segments of the corresponding time period are extracted, and the SOC calibration is performed using the preset deep neural network to obtain the calibration value SOCnet.

[0026] Furthermore, after obtaining the SOC calibration value of the energy storage battery, the method further includes: S5. If the SOC calibration value meets the preset physical constraint conditions, then the SOC calibration value is taken as the final SOC calibration value. S6. If the SOC calibration value does not meet the preset physical constraints, the SOC calibration value is corrected and the corrected SOC calibration value is used as the final SOC calibration value.

[0027] For example, the preset physical constraints are: , In the formula, The estimated value obtained at the current moment. This is the SOC estimate obtained at the previous moment. Considering that the SOC value of the energy storage battery will not change abruptly under normal operating conditions, the SOC change factor is increased. Constraints It can be set to 3%.

[0028] Specifically, the method for obtaining the preset deep neural network model includes: Obtain historical operating data of user-side energy storage batteries and preprocess the historical operating data; Based on the preset battery determination rules, the processed historical operating data is processed to extract battery feature segments, thereby obtaining dynamic process feature segment sample data of the energy storage battery. Correlation analysis was performed on the sample data of the dynamic process feature segments to obtain the training dataset; The deep neural network model is iteratively trained using the training dataset through sliding window sampling until the model accuracy reaches a preset accuracy threshold, thus obtaining the preset deep neural network model.

[0029] For example, correlation analysis is used to extract feature input data for a deep neural network model, and small-batch gradient dispersion optimization training is performed to obtain the initial SOC value of the energy storage battery under operating conditions. Current, voltage, power, and SOC data for corresponding time segments are extracted, and data correlation analysis is conducted to determine the correlation of feature segments at the same time period on different days. A correlation coefficient greater than 0.7 indicates that the correlation meets the requirements and can be used as input segments for SOC estimation in deep learning. A window of length m is used to scan the entire input and output time series to obtain samples containing both input and output. For a time series of length N, N-m+1 samples can be obtained by moving the window at a step size of 1. A window size of 1200s is selected, collecting 20 minutes of charging data as input; 12 days of feature segment data are extracted as the training set, consisting of 10-minute power and voltage segments labeled with the SOC value at the end of the 10th minute. Based on the extracted data, the SOC is calculated using the ampere-hour integration method.

[0030] Deep neural network models are trained using mini-batch gradient scattering optimization. The training dataset is divided into mini-batches, each consisting of n training samples. Mean squared error is used as the training loss, and the model parameters are updated through batch iterative calculation. The loss function is expressed as: , In the formula, Here, n represents the mean squared error loss value, and n is the size of each batch of data, which can be set to 1024. For the first The true SOC value of each sample; Let be the SOC prediction value of the deep neural network model for the i-th sample. Then, the Adam algorithm is used as a gradient descent method to iteratively update the model parameters.

[0031] Specifically, the structure of the preset deep neural network model includes: The system consists of a first convolutional layer, a max pooling layer, a second convolutional layer, a GRU layer, and a fully connected layer.

[0032] For example, the structure and parameters of a deep neural network model are designed. The first layer is a convolutional layer, consisting of a set of filters. Each filter is convolved with the layer input to extract features. After scanning the entire input sequence through a window, the output sequence is obtained. This layer contains 32 filters, with a window size of 5 and a stride of 1. A max-pooling layer is used to further compress the information extracted by the first convolutional layer. In this layer, a pooling size of 3 is chosen, meaning that the maximum value comes from a window of 3 at the output of the convolutional layer. Another convolutional layer with the same parameters as the first convolutional layer is connected using the same settings to further extract information.

[0033] The second convolutional layer uses a gated recurrent unit (GRU) layer to learn the sequential dependencies of the input data, processing the input data x in a cyclic manner. The state h is used to store historical information. At time k, the input to the GRU includes the input vector of the GRU layer at time k. and the Time-bound GRU layer hidden state Then the reset gate r and update gate z are calculated as follows: , In the formula, and These are the weight matrices corresponding to the reset gate r and the update gate z, respectively. It is defined as The sigmoid function. Then, the prior states are calculated. for: , In the formula, Hyperbolic tangent activation function, , Let be the prior state weight matrix, with the symbol . This represents the element-wise multiplication of the vectors. Finally, we obtain the updated state. , represented as: , The reset gate r restricts The GRU layer affects the weights between the previous state and the new prior state, while the update gate affects the weights between the previous state and the new prior state. This ensures long-term memory of the input sequence. In the designed deep neural network model, the GRU layer has 32 units.

[0034] A fully connected layer consisting of 5 neurons is added after the GRU layer. The neurons in the fully connected layer independently compute the dot product between the input and the weights, and then input it into the activation function. Rectified linear units (ReLU) are applied in the fully connected layer and the convolutional layer to increase nonlinearity by simply outputting the positive part of their parameters.

[0035] Specifically, the method for obtaining the modified OCV-SOC relationship curve includes: Obtain historical operating data of user-side energy storage batteries and preprocess the historical operating data; Based on the preset battery determination rules, the battery feature segments are extracted from the processed historical operating data to obtain the stable process feature segment sample data of the energy storage battery. The OCV-SOC relationship curve was obtained by segmented static placement method, and the battery temperature of the individual cells in the energy storage battery was collected. Based on the battery temperature and the sample data of the characteristic segments of the stabilization process, the current-temperature relationship curve of the energy storage battery is reproduced; Based on the current-temperature relationship curve, the relationship between OCV and SOC at different temperatures is linearly fitted, and the temperature compensation coefficient is calculated. The temperature deviation of the OCV-SOC relationship curve is corrected according to the temperature compensation coefficient to obtain the corrected OCV-SOC relationship curve.

[0036] For example, using a segmented resting method and a dynamic operating condition reproduction method, charge-discharge tests were conducted on individual cells from the same batch: First, the segmented resting method was used, with 10% SOC intervals. Constant current charge / discharge was performed to the target SOC in each stage, with a rate less than 0.5C, and the resting time was no less than 2 hours. The voltage stability value was recorded, and the OCV-SOC relationship curve was obtained, as shown below. Figure 3 As shown, Figure 3 This invention provides an OCV-SOC relationship curve diagram and simultaneously collects battery temperature; based on the sample data of the characteristic segment of the stable process, the current-temperature curve in actual operation is reproduced, and OCV is measured after resting under the same operating conditions to establish the operating condition-OCV compensation relationship: OCV(SOC,T)=OCV 25℃ (SOC)+β(T-25), Where OCV(SOC,T) is the corrected open-circuit voltage at a specific SOC and temperature T, and OCV25℃(SOC) is the open-circuit voltage corresponding to SOC at room temperature (25℃). β This is the temperature compensation coefficient.

[0037] Specifically, the battery state determination rules include current stability criteria and voltage stability criteria; The battery operation event determination rules include charging end event criteria and low-rate charge / discharge event criteria.

[0038] For example, battery state determination rules are defined to identify and define steady-state or quasi-steady-state characteristic segments of battery data. The current stability criterion is as follows: a current fluctuation threshold delta_I_threshold and a time window T_width are set. If the absolute value of the current fluctuation range is less than delta_I_threshold within the continuous time window T_width, the battery is determined to be in a static or low-power operation. The voltage stability criterion is as follows: a voltage fluctuation threshold delta_V_threshold and a time window T_width are set. If the absolute value of the voltage change rate is less than delta_V_threshold within the continuous time window T_width, the battery is determined to be in a static or constant-voltage steady state. If the voltage fluctuation rate is constant, it can be determined to be a low-rate constant-current charging and discharging process, which is considered a quasi-steady state. Define battery operation event judgment rules to identify characteristic segments of data at specific times. The criteria for the end of charging event are: the charging current is detected to drop to near zero or less than I_cutoff and remain there for a set time T_event, or the battery voltage reaches the set charging cutoff voltage V_cutoff. The criteria for the low-rate charge and discharge event are: the absolute value of the charge and discharge current is detected to be less than I_threshold.

[0039] This invention discloses a method for SOC calibration of a user-side energy storage battery. The method involves acquiring real-time operating data of the user-side energy storage battery over a fixed time period and preprocessing the data. Battery feature segments are extracted from the processed data according to preset battery determination rules to obtain battery feature segment data. If the battery feature segment data is dynamic process feature segment data, the SOC calibration value of the energy storage battery is obtained based on the dynamic process feature segment data and a preset deep neural network model. If the battery feature segment data is stable process feature segment data, the SOC calibration value of the energy storage battery is calculated based on the stable process feature segment data, using a modified OCV-SOC relationship curve and a preset deep neural network model. When a static segment is detected, high-precision calibration is achieved using the modified OCV-SOC curve and the preset deep network model. When no static segment is detected, calibration is completed using stable charging data through the preset deep network model. This method adapts to the complex dynamic characteristics of the discharge process and improves the accuracy of the initial SOC calibration, meeting the requirements for high-precision evaluation.

[0040] See Figure 4 , Figure 4 This is a schematic diagram of a user-side energy storage battery SOC calibration system 10 provided in an embodiment of the present invention. The user-side energy storage battery SOC calibration system 10 includes: The real-time data acquisition module 11 is used to acquire real-time operating data of the user-side energy storage battery within a fixed time period and to preprocess the real-time operating data. The feature segment extraction module 12 is used to extract battery feature segments from the processed real-time operating data according to the preset battery determination rules, so as to obtain the battery feature segment data of the energy storage battery. The initial calibration value acquisition module 13 is used to obtain the SOC calibration value of the energy storage battery based on the dynamic process feature segment data and the preset deep neural network model if the battery feature segment data is dynamic process feature segment data. The initial calibration value calculation module 14 is used to calculate the SOC calibration value of the energy storage battery based on the stable process feature segment data, using the corrected OCV-SOC relationship curve and a preset deep neural network model, if the battery feature segment data is stable process feature segment data.

[0041] Furthermore, the user-side energy storage battery SOC calibration system 10 also includes: The final calibration value determination module is used to determine the final SOC calibration value if the SOC calibration value meets the preset physical constraints. The initial calibration value correction module is used to correct the SOC calibration value if the SOC calibration value does not meet the preset physical constraints, and to use the corrected SOC calibration value as the final SOC calibration value.

[0042] The user-side energy storage battery SOC calibration system 10 provided in this embodiment of the invention can realize all the processes of the user-side energy storage battery SOC calibration method of the above embodiments. The functions and technical effects of each module in the system are the same as those of the user-side energy storage battery SOC calibration method of the above embodiments, and will not be repeated here.

[0043] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for calibrating the State of Charge (SOC) of a user-side energy storage battery, characterized in that, include: Acquire real-time operating data of user-side energy storage batteries within a fixed time period, and preprocess the real-time operating data; Based on the preset battery determination rules, the battery feature segments are extracted from the processed real-time operating data to obtain the battery feature segment data of the energy storage battery. If the battery feature segment data is dynamic process feature segment data, then the SOC calibration value of the energy storage battery is obtained based on the dynamic process feature segment data and the preset deep neural network model. If the battery feature segment data is stable process feature segment data, then based on the stable process feature segment data, the SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve and a preset deep neural network model.

2. The user-side energy storage battery SOC calibration method as described in claim 1, characterized in that, If the preset battery determination rules include battery operation event determination rules and battery status determination rules, The step of extracting battery feature segments from the processed real-time operating data according to preset battery determination rules to obtain battery feature segment data of the energy storage battery includes: If the processed real-time operating data is determined to be after the charging end event has occurred according to the battery operation event determination rules, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rules to obtain the stable process feature segment data of the energy storage battery. If the processed real-time operating data is determined to be in a low-rate charge / discharge event process according to the battery operation event determination rule, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rule to obtain the dynamic process feature segment data of the energy storage battery.

3. The user-side energy storage battery SOC calibration method as described in claim 1, characterized in that, If the battery feature segment data is stable process feature segment data, then based on the stable process feature segment data, using the corrected OCV-SOC relationship curve and a preset deep neural network model, the SOC calibration value of the energy storage battery is calculated, including: If the battery feature segment data is stable process feature segment data, then the first SOC calibration value of the energy storage battery is obtained based on the stable process feature segment data and the preset deep neural network model. Based on the stable process characteristic segment data, the second SOC calibration value of the energy storage battery is calculated using the corrected OCV-SOC relationship curve; The SOC calibration value of the energy storage battery is determined based on the first SOC calibration value, the second SOC calibration value, and the preset SOC threshold.

4. The user-side energy storage battery SOC calibration method as described in claim 1, characterized in that, After obtaining the SOC calibration value of the energy storage battery, the method further includes: If the SOC calibration value meets the preset physical constraints, then the SOC calibration value is taken as the final SOC calibration value. If the SOC calibration value does not meet the preset physical constraints, the SOC calibration value is corrected, and the corrected SOC calibration value is used as the final SOC calibration value.

5. The user-side energy storage battery SOC calibration method as described in claim 1, characterized in that, The method for obtaining the preset deep neural network model includes: Obtain historical operating data of user-side energy storage batteries and preprocess the historical operating data; Based on the preset battery determination rules, the processed historical operating data is processed to extract battery feature segments, thereby obtaining dynamic process feature segment sample data of the energy storage battery. Correlation analysis was performed on the sample data of the dynamic process feature segments to obtain the training dataset; The deep neural network model is iteratively trained using the training dataset through sliding window sampling until the model accuracy reaches a preset accuracy threshold, thus obtaining the preset deep neural network model.

6. The user-side energy storage battery SOC calibration method as described in claim 1, characterized in that, The structure of the preset deep neural network model includes: The system consists of a first convolutional layer, a max pooling layer, a second convolutional layer, a GRU layer, and a fully connected layer.

7. The user-side energy storage battery SOC calibration method as described in claim 1, characterized in that, The method for obtaining the modified OCV-SOC relationship curve includes: Obtain historical operating data of user-side energy storage batteries and preprocess the historical operating data; Based on the preset battery determination rules, the battery feature segments are extracted from the processed historical operating data to obtain the stable process feature segment sample data of the energy storage battery. The OCV-SOC relationship curve was obtained by segmented static placement method, and the battery temperature of the individual cells in the energy storage battery was collected. Based on the battery temperature and the sample data of the characteristic segments of the stabilization process, the current-temperature relationship curve of the energy storage battery is reproduced; Based on the current-temperature relationship curve, the relationship between OCV and SOC at different temperatures is linearly fitted, and the temperature compensation coefficient is calculated. The temperature deviation of the OCV-SOC relationship curve is corrected according to the temperature compensation coefficient to obtain the corrected OCV-SOC relationship curve.

8. The user-side energy storage battery SOC calibration method as described in claim 2, characterized in that, The battery state determination rules include current stability criteria and voltage stability criteria; The battery operation event determination rules include charging end event criteria and low-rate charge / discharge event criteria.

9. A user-side energy storage battery SOC calibration system, characterized in that, include: The real-time data acquisition module is used to acquire real-time operating data of user-side energy storage batteries within a fixed time period and to preprocess the real-time operating data. The feature segment extraction module is used to extract battery feature segments from the processed real-time operating data according to preset battery determination rules, so as to obtain the battery feature segment data of the energy storage battery. The initial calibration value acquisition module is used to obtain the SOC calibration value of the energy storage battery based on the dynamic process feature segment data and the preset deep neural network model if the battery feature segment data is dynamic process feature segment data. The initial calibration value calculation module is used to calculate the SOC calibration value of the energy storage battery based on the stable process characteristic segment data, using the corrected OCV-SOC relationship curve and a preset deep neural network model, if the battery characteristic segment data is stable process characteristic segment data.

10. The user-side energy storage battery SOC calibration system as described in claim 9, characterized in that, If the preset battery determination rules include battery operation event determination rules and battery status determination rules, The feature fragment extraction module is used for: If the processed real-time operating data is determined to be after the charging end event has occurred according to the battery operation event determination rules, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rules to obtain the stable process feature segment data of the energy storage battery. If the processed real-time operating data is determined to be in a low-rate charge / discharge event process according to the battery operation event determination rule, then the battery feature segments are extracted from the processed real-time operating data according to the battery state determination rule to obtain the dynamic process feature segment data of the energy storage battery.

Images