Balanced charging and discharging control method for battery pack of energy storage cabinet based on AI prediction

Abstract

The invention discloses an energy storage cabinet battery pack balanced charging and discharging control method based on AI prediction, and relates to the technical field of balanced charging and discharging control. The AI prediction-based balanced charging and discharging control method for the battery packs of the energy storage cabinet comprises the following steps: acquiring operation state data, performance state data and thermal imaging data of a plurality of battery packs in each time period, and analyzing battery charging and discharging degradation characteristic values and energy attenuation characteristic values in corresponding time periods; on the basis of a pre-trained thermal evolution charge-discharge correlation model, charge-discharge thermal regulation characteristic values of the corresponding time periods are analyzed, charge-discharge regulation demand characteristic values of the corresponding time periods are analyzed in combination with the characteristic values, and predicted charge-discharge regulation demand characteristic values of the corresponding battery packs in the next time period are analyzed; balanced charge-discharge control is performed on the next time period of the corresponding battery pack of the set energy storage cabinet by predicting the charge-discharge adjustment demand characteristic value, so that the balanced charge-discharge control precision is improved, and the energy utilization rate is further remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of balanced charge and discharge control technology, specifically to an AI-predictive balanced charge and discharge control method for battery packs in energy storage cabinets. Background Technology

[0002] With the rapid development of new energy power generation and large-scale energy storage, energy storage cabinets, as important units for energy storage and regulation, are typically composed of several series-parallel battery packs. Their overall performance directly depends on the operating status and energy output capacity of each battery pack. However, in actual operation, due to differences in manufacturing processes, material aging, usage environment, and cycle count, battery packs often exhibit inconsistencies in parameters such as voltage, capacity, and internal resistance. When these differences accumulate during charging, some battery packs are prone to overcharging, leading to local temperature rise, increased polarization, and even the risk of thermal runaway. During discharging, some battery packs may be over-discharged, causing a sharp increase in internal resistance and energy output decay, thereby affecting the lifespan and safety of the entire energy storage system.

[0003] Existing technologies, such as the battery pack charging and discharging control method, battery pack, energy storage system, and storage medium disclosed in patent application CN113162028B, involve the battery pack discharging or receiving charge through a power conversion device. The charging and discharging control method includes: in a discharging state, when the battery voltage of the battery pack is detected to meet the undervoltage protection condition, accumulating undervoltage protection counts within a first preset time period; if the undervoltage protection counts reach a preset number and the battery voltage is less than or equal to a first voltage threshold, setting a prohibition-discharge flag; while the prohibition-discharge flag is set, sending a first control signal to the power conversion device; the first control signal is used to prohibit the power conversion device from drawing power from the battery pack. This avoids the power supply device drawing power from the battery pack to supply power to the load when the battery pack is in an undervoltage state, continuously consuming battery pack energy, causing the battery pack voltage to drop too low, affecting battery pack life, and ensuring the safe use of the equipment.

[0004] Based on the above findings, the limitations of existing technologies include at least the following problems: Existing technologies lack the ability to proactively analyze and precisely control the operating trends of battery packs. As the voltage, current, and heat distribution of the battery pack are coupled during charging and discharging, and these factors continuously evolve over time and under changing operating conditions, existing technologies struggle to accurately reflect the evolution of battery states. This results in a lag in the response to battery operating trends, especially before the risk of overcharging or over-discharging occurs. It is difficult to detect early characteristic signals in a timely manner, making it difficult to dynamically adjust for differences between batteries. This can easily lead to a decrease in energy utilization, thereby shortening the lifespan of the entire battery pack, increasing operating and maintenance costs, and creating potential safety hazards. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an AI-based predictive method for balanced charging and discharging control of battery packs in energy storage cabinets, which solves the problem that existing technologies are unable to predict and control in advance, leading to a decrease in energy utilization.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an AI-predictive method for balanced charge and discharge control of battery packs in an energy storage cabinet, comprising the following steps: continuously acquiring operational status data, performance status data, and thermal imaging data of several battery packs in a designated energy storage cabinet for each time period; analyzing the battery charge and discharge degradation characteristic values ​​and energy decay characteristic values ​​for each time period based on the operational status data and energy status data of each battery pack in the designated energy storage cabinet for each time period; analyzing the charge and discharge thermal regulation characteristic values ​​for each time period based on a pre-trained thermal evolution charge and discharge correlation model, combined with the thermal imaging time-series data of each battery pack in the designated energy storage cabinet for each time period, and analyzing the charge and discharge regulation demand characteristic values ​​for each time period based on the battery charge and discharge degradation characteristic values ​​and energy decay characteristic values; analyzing the predicted charge and discharge regulation demand characteristic values ​​for the next time period of each battery pack in the designated energy storage cabinet based on the predicted charge and discharge regulation demand characteristic values; and performing balanced charge and discharge control on the corresponding battery pack in the next time period based on the predicted charge and discharge regulation demand characteristic values.

[0007] Furthermore, the operational status data includes voltage fluctuation amplitude, SOC value, internal resistance change rate, interface electrical micro-disturbance value, electromagnetic leakage field strength value, and terminal current pulsation response value. The specific steps for analyzing the battery charging and discharging degradation characteristics of each battery pack in each time period of the set energy storage cabinet are as follows: Based on the operational status data of each battery pack in each time period of the set energy storage cabinet, analyze the corresponding operational anomaly characteristic set, including battery operational imbalance characteristic value and electrical transmission anomaly characteristic value; Based on the operational anomaly characteristic set of each battery pack in each time period of the set energy storage cabinet, analyze the corresponding battery charging and discharging degradation characteristic value.

[0008] Furthermore, the specific steps for analyzing the abnormal operation characteristic set of each battery pack in the energy storage cabinet for each time period are as follows: Based on the voltage fluctuation amplitude, SOC value, and internal resistance change rate value of each battery pack in the energy storage cabinet for each time period, analyze the battery operation imbalance characteristic value of the corresponding time period; Based on the interface electrical micro-disturbance value, electromagnetic leakage field strength value, and terminal current pulsation response value of each battery pack in the energy storage cabinet for each time period, analyze the electrical transmission abnormal characteristic value of the corresponding time period.

[0009] Furthermore, the specific steps for analyzing the energy decay characteristic value of each battery pack in the set energy storage cabinet for each time period are as follows: read the energy state data of each battery pack in the set energy storage cabinet for each time period and perform standardization processing; perform comprehensive analysis on the standardized energy state data of each battery pack in the set energy storage cabinet for each time period to obtain the corresponding energy decay characteristic value.

[0010] Furthermore, the thermal imaging data includes the pixel value and two-dimensional coordinates of each pixel in several frames of thermal imaging, and the thermal evolution charge-discharge correlation model includes an input layer, a heat flow analysis layer, a charge-discharge thermal evolution layer, and an output layer.

[0011] Furthermore, the specific steps for analyzing the charge-discharge thermal regulation characteristic values ​​of each battery pack in the energy storage cabinet for each time period are as follows: Input the thermal imaging data of each battery pack in the energy storage cabinet for each time period into the pre-trained thermal evolution charge-discharge correlation model, and analyze the charge-discharge thermal coupling characteristic set of the corresponding time period, including hot spot link offset characteristic value, hierarchical thermal disturbance intensity characteristic value, and thermal spectrum misalignment characteristic value; Based on the charge-discharge thermal coupling characteristic set of each battery pack in the energy storage cabinet for each time period, analyze the charge-discharge thermal regulation characteristic value of the corresponding time period.

[0012] Further, the specific steps for analyzing the charge-discharge thermal coupling feature set of each battery pack in each time period of the set energy storage cabinet are as follows: In the input layer of the thermal evolution charge-discharge correlation model, the thermal imaging data of each battery pack in each time period of the set energy storage cabinet is received and preprocessed; in the thermal flow analysis layer of the thermal evolution charge-discharge correlation model, based on the preprocessed thermal imaging data of each battery pack in each time period of the set energy storage cabinet, the thermal anomaly feature vector in the corresponding frame thermal imaging is extracted; in the charge-discharge thermal evolution layer of the thermal evolution charge-discharge correlation model, based on the thermal anomaly feature vector in the thermal imaging of each frame thermal imaging of each battery pack in each time period of the set energy storage cabinet, the thermal evolution time sequence feature vector in the corresponding time period is extracted; in the output layer of the thermal evolution charge-discharge correlation model, based on the thermal evolution time sequence feature vector in the thermal evolution of each battery pack in each time period of the set energy storage cabinet, the charge-discharge thermal coupling feature set in the corresponding time period is output.

[0013] Furthermore, the specific steps for analyzing the predicted charge and discharge regulation demand characteristic values ​​for each battery pack in the energy storage cabinet for the next time period are as follows: Based on the predicted charge and discharge regulation demand characteristic values ​​for each battery pack in the energy storage cabinet for each time period, analyze the predicted demand baseline value, demand fluctuation factor, and regulation demand inertia factor of the corresponding battery pack; based on the predicted demand baseline value, demand fluctuation factor, and regulation demand inertia factor of each battery pack in the energy storage cabinet, analyze the predicted charge and discharge regulation demand characteristic values ​​for the corresponding battery pack for the next time period.

[0014] Furthermore, the specific steps for analyzing the regulation demand inertia factor of each battery pack in the energy storage cabinet are as follows: Based on the charging and discharging regulation demand characteristic value of each battery pack in the energy storage cabinet for each time period, analyze the regulation demand change characteristic value of several groups of adjacent time periods for the corresponding battery pack; Based on the regulation demand change characteristic value of each group of adjacent time periods for each battery pack in the energy storage cabinet, analyze the regulation demand inertia factor of the corresponding battery pack.

[0015] Furthermore, the specific steps for equalizing charge and discharge control of the corresponding battery packs in the next time period based on the predicted charge and discharge regulation demand characteristic values ​​of the set energy storage cabinet are as follows: compare the predicted charge and discharge regulation demand characteristic values ​​of each battery pack in the next time period of the set energy storage cabinet with the preset charge and discharge regulation demand threshold range; and take preset equalizing charge and discharge control measures for the corresponding battery packs in the next time period of the set energy storage cabinet based on the comparison results.

[0016] The present invention has the following beneficial effects: (1) The AI-based prediction-based balanced charging and discharging control method for battery packs in energy storage cabinets continuously acquires multi-source data of the battery pack during operation, thereby deeply analyzing the characteristics of energy decay and charging and discharging degradation generated during the charging and discharging process of the battery pack, and generating characteristic values ​​of charging and discharging regulation demand for each period, thereby accurately describing the dynamic trend of the charging and discharging process of the battery pack, and predicting the demand changes in the next period, thus realizing forward-looking operation trend perception, capturing potential characteristics before the risk of overcharging or over-discharging occurs, and implementing control strategies in advance based on the prediction results, thereby significantly improving energy utilization and operational stability, extending the effective working cycle of the battery pack, and reducing the operational risk of the battery pack.

[0017] (2) The AI-based prediction-based equalization charging and discharging control method for battery packs in energy storage cabinets introduces a pre-trained thermal evolution charging and discharging correlation model to perform in-depth analysis of the thermal imaging data generated by the battery pack during charging and discharging, and extracts the corresponding features. This enables time-series tracking of thermal anomalies throughout the charging and discharging process. The model can simultaneously capture the connectivity of local hotspots and the penetration pattern of heat from the outside to the inside, as well as the shift of heat distribution in scale and direction. Through time-series modeling, it reveals the dynamic changes of these features as they evolve with the charging and discharging process, thereby achieving static analysis of single-frame thermal imaging. It can also identify the evolution pattern of features during the charging and discharging process of the battery pack, thus significantly improving the depth of battery pack operating status analysis and improving the accuracy of equalization charging and discharging control.

[0018] (3) The AI-based energy storage cabinet battery pack equalization charge and discharge control method introduces the predicted demand benchmark value, demand fluctuation factor and adjustment demand inertia factor to predict the characteristic value of charge and discharge adjustment demand in the future period. It can capture the inertial trend and potential sudden change characteristics of battery operation status, thereby avoiding the deviation of monotonous increase or decrease in the prediction result. In this way, the risk of overcharging or over-discharging can be identified in advance in the next period, and equalization regulation can be achieved through control measures, thereby improving the timeliness of regulation and enhancing the safety of battery pack operation.

[0019] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0020] Figure 1 This is a flowchart of the AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets, as described in this invention.

[0021] Figure 2 This is a flowchart illustrating the specific steps involved in analyzing and setting the battery charging and discharging degradation characteristic values ​​of each battery pack in the energy storage cabinet for each time period in the AI-predictive battery pack equalization charging and discharging control method of the present invention.

[0022] Figure 3 This is a schematic diagram of the timing set data of the charging and discharging thermal coupling characteristics of a certain battery pack in the energy storage cabinet in the AI ​​prediction-based balanced charging and discharging control method for battery packs in the energy storage cabinet of the present invention.

[0023] Figure 4 This describes the specific steps involved in the AI-predictive-based balanced charge and discharge control method for battery packs in energy storage cabinets, specifically analyzing and setting the thermal coupling feature set of each battery pack in each time period within the energy storage cabinet. Detailed Implementation

[0024] Please see Figure 1This invention provides a technical solution: an AI-predictive-based method for equalizing charge and discharge control of battery packs in an energy storage cabinet, comprising the following steps: continuously acquiring operating status data, performance status data, and thermal imaging data of several battery packs in a designated energy storage cabinet for each time period (e.g., 3 seconds); based on the operating status data and energy status data of each battery pack in the designated energy storage cabinet for each time period, analyzing the battery charge and discharge degradation characteristic values ​​and energy decay characteristic values ​​for the corresponding time periods; and based on a pre-trained thermal evolution charge and discharge correlation model, combined with the thermal imaging time series data of each battery pack in the designated energy storage cabinet for each time period, analyzing its phase... Based on the characteristic values ​​of charge and discharge thermal regulation for each time period, and combined with the characteristic values ​​of battery charge and discharge degradation and energy decay, the characteristic values ​​of charge and discharge regulation demand for the corresponding time period are analyzed (the magnitude of these values ​​directly reflects the degree of intervention required during that time period; larger values ​​indicate that the battery operation deviates from a stable state, accompanied by the risk of overcharging and over-discharging). Based on the characteristic values ​​of charge and discharge regulation demand for each battery pack in each time period of the set energy storage cabinet, the predicted characteristic values ​​of charge and discharge regulation demand for the corresponding battery pack in the next time period are analyzed. Based on the predicted characteristic values ​​of charge and discharge regulation demand, balanced charge and discharge control is performed on the corresponding battery pack in the set energy storage cabinet for the next time period.

[0025] The specific formula for calculating the characteristic value of the charging and discharging regulation demand of a certain battery pack in a certain energy storage cabinet at a certain time period is as follows: ;in, To set the characteristic value of the charging and discharging regulation demand of a certain battery pack in an energy storage cabinet for a certain period of time, To set the battery charge-discharge degradation characteristic value of a specific battery pack in an energy storage cabinet at a specific time period, The charge / discharge degradation adjustment coefficients are stored in the database. To set the energy decay characteristic value of a specific battery pack in an energy storage cabinet for a specific time period, The energy attenuation adjustment coefficient is stored in the database. To set the charging and discharging thermal regulation characteristic value of a certain battery pack in an energy storage cabinet for a certain period of time, These are the thermal regulation adjustment coefficients stored in the database. , The difference adjustment coefficient is stored in the database, and in this embodiment, the charge-discharge degradation adjustment coefficient is stored in the database. Energy attenuation adjustment coefficient Thermal regulation adjustment coefficient Difference adjustment coefficient The values ​​were 0.381, 0.267, 0.352, and 0.724, respectively.

[0026] Specifically, such as Figure 2As shown, the operating status data includes voltage fluctuation amplitude, SOC value, internal resistance change rate, interface electrical micro-disturbance value, electromagnetic leakage field strength value, and terminal current pulsation response value. The specific steps for analyzing the battery charging and discharging degradation characteristics of each battery pack in the set energy storage cabinet for each time period are as follows: Based on the operating status data of each battery pack in the set energy storage cabinet for each time period, analyze the corresponding abnormal operating characteristic set, including battery operating imbalance characteristic value and electrical transmission abnormal characteristic value; Based on the abnormal operating characteristic set of each battery pack in the set energy storage cabinet for each time period, analyze the corresponding battery charging and discharging degradation characteristics, specifically: weight the battery operating imbalance characteristic value and electrical transmission abnormal characteristic value of each battery pack in the set energy storage cabinet for each time period to obtain the corresponding battery charging and discharging degradation characteristics. The larger the value, the more unstable the battery pack is during the charging and discharging process in that time period.

[0027] The voltage fluctuation amplitude value is the degree of fluctuation of the battery pack terminal voltage during the specified period. It can be obtained by setting a high-precision voltage sensor at the battery pack terminal to sample the voltage signal, extract the maximum voltage, minimum voltage, and average voltage during the specified period, and perform ratio processing, i.e. (maximum voltage - minimum voltage) / average voltage. The result is used as the voltage fluctuation amplitude value. If the voltage fluctuation amplitude value increases significantly during charging, it indicates that the polarization of the battery pack is enhanced or close to the charging limit. If the voltage fluctuation amplitude value increases during discharging, it indicates that the electrode reaction of the battery pack is unstable during the discharging process.

[0028] The State of Charge (SOC) value represents the proportion of the battery pack's remaining charge to its rated capacity. It can be obtained by installing a high-precision Hall current sensor at the battery pack terminals to collect the current signal of the battery pack during charging or discharging in real time, performing time integration processing, and then comparing the result with the rated capacity. The result is then used as the SOC value.

[0029] The internal resistance change rate is the rate at which the equivalent internal resistance of the battery pack changes over time. It can be obtained by acquiring the voltage and current values ​​at each time point within a given period using a high-precision voltage sensor and a high-precision Hall current sensor. The voltage and current values ​​at adjacent time points are then compared, and the instantaneous internal resistance change value is obtained by performing a moving average process. This result is used as the internal resistance change rate value. During charging, if the internal resistance change rate of the battery pack increases rapidly, it indicates severe polarization or accelerated aging. During discharging, if the internal resistance change rate rises rapidly, it indicates a decline in the energy output capability of the battery pack.

[0030] The interface electrical micro-perturbation value is the minute electrical disturbance generated during current flow at the connection point between the battery pack and the busbar. This can be addressed by connecting a high-speed voltage sensor in parallel at the connection terminals and a high-speed current sensor in series at the same location to acquire the interface voltage and current signals. A bandpass filter (e.g., 100Hz–10kHz) is then used to retain the high-frequency disturbance components (i.e., voltage and current disturbance signals). A Fast Fourier Transform (FFT) is then performed on these signals to obtain the amplitude-frequency response curves of the voltage and current disturbance signals in the frequency domain, i.e., the amplitude distribution at each frequency point. The amplitude sequence within a preset main frequency band (e.g., 500Hz–5kHz) is selected from the amplitude-frequency characteristic curve, and the root mean square values ​​of voltage perturbation and current perturbation are extracted and standardized. Based on the results, a weighted average is performed to obtain the interface electrical micro-perturbation value. During charging: if the interface electrical micro-perturbation value increases, it indicates that the contact resistance fluctuates during high current injection, which may lead to local temperature rise and contact point heating, thereby limiting the charging current of the battery pack. During discharging: if the interface electrical micro-perturbation value is abnormal, it indicates that the current conduction path is unstable, which may cause transient arcing or energy loss during high current output, and the output path is unstable.

[0031] The electromagnetic leakage field strength is the electromagnetic field strength caused by the change in current during the charging and discharging process of the battery pack. It can be obtained by arranging Hall flux sensors around the battery pack to collect the electromagnetic induction signal of the battery pack in real time during the charging and discharging process; the signal is converted into a time series of magnetic induction intensity through a signal conditioning circuit, and the amplitude tracking processing is performed on the time series to obtain the magnetic field strength at different time points; based on this, the maximum magnetic field strength of the time period is extracted and used as the electromagnetic leakage field strength value. During charging: a large electromagnetic leakage field strength value indicates abnormal current distribution or significant local inrush current phenomenon, which can easily lead to electromagnetic interference and local heating; during discharging, an increase in the electromagnetic leakage field strength value indicates that there is a risk of overcurrent or short circuit in the discharge path.

[0032] The terminal current ripple response value is the fluctuation characteristic of the current signal contained in the battery pack terminals during charging and discharging. It can be obtained by placing a high-frequency Hall current sensor at the battery pack terminals to collect the current waveform of the battery pack in real time during charging and discharging. After signal conditioning, the current waveform is analyzed in both the time and frequency domains. In the time domain, the ripple amplitude of the current waveform is extracted, and in the frequency domain, the current ripple amplitude of specific frequency components (such as the switching frequency component in the range of 5–20kHz) is extracted using Fast Fourier Transform (FFT). The square root of the sum of the squares of the current ripple amplitude and the ripple amplitude is taken as the terminal current ripple response value. During charging: a large terminal current ripple response value indicates that strong ripple or transient pulsation is superimposed in the charging current, which can easily lead to enhanced polarization and unstable electrochemical reaction. During discharging: an increased terminal current ripple response value indicates that the discharge current output fluctuates violently and the power transmission quality is degraded.

[0033] The specific steps for analyzing the abnormal operation feature set of each battery pack in the set energy storage cabinet for each time period are as follows: Based on the voltage fluctuation amplitude, SOC value, and internal resistance change rate value of each battery pack in the set energy storage cabinet for each time period, analyze the battery operation imbalance feature value for the corresponding time period. Specifically, the voltage fluctuation amplitude, SOC value, and internal resistance change rate value of each battery pack in the set energy storage cabinet for each time period are standardized. A weighted average is then applied based on the standardized values. During discharge, the standardized SOC value is inverted, i.e., 1 / (1 + standardized SOC value), to obtain the battery operation characteristic value for the corresponding time period. Imbalance characteristic values ​​are used to characterize the degree of imbalance of the battery pack body during charging and discharging. Based on the interface electrical micro-perturbation value, electromagnetic leakage field strength value, and terminal current pulsation response value of each battery pack in the set energy storage cabinet for each time period, the electrical transmission abnormal characteristic values ​​of the corresponding time period are analyzed. Specifically, the interface electrical micro-perturbation value, electromagnetic leakage field strength value, and terminal current pulsation response value of each battery pack in the set energy storage cabinet for each time period are standardized, and the standardized results are weighted to obtain the electrical transmission abnormal characteristic values ​​of the corresponding time period, which are used to comprehensively characterize the transmission instability of the battery pack during charging and discharging.

[0034] In this implementation plan, by conducting in-depth analysis of operational status data, operational imbalance characteristic values ​​and electrical transmission anomaly characteristic values ​​are constructed. This standardizes and weights the relationships between different parameters into a unified framework, avoiding confusion caused by differences in parameter dimensions. For example, during charging, if voltage fluctuations and internal resistance changes are increasing while the SOC is abnormally low, this abnormal combination can be directly reflected by the imbalance characteristic value. Similarly, interface disturbances, electromagnetic leakage, and current pulsations may not seem significant individually, but when combined, they can greatly affect transmission stability. This can be intuitively reflected by the electrical transmission anomaly characteristic value. Finally, the comprehensive battery charging and discharging degradation characteristic value can quickly reflect whether the battery pack has entered an unstable state, thereby reducing the complexity of manual judgment and improving operational safety.

[0035] Specifically, the energy state data includes energy conversion efficiency, electromagnetic energy leakage factor, and pulsating energy ratio. The specific steps for analyzing the energy decay characteristic value of each battery pack in the set energy storage cabinet for each time period are as follows: Read the energy state data of each battery pack in the set energy storage cabinet for each time period and perform standardization processing (i.e., standardize the energy conversion efficiency, electromagnetic energy leakage factor, and pulsating energy ratio of each battery pack in the set energy storage cabinet for each time period); Perform comprehensive analysis on the standardized energy state data of each battery pack in the set energy storage cabinet for each time period to obtain the corresponding energy decay characteristic value. Specifically, perform weighted processing on the standardized energy conversion efficiency, electromagnetic energy leakage factor, and pulsating energy ratio of each battery pack in the set energy storage cabinet for each time period. During the weighted processing, the standardized energy conversion efficiency value is inverted, i.e., 1 / (1+standardized energy conversion efficiency value), to obtain the corresponding energy decay characteristic value, which is used to characterize the overall energy transfer degradation level of the battery pack during the charging and discharging process.

[0036] The energy conversion efficiency (ECE) value represents the degree of energy utilization by the battery pack during charging and discharging. High-precision voltage and current sensors can be placed at the battery pack terminals to collect terminal voltage and current signals during charging and discharging. By multiplying the collected voltage and current signals, the power sequence for that period is obtained, and time integration is performed to obtain the charging input energy or discharging output energy for that period. By integrating the collected current signal, the charge increment for that period is obtained, and this increment is compared with the battery's rated capacity (which can be obtained from the technical specifications stored in the database) to obtain the SOE for that period. The change in SOC is mapped to the battery's open-circuit voltage – SOC calibration curve. This curve is used to obtain the voltage level corresponding to the change in SOC. This voltage level is then combined with the charge increment to obtain the actual stored or released usable energy during that period. The ratio of the actual stored or released usable energy to the corresponding input or output energy is then calculated to obtain the energy conversion efficiency value. During charging, the energy conversion efficiency value is equal to the ratio of the actual stored usable energy to the charging input energy; during discharging, the energy conversion efficiency value is equal to the ratio of the actual released usable energy to the discharging output energy.

[0037] The electromagnetic energy leakage factor is the relative proportion of electrical energy leaked into the external environment in the form of an electromagnetic field during the charging and discharging process of a battery pack. It can be determined by placing Hall effect magnetic induction sensors around the battery pack busbars or current paths to collect the magnetic induction signals generated during charging and discharging. These signals are then converted into voltage signals by a signal conditioning circuit, and the time-series value of the magnetic induction intensity is obtained through an analog-to-digital converter. Based on this, the equivalent electromagnetic energy leakage is obtained by square integration of the magnetic induction intensity. This value is then compared with the input or output energy during that time period to calculate the electromagnetic energy leakage factor. The pulsating energy ratio is the proportion of the energy corresponding to the AC component (ripple, pulsation) in the current signal during the charging and discharging process of the battery pack to the total energy. It can be obtained by placing a high-frequency Hall current sensor at the battery pack terminals to collect the current waveform during the charging and discharging process in real time. After filtering, the collected current signal can be separated into DC and AC components. The DC component is used to characterize the average flow part of the current, and the AC component is used to characterize the ripple and pulsation components superimposed on the current. The DC component is multiplied by the terminal voltage signal and accumulated over time to obtain the DC energy in that time period. Then, the AC component is multiplied by the terminal voltage signal and accumulated over time to obtain the AC energy in that time period. By comparing the ratio of the sum of AC energy and DC energy, i.e., AC energy / (AC energy + DC energy), the pulsating energy ratio is obtained.

[0038] In this implementation scheme, the energy state data of the battery pack during charging and discharging is uniformly quantified to comprehensively reflect the degradation level of the battery pack's energy transmission. Secondly, the electromagnetic leakage energy factor and pulsating energy ratio are introduced to quantify and supplement the implicit channels of energy loss. For example, in the charging scenario, even if the energy conversion efficiency is high, if the electromagnetic leakage energy factor increases, it indicates that some electrical energy is leaked out in the form of electromagnetic fields, which has a substantial impact on the overall utilization rate. In the discharging scenario, if the pulsating energy ratio increases significantly, it means that there is a strong ripple component in the output energy, and the transmission stability decreases. Finally, through standardization and weighted fusion, the influence of the difference in dimensions between different physical quantities can be avoided. At the same time, by inverting the energy conversion efficiency value, its correlation in the energy decay process is strengthened, thereby enabling a more comprehensive characterization of the energy transmission degradation of the battery pack under actual working conditions.

[0039] Specifically, the thermal imaging data includes the pixel value (i.e., temperature value) and two-dimensional coordinates of each pixel in several frames of thermal imaging. The thermal evolution charge-discharge correlation model includes an input layer, a heat flow analysis layer, a charge-discharge thermal evolution layer, and an output layer.

[0040] The specific steps for analyzing the charge-discharge thermal regulation characteristics of each battery pack in the energy storage cabinet for each time period are as follows: Input the thermal imaging data of each battery pack in the energy storage cabinet for each time period into the pre-trained thermal evolution charge-discharge correlation model, and analyze the charge-discharge thermal coupling characteristic set of the corresponding time period, including hot spot link offset characteristic value, hierarchical thermal disturbance intensity characteristic value, and thermal spectrum misalignment characteristic value; Based on the charge-discharge thermal coupling characteristic set of each battery pack in the energy storage cabinet for each time period, analyze the charge-discharge thermal regulation characteristic value of the corresponding time period (used to characterize the temperature change of the battery pack during overcharging or over-discharging. During charging, if a local cell in the battery pack is in an overcharged state, its electrode reaction will enhance the polarization effect, which manifests as a regional temperature rise; during discharging, if a local cell in the battery pack is in an over-discharged state, its internal resistance will rise sharply, which manifests as abnormal heating).

[0041] The specific formula for calculating the charging and discharging thermal regulation characteristic value of a certain battery pack in a certain time period of a set energy storage cabinet is as follows: ;in, To set the charging and discharging thermal regulation characteristic value of a certain battery pack in an energy storage cabinet for a certain period of time, To set the hotspot link offset characteristic value of a certain battery pack in an energy storage cabinet for a certain period of time, These are the hotspot link adjustment coefficients stored in the database. To set the characteristic value of the hierarchical thermal disturbance intensity of a certain battery pack in an energy storage cabinet at a certain time period, These are the hierarchical thermal disturbance adjustment coefficients stored in the database. To set the thermal spectral imbalance characteristic value of a certain battery pack in an energy storage cabinet for a certain period of time, These are the thermal spectral misalignment adjustment coefficients stored in the database. These are the smoothing adjustment coefficients stored in the database, and in this embodiment, the hotspot link adjustment coefficients stored in the database. Hierarchical thermal disturbance adjustment coefficient Thermal spectral misalignment adjustment coefficient Smoothing adjustment coefficient The values ​​were 0.384, 0.462, 0.421, and 2.000, respectively.

[0042] The following is a specific implementation example for calculating the charging and discharging thermal regulation characteristic values ​​of a battery pack in a set energy storage cabinet during a certain time period. The available data includes: hotspot link offset characteristic values, hierarchical thermal disturbance intensity characteristic values, and thermal spectrum imbalance characteristic values ​​for five randomly selected time periods of a battery pack in the set energy storage cabinet, as detailed in Table 1 and... Figure 3 As shown: Table 1. Example of time-series data set showing the charging and discharging thermal coupling characteristics of a specific battery pack in an energy storage cabinet.

[0043] Hotspot link adjustment coefficients stored in the database The value is: 0.384; Hierarchical thermal disturbance adjustment coefficients stored in the database The value is: 0.462; Thermal spectral offset adjustment coefficients stored in the database The value is: 0.421; Smoothing adjustment coefficients stored in the database The value is: 2.000; Substituting the data from Table 1 and the aforementioned coefficients into the specific formula for calculating the charging and discharging thermal regulation characteristic value of a certain battery pack in a certain time period of the energy storage cabinet, we obtain: The characteristic value of the charge-discharge thermal regulation of a certain battery pack in the energy storage cabinet during the first period is set as ln(1+((0.254^0.384+0.312^0.462+0.332^0.421) / 2.000))≈0.643; The characteristic value of charge-discharge thermal regulation for a certain battery pack in the energy storage cabinet during the second time period is set as ln(1+((0.324^0.384+0.364^0.462+0.373^0.421) / 2.000))≈0.677; The characteristic value of charge-discharge thermal regulation for a certain battery pack in the energy storage cabinet during the third time period is set as ln(1+((0.294^0.384+0.415^0.462+0.352^0.421) / 2.000))≈0.676; The characteristic value of the charge-discharge thermal regulation of a certain battery pack in the energy storage cabinet during the fourth time period is set as ln(1+((0.362^0.384+0.402^0.462+0.428^0.421) / 2.000))≈0.701; The characteristic value of the charge and discharge thermal regulation of a certain battery pack in the energy storage cabinet during the fifth period is set as ln(1+((0.423^0.384+0.436^0.462+0.442^0.421) / 2.000))≈0.720.

[0044] like Figure 4 As shown, the specific steps for analyzing the charge-discharge thermal coupling feature set of each battery pack in each time period of the energy storage cabinet are as follows: In the input layer of the thermal evolution charge-discharge correlation model, the thermal imaging data (i.e., the pixel value and two-dimensional coordinates of each pixel point in each frame of thermal imaging) of each battery pack in the energy storage cabinet for each time period are received and preprocessed. Specifically, this involves calling a semantic segmentation network (such as SegNet, U-Net, Mask) R-CNN performs recognition processing on each frame of thermal imaging (i.e., the semantic segmentation network is a deep convolutional neural network structure that can classify each pixel in the input image to distinguish each battery pack region from the background region. In specific implementation, the semantic segmentation network usually adopts an encoder-decoder structure: the encoder part is used to extract multi-level features of the input image layer by layer, and the decoder part is used to restore the features to the same resolution as the input image step by step. The encoder part extracts the edge features, temperature distribution patterns and geometric structural features of each battery pack through convolution and downsampling operations; the decoder part restores the above features to the pixel-by-pixel predicted image through deconvolution and feature fusion operations, thereby generating the pixel classification results of battery packs and non-battery packs and the corresponding confidence scores. The pixel classification results output by the network are thresholded and filtered, and pixels with confidence scores lower than the set threshold are removed. Connectivity analysis is performed, such as using 4-neighborhood or 8-neighborhood, to obtain several candidate regions, and the candidate regions with an area higher than the set area threshold are selected as battery pack regions, and the area of ​​the candidate region is the number of pixels in the candidate region), so as to obtain the pixel set of each battery pack region. In the thermal flow analysis layer of the thermal evolution charge-discharge correlation model, based on the preprocessed thermal imaging data of each battery pack in each time period of the set energy storage cabinet, thermal anomaly feature vectors are extracted from the corresponding frame thermal imaging. In the charge-discharge thermal evolution layer of the thermal evolution charge-discharge correlation model, based on the thermal anomaly feature vectors in each frame thermal imaging of each battery pack in each time period of the set energy storage cabinet, thermal evolution time-series feature vectors for the corresponding time period are extracted. Specifically, the thermal anomaly feature vectors are constructed into a time-series feature sequence according to the frame order and input into an LSTM. The LSTM dynamically adjusts the weights of historical information and current input information through a gating mechanism, thereby capturing the change pattern of feature vectors over time during charging and discharging. In this process, the update gate and reset gate of the cell control the retention and forgetting of historical features, respectively, so that the model can simultaneously focus on long-term evolution trends and instantaneous fluctuation features, thereby generating thermal evolution time-series feature vectors that can reflect the thermal anomaly evolution law of the battery pack during charging and discharging, such as: For the hotspot link connectivity features in the thermal anomaly feature vector of each frame of thermal imaging, a ratio processing is performed, such as the ratio of the hotspot link connectivity features of the first frame to the second frame / the hotspot link connectivity features of the second frame, and then the mean is applied. Simultaneously, a moving average processing is performed on the hotspot link connectivity features in the frame thermal imaging. This is then standardized along with the mean processing result, and the standardized result is weighted to extract the hotspot link offset feature. This feature is used to characterize the excessive extension trend of hotspot links over time during charging and discharging. During charging, a large feature value indicates that the connectivity of the hotspot links is continuously increasing, suggesting that the charging current is accumulating in local paths, which can easily lead to overheating channels and local heat accumulation between cells. During discharging, a large feature value indicates that the hotspot path continues to expand during energy release, suggesting uneven current distribution and the risk of local overcurrent and unstable discharge paths. For the thermal anomaly feature vector in each frame of thermal imaging, variance processing is performed on the thermal penetration level feature to extract the level thermal disturbance intensity feature. This feature is used to characterize the temperature difference expansion between different ring-shaped levels of the battery pack during charging and discharging. During charging, if the feature value is large, it indicates that the temperature rise of the outer cell is significant while the temperature rise of the inner cell is not synchronous, reflecting that the charging rate is too fast or the heat conduction is uneven. During discharging, if the feature value is large, it indicates that the energy release of the inner cell is aggravated while the heat dissipation of the periphery is insufficient, reflecting the internal heat accumulation and the uneven effect between levels during discharging. For the thermal spectrum offset feature in the thermal anomaly feature vector of each frame of thermal imaging, the minimum and maximum values ​​of the thermal spectrum offset feature are extracted and their ratios are processed to extract the thermal spectrum misalignment feature. This feature is used to characterize the energy shift of heat distribution in direction and scale during charging and discharging. During charging, if this feature value is large, it indicates that there is unstable contact resistance in the current path or overcharging of a single cell, which manifests as sharp local heat accumulation. During discharging, if this feature value is large, it indicates abnormal discharge current shunting, resulting in instability of the overall output path of the battery pack. The hot spot link offset feature, hierarchical thermal disturbance intensity feature, and thermal spectrum misalignment feature are concatenated into a thermal evolution time series feature vector. In the output layer of the thermal evolution charge-discharge correlation model, based on the thermal evolution time series feature vector of each battery pack in the energy storage cabinet for each time period, the charge-discharge thermal coupling feature set of the corresponding time period is output. Specifically, the hot spot link offset feature, hierarchical thermal disturbance intensity feature, and thermal spectrum misalignment feature in the thermal evolution time series feature vector of each battery pack for each time period are processed by the Sigmoid function, and the results are mapped between 0 and 1 to obtain the hot spot link offset feature value, hierarchical thermal disturbance intensity feature value, and thermal spectrum misalignment feature value of the corresponding time period.

[0045] The specific steps for extracting the thermal anomaly feature vector from each frame of thermal imaging for each battery pack in each time period of the set energy storage cabinet are as follows: For the pixel value (i.e., temperature value) of each pixel in the pixel set of each battery pack region in each frame of thermal imaging for each time period of the preprocessed set energy storage cabinet, extract the mean temperature and standard deviation of each battery pack region in the corresponding frame of thermal imaging, and set a temperature threshold based on this, i.e., mean temperature + 2 × standard deviation of temperature. Mark the pixels with a temperature higher than the set temperature threshold as hotspot pixels. Based on the 8-neighborhood connectivity principle, mark the hotspot pixels with connected components to obtain several hotspot connected clusters. For the two-dimensional coordinates of each hotspot pixel in each hotspot connected cluster, construct a covariance matrix and perform eigenvalue decomposition. Align the direction with larger eigenvalues ​​with the principal axis direction of the hotspot connected cluster, and align the direction with smaller eigenvalues ​​with the vertical principal axis direction of the hotspot connected cluster. Project the two-dimensional coordinates of each hotspot pixel in the hotspot connected cluster onto the principal axis direction to obtain the projection value of the corresponding hotspot pixel. Extract the maximum and minimum projection values, and take the difference between the maximum and minimum projection values ​​as the length value of the hotspot connected cluster. Similarly, the two-dimensional coordinates of each hot spot pixel in the hot spot connected cluster are projected onto the vertical principal axis to obtain the width value of the hot spot connected cluster. The aspect ratio of the hot spot connected cluster is then processed. Hot spot connected clusters with aspect ratios greater than a set aspect ratio threshold are selected and regarded as hot spot links. The sum of the length values ​​of all hot spot links is calculated and compared with the diagonal length of each battery pack region (i.e., the maximum and minimum values ​​of the horizontal and vertical coordinates are counted in all pixels to form an circumscribed rectangle, and the Euclidean distance between the diagonal points of the rectangle is calculated, i.e., the Euclidean distance between the diagonal point corresponding to the maximum value of the horizontal and vertical coordinates and the diagonal point corresponding to the minimum value of the horizontal and vertical coordinates). This is used to extract the hot spot link connectivity features in the corresponding thermal imaging. During the charging process, if this feature is large, it indicates that the local cell or busbar area is continuously heated during the current injection process, and the hot spots gradually connect into a sheet, which can easily lead to uneven charging channels and local overheating. During the discharging process, if this feature is large, it indicates that the energy release is uneven, the local cell bears too much current, forming a hot spot path, and there is a risk of overcurrent. For each battery pack in each time period of each frame of thermal imaging of each battery pack region in the preprocessed energy storage cabinet, the two-dimensional coordinates of each pixel are averaged to obtain the center two-dimensional coordinates of each battery pack region. The distance value from each pixel to the center two-dimensional coordinates of the corresponding battery pack region is extracted, and the maximum distance value is selected. Then, it is divided into n sub-intervals (rings). The annular temperature mean of each sub-interval (i.e., the mean temperature value of each pixel in the sub-interval) is extracted. The annular temperature difference value (absolute value) between adjacent sub-intervals is extracted. The annular temperature difference variance, annular temperature difference mean, and annular temperature difference maximum are extracted and weighted to extract the heat penetration layer feature. During charging, if this feature is large, it indicates that the temperature of the outer cell rises rapidly, but the inner cell is not balanced in time, indicating that the charging rate is too fast and there are uneven current distribution and boundary heat accumulation problems. During discharging, if this feature is large, it indicates that the inner cell heats up first, while the outer cell heats up insufficiently, the consistency between cells decreases, and the discharge path is unbalanced. For the temperature value of each pixel in the pixel set of each battery pack region in each frame of thermal imaging for each time period of each battery pack in the preprocessed setting energy storage cabinet, it is first reconstructed into a temperature matrix according to two-dimensional coordinates, and then a two-dimensional fast Fourier transform (FFT) is performed on the temperature matrix to obtain the corresponding spectrum distribution (including the frequency coordinates and temperature amplitude corresponding to each frequency component). The point where the zero frequency component is located is taken as the spectrum center. With the spectrum center as the center, the relative radius distance of each frequency component is calculated. The frequency domain is divided into several ring band regions according to the relative radius distance. If the radius is lower than the lower limit of the preset relative radius distance interval, it is divided into a low frequency ring band region. If it is within the preset relative radius distance interval, it is divided into a mid frequency ring band region. If it is higher than the upper limit of the preset relative radius distance interval, it is divided into a high frequency ring band region. The battery pack is divided into several sectors according to angular direction (e.g., every 30°). Based on this, the squares of the temperature amplitudes within each annular zone and sector are accumulated to obtain the sum of temperature energy values ​​for each annular zone and each sector. The ratio of the sum of temperature energy values ​​between adjacent annular zones is extracted and averaged to obtain the inter-ring energy ratio. Similarly, the ratio of the sum of temperature energy values ​​between adjacent sectors is extracted and averaged to obtain the sector energy ratio. This ratio is then weighted with the inter-ring energy ratio to extract thermal spectrum shift features. During charging, a large feature indicates that the charging current is unstable in that direction, manifesting as sharp local thermal fluctuations. During discharging, a large feature indicates uneven distribution of the discharging current along the path, leading to unstable energy release paths in the battery pack. The hotspot link connectivity feature, thermal penetration level feature, and thermal spectrum shift feature are concatenated into a thermal anomaly feature vector.

[0046] Furthermore, the pre-training steps for the thermal evolution charge-discharge correlation model are as follows: A labeled dataset is obtained, which consists of time-series thermal imaging data of battery packs under different operating conditions and operational status monitoring data of energy storage cabinets. The data comes from long-term operation records of actual energy storage systems and is combined with risk event information manually annotated by experts. Each sample in the labeled dataset includes: thermal imaging frame sequences of each battery pack in multiple consecutive time periods, corresponding operational status labels (such as overcharge, over-discharge, thermal imbalance, etc.), and ground truth labels of thermal anomaly features. The dataset is divided into training set, validation set, and test set. For example, 80% of the samples are used for training, 10% for validation, and 10% for testing.

[0047] The thermal evolution charge-discharge correlation model is trained. Taking the charge-discharge thermal evolution layer as an example, the extracted thermal anomaly feature vectors of each frame are constructed into a time sequence and input into LSTM or Gated Recurrent Unit (GRU). Through its gating mechanism, long-term dependence and short-term fluctuations are captured. During the training process, the network optimizes the weights of the time sequence layer through the backpropagation algorithm (BPTT) to minimize the error between the predicted value and the true labeled value. The loss function can adopt mean squared error (MSE) or weighted loss to take into account the imbalanced samples under overcharge and over-discharge scenarios.

[0048] Model training and optimization employs the Adam optimizer or RMSProp optimizer to dynamically adjust weight parameters. Hyperparameters such as learning rate, batch size, and number of LSTM hidden units are adjusted during training through grid search or Bayesian optimization. The validation set is used to monitor model performance during training to avoid overfitting, and an early stopping mechanism is used to determine the optimal number of training epochs.

[0049] The model was evaluated on the test set to verify its generalization ability on unseen thermal imaging time series data. Evaluation metrics included accuracy, recall, and F1-score. After the test, the trained thermal evolution charge-discharge correlation model and its parameter files were saved for subsequent online operation to achieve the extraction of thermal anomaly features and prediction of adjustment requirements during the battery pack charge-discharge process.

[0050] In this implementation scheme, a pre-trained thermal evolution charge-discharge correlation model is introduced, and in-depth analysis is performed using thermal imaging data of each battery pack at each time period. This significantly improves the precision of balanced charge-discharge control of the battery pack in the energy storage cabinet. Specifically, by calling a semantic segmentation network to extract pixels in the battery pack area, background interference can be avoided, ensuring that the thermal imaging data is accurately focused on the battery pack itself. Secondly, in the thermal flow analysis layer, feature extraction methods such as hotspot links, hierarchical thermal penetration, and thermal spectrum shift are used to reveal complex phenomena such as local overheating of the battery and to form quantifiable features that are easy to correlate directly with the charge-discharge state. Furthermore, in the charge-discharge thermal evolution layer, time-series modeling is used to capture the dynamic changes of features over time, which can identify the thermal anomaly evolution trend of the battery during fast charging or high-power discharge, helping to detect potential overcharging or over-discharging runaway risks in advance. Finally, the output thermal coupling feature set can be mapped to a standardized range of 0–1, which is convenient for fusion analysis with other operating features and can achieve high-precision evaluation of the battery operating state.

[0051] Specifically, the steps for analyzing the predicted charge and discharge regulation demand characteristic values ​​for each battery pack in the next time period of the energy storage cabinet are as follows: Based on the predicted charge and discharge regulation demand characteristic values ​​for each battery pack in the energy storage cabinet for each time period, analyze the predicted demand baseline value (i.e., the mean of the charge and discharge regulation demand characteristic values ​​for each time period), demand fluctuation factor (i.e., the variance of the charge and discharge regulation demand characteristic values ​​for each time period), and regulation demand inertia factor for the corresponding battery pack; Based on the predicted demand baseline value, demand fluctuation factor, and regulation demand inertia factor for each battery pack in the energy storage cabinet, analyze the predicted charge and discharge regulation demand characteristic values ​​for the next time period of the corresponding battery pack, i.e., predicted demand baseline value + demand fluctuation factor × tanh (regulation demand inertia factor), to obtain the predicted charge and discharge regulation demand characteristic values ​​for each battery pack in the next time period.

[0052] The specific steps for analyzing the regulation demand inertia factor of each battery pack in the energy storage cabinet are as follows: Based on the charging and discharging regulation demand characteristic values ​​of each battery pack in the energy storage cabinet for each time period, analyze the regulation demand change characteristic values ​​of several groups of adjacent time periods for the corresponding battery pack. That is, perform ratio processing on the charging and discharging regulation demand characteristic values ​​of adjacent time periods, such as the difference between the regulation demand change characteristic values ​​of the first time period and the second time period / the regulation demand change characteristic value of the second time period; Based on the regulation demand change characteristic values ​​of each battery pack in the energy storage cabinet for each group of adjacent time periods, analyze the regulation demand inertia factor of the corresponding battery pack. That is, perform ratio processing on any two groups of adjacent time periods for the regulation demand change characteristic values, and perform weighted processing based on the ratio processing to obtain the regulation demand inertia factor of the corresponding battery pack, thereby characterizing the inertial effect of the regulation demand evolution of the battery pack during charging and discharging.

[0053] In this implementation scheme, by introducing a predicted demand baseline value, a demand fluctuation factor, and a regulating demand inertia factor, the baseline level, fluctuation intensity, and inertial continuity of the battery pack's regulating demand during charging and discharging can be extracted, thereby effectively revealing the potential deviation trend of battery operation. Secondly, the introduction of the regulating demand inertia factor enables the model to capture the accelerated change characteristics of regulating demand over time, avoiding the one-sidedness of prediction based solely on instantaneous data. Finally, the inertia factor is normalized to ensure that the prediction results maintain both positive and negative fluctuations while avoiding abnormal amplification effects, thereby achieving dynamic and smooth output of the predicted values. This allows for early identification of the battery pack's regulating risk level in the next period, thereby improving the overall safety and stability of operation.

[0054] Specifically, the steps for equalizing charge and discharge control of the corresponding battery packs in the next time period based on the predicted charge and discharge regulation demand characteristic values ​​of the set energy storage cabinet are as follows: The predicted charge and discharge regulation demand characteristic values ​​of each battery pack in the next time period of the set energy storage cabinet are compared with the preset (next time period) charge and discharge regulation demand threshold range. The steps for obtaining the upper and lower limits of the charge and discharge regulation demand threshold range of the next time period are as follows: Based on the predicted charge and discharge regulation demand characteristic values ​​of each battery pack in the next time period of the set energy storage cabinet, the mean value of the predicted charge and discharge regulation demand characteristic and the standard deviation of the predicted charge and discharge regulation demand characteristic are extracted respectively. The upper limit = the mean value of the predicted charge and discharge regulation demand characteristic + 3 × the standard deviation of the predicted charge and discharge regulation demand characteristic, and the lower limit = the mean value of the predicted charge and discharge regulation demand characteristic - 3 × the standard deviation of the predicted charge and discharge regulation demand characteristic. Based on the comparison and processing results, preset equalization charge and discharge control measures are adopted for the corresponding battery packs in the next time period of the set energy storage cabinet. Specifically: if the predicted charge and discharge regulation demand characteristic value of each battery pack in the next time period of the set energy storage cabinet is higher than the upper limit of the preset charge and discharge regulation demand threshold range, then during charging: it indicates that the battery pack is at extremely high risk of overcharging and is identified as an overcharge risk source; during discharging: this situation indicates that the battery pack is at extremely high risk of over-discharging and is identified as an over-discharge risk source; if the predicted charge and discharge regulation demand characteristic value of each battery pack in the next time period of the set energy storage cabinet is within the preset charge and discharge regulation demand threshold range, then it is determined that the charge and discharge of the battery pack is close to stable and only normal charge and discharge is maintained; if the predicted charge and discharge regulation demand characteristic value of each battery pack in the next time period of the set energy storage cabinet is lower than the lower limit of the preset charge and discharge regulation demand threshold range, then during charging: this situation indicates that the battery pack's power is significantly low and is identified as an object that needs energy replenishment; during discharging: this situation indicates that the battery pack's power is sufficient and its health is good, and it is stable energy. During charging, for battery packs at risk of overcharging, a powerful active balancing mechanism is immediately activated, treating them as energy senders. Excess energy in these battery packs is transferred out through an active balancing circuit (such as a bidirectional DC-DC converter). The transfer process, specifically to battery packs requiring additional energy, involves the following steps: For several battery packs at risk of overcharging, the difference between the average predicted charge / discharge regulation demand characteristic and the upper limit of the charge / discharge regulation demand threshold range is extracted to obtain the risk overflow value for each overcharging risk source. These values ​​are then sorted in descending order to generate an overcharging risk ranking table. For several battery packs requiring additional energy, the difference between the average predicted charge / discharge regulation demand characteristic and the upper limit of the charge / discharge regulation demand threshold range is also extracted. The absolute value of the difference between the lower limits is used to obtain the energy shortage value of each energy replenishment object. These values ​​are then sorted in descending order to generate an energy replenishment sorting table. Battery packs in the same sequence as those in the overcharge risk sorting table are matched to establish matching pairs for energy transfer. The transferred energy is selected based on the minimum value between the risk overflow value and the energy shortage value in the matching pair. If the number of overcharge risk sources is greater than the number of energy replenishment objects, a single energy replenishment object is allowed to receive energy from multiple risk sources simultaneously. The system scheduling module limits the power of each channel to avoid overload. If the number of energy replenishment objects is greater than the number of risk sources, the risk sources allocate energy to multiple energy replenishment objects according to priority until their transferable energy is exhausted. During discharge, for sources of over-discharge risk, protective balancing is immediately initiated, treating them as objects requiring protection and prioritizing their energy replenishment. Through the active balancing circuit, energy is transferred from the battery pack with stable energy to this battery pack, and the transfer steps are consistent with the logic during charging.

[0055] In this implementation scheme, a method for comparing predicted charge and discharge regulation demand characteristic values ​​with dynamic threshold ranges is introduced to achieve differentiated control of the charge and discharge state of the battery pack. Secondly, through a bidirectional sorting and matching mechanism of risk overflow value and energy shortage value, optimal pairing of energy transfer is achieved to ensure that the excess energy of high-risk batteries can preferentially flow to batteries with insufficient energy, thereby balancing safety and energy utilization efficiency. In addition, in the active balancing and protective balancing corresponding to overcharge risk and over-discharge risk, respectively, the system can automatically select the energy transfer direction and avoid overload of the balancing circuit by limiting energy transfer, thereby ensuring the reliability and stability of operation. Finally, this step not only improves the accuracy and dynamism of energy distribution between battery packs, but also significantly enhances the reliability of the energy storage cabinet under complex operating conditions.

[0056] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0057] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for balanced charging and discharging control of battery packs in energy storage cabinets based on AI prediction, characterized in that, Includes the following steps: Continuously acquire the operating status data, performance status data, and thermal imaging data of several battery packs in a set energy storage cabinet for each time period; Based on the operating status data and energy status data of each battery pack in the energy storage cabinet for each time period, the battery charging and discharging degradation characteristic value and energy decay characteristic value for the corresponding time period are analyzed respectively. Based on the pre-trained thermal evolution charge-discharge correlation model, and combined with the thermal imaging time series data of each battery pack in the set energy storage cabinet for each time period, the charge-discharge thermal regulation characteristic value of the corresponding time period is analyzed. Combined with the battery charge-discharge degradation characteristic value and energy decay characteristic value, the charge-discharge regulation demand characteristic value of the corresponding time period is analyzed. Based on the characteristic values ​​of the charging and discharging regulation demand of each battery pack in each time period of the set energy storage cabinet, the predicted characteristic values ​​of the charging and discharging regulation demand of the corresponding battery pack in the next time period are analyzed. Based on the predicted charging and discharging regulation demand characteristics, the next time period is set for the corresponding battery pack of the energy storage cabinet to perform balanced charging and discharging control.

2. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 1, characterized in that, The operational status data includes voltage fluctuation amplitude, SOC value, internal resistance change rate, interface electrical micro-perturbation value, electromagnetic leakage field strength value, and terminal current pulsation response value. The specific steps for analyzing and setting the battery charging and discharging degradation characteristic values ​​of each battery pack in the energy storage cabinet for each time period are as follows: Based on the operational status data of each battery pack in the energy storage cabinet for each time period, the abnormal operation feature set of the corresponding time period is analyzed, including battery operation imbalance feature value and electrical transmission abnormal feature value. Based on the set of abnormal operation characteristics of each battery pack in the energy storage cabinet for each time period, the battery charging and discharging degradation characteristic values ​​of the corresponding time period are analyzed.

3. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 2, characterized in that, The specific steps for analyzing and setting the abnormal operation feature set of each battery pack in the energy storage cabinet for each time period are as follows: Based on the voltage fluctuation amplitude, SOC value, and internal resistance change rate of each battery pack in the energy storage cabinet for each time period, the battery operation imbalance characteristics of the corresponding time period are analyzed. Based on the interface electrical micro-perturbation value, electromagnetic leakage field strength value, and terminal current pulsation response value of each battery pack in the energy storage cabinet for each time period, the abnormal electrical transmission characteristic values ​​of the corresponding time period are analyzed.

4. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 1, characterized in that, The specific steps for analyzing and setting the energy decay characteristic value of each battery pack in the energy storage cabinet for each time period are as follows: Read the energy status data of each battery pack in the designated energy storage cabinet for each time period and perform standardization processing; The energy state data of each battery pack in the standardized energy storage cabinet for each time period are comprehensively analyzed to obtain the energy decay characteristic value of the corresponding time period.

5. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 1, characterized in that, The thermal imaging data includes the pixel value and two-dimensional coordinates of each pixel in several frames of thermal imaging. The thermal evolution charge-discharge correlation model includes an input layer, a heat flow analysis layer, a charge-discharge thermal evolution layer, and an output layer.

6. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 5, characterized in that, The specific steps for analyzing and setting the charging and discharging thermal regulation characteristic values ​​of each battery pack in the energy storage cabinet for each time period are as follows: The thermal imaging data of each battery pack in the energy storage cabinet for each time period is input into the pre-trained thermal evolution charge-discharge correlation model to analyze the charge-discharge thermal coupling feature set of the corresponding time period, including hot spot link offset feature value, hierarchical thermal disturbance intensity feature value, and thermal spectrum misalignment feature value. Based on the set of charge-discharge thermal coupling characteristics of each battery pack in the energy storage cabinet for each time period, the charge-discharge thermal regulation characteristic values ​​of the corresponding time period are analyzed.

7. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 6, characterized in that, The specific steps for analyzing and setting the charge-discharge thermal coupling characteristic set of each battery pack in the energy storage cabinet for each time period are as follows: In the input layer of the thermal evolution charge-discharge correlation model, thermal imaging data of each battery pack in each time period of the set energy storage cabinet is received and preprocessed. In the thermal flow analysis layer of the thermal evolution charge-discharge correlation model, based on the preprocessed thermal imaging data of each battery pack in each time period of the set energy storage cabinet, the thermal anomaly feature vector in the corresponding frame thermal imaging is extracted. In the charge-discharge thermal evolution layer of the thermal evolution charge-discharge correlation model, based on the thermal anomaly feature vector in each frame of thermal imaging of each battery pack of each time period in the set energy storage cabinet, the thermal evolution time sequence feature vector of the corresponding time period is extracted. In the output layer of the thermal evolution charge-discharge correlation model, based on the thermal evolution time sequence feature vector of each battery pack in the energy storage cabinet for each time period, the charge-discharge thermal coupling feature set of the corresponding time period is output.

8. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 1, characterized in that, The specific steps for analyzing and setting the predicted charge and discharge regulation demand characteristic values ​​for each battery pack in the energy storage cabinet for the next time period are as follows: Based on the characteristic values ​​of charging and discharging regulation demand for each battery pack in each time period of the energy storage cabinet, the predicted demand benchmark value, demand fluctuation factor, and regulation demand inertia factor of the corresponding battery pack are analyzed. Based on the predicted demand baseline, demand fluctuation factor, and adjustment demand inertia factor for each battery pack in the energy storage cabinet, the predicted charging and discharging adjustment demand characteristic value of the corresponding battery pack in the next period is analyzed.

9. The AI-predictive-based balanced charging and discharging control method for battery packs in energy storage cabinets according to claim 8, characterized in that, The specific steps for analyzing and setting the adjustment demand inertia factor for each battery pack in the energy storage cabinet are as follows: Based on the characteristic values ​​of the charging and discharging regulation demand of each battery pack in the energy storage cabinet for each time period, the characteristic values ​​of the regulation demand change of several groups of adjacent time periods of the corresponding battery pack are analyzed. Based on the characteristic values ​​of the regulation demand change of each battery pack in each adjacent time period of the set energy storage cabinet, the regulation demand inertia factor of the corresponding battery pack is analyzed.

10. The AI-predictive-based method for equalizing charge and discharge control of battery packs in energy storage cabinets according to claim 1, characterized in that, The specific steps for equalizing charge and discharge control of the corresponding battery pack in the energy storage cabinet for the next time period based on the predicted charge and discharge regulation demand characteristics are as follows: The predicted charge and discharge regulation demand characteristic value for each battery pack in the next time period of the energy storage cabinet is compared with the preset charge and discharge regulation demand threshold range. Based on the comparison and processing results, preset equalization charging and discharging control measures are adopted for the corresponding battery packs in the next time period of the set energy storage cabinet.

Images