Intelligent electric energy adjusting device and method for distributed energy storage transformer area

Abstract

The invention relates to the technical field of energy storage system scheduling, and discloses an intelligent electric energy adjusting device and method for a distributed energy storage area. Performing weighted calculation on a health degree index through the acquisition module; converting into a power allocation weight through a calculation module; a monitoring module predicts a health degree decline amount and dynamically adjusts a protection coefficient; and adaptively switching a scheduling mode through a switching module according to the global health balance degree. According to the method, the health degree evaluation precision, the power distribution rationality, the health state balance, the overall service life of the system and the operation economy of the distributed energy storage transformer area are improved.

Description

Technical Field

[0001] This application relates to the field of energy storage system scheduling technology, and in particular to a power intelligent regulation device and method for distributed energy storage areas. Background Technology

[0002] With the rapid development of distributed energy systems, energy storage technology is playing an increasingly important role in grid peak shaving and valley filling, renewable energy consumption, and power quality regulation. In distributed energy storage areas, multiple energy storage units are typically deployed to meet the area's load demand. Existing energy storage dispatching methods are mainly based on state-of-charge (SOC) balancing strategies, which monitor the SOC of each energy storage unit and allocate power according to demand to achieve charge and discharge control of the energy storage system. Some existing technologies also consider single health indicators such as cycle count and capacity decay, allocating more usage opportunities to energy storage units with higher health during power allocation to extend the overall system lifespan. These methods, to some extent, improve the economy and reliability of energy storage systems.

[0003] However, existing technologies have the following shortcomings: First, the health assessment is based on a single dimension, relying solely on individual parameters such as cycle count or capacity decay, which cannot comprehensively reflect the true health status of the energy storage unit, leading to a discrepancy between the health assessment results and the actual degree of degradation. Second, the power allocation strategy lacks a dynamic response mechanism to the health status. When using fixed weighting coefficients for power allocation, it is impossible to flexibly adjust the allocation strategy according to the real-time health status and load changes of the energy storage unit. Third, it is difficult to balance health protection and economic operation. Overly pursuing economic efficiency will accelerate the degradation of high-health energy storage, while over-protection will reduce the overall utilization efficiency of the system. Summary of the Invention

[0004] This application provides a power intelligent regulation device and method for distributed energy storage areas, which solves a series of technical problems in existing energy storage scheduling technologies, such as incomplete health assessment, lack of dynamic response mechanism in power allocation strategy, Matthew effect reversal caused by health-aware scheduling, lack of forward-looking health protection based on future prediction, and inability to adaptively switch scheduling modes according to global health balance. It improves the accuracy of health assessment, rationality of power allocation, balance of health status, overall system lifespan and operating economy of distributed energy storage areas.

[0005] In a first aspect, this application provides a power intelligent regulation device for a distributed energy storage area. The power intelligent regulation device for the distributed energy storage area includes: a data acquisition module, used to acquire the cycle number, internal resistance growth rate, capacity decay rate and charge / discharge depth historical sequence of each energy storage unit, and to calculate the health index of each energy storage unit by weighted fusion. The calculation module is used to convert the health index into power allocation attention weights, calculate the charging and discharging power commands of each energy storage unit in combination with the state of charge and the load demand of the distribution area, and execute power allocation. The monitoring module includes a monitoring unit, a prediction unit, and a calculation unit. The monitoring unit is used to monitor the difference between the health change rate of the high-load energy storage group and the health change rate of the normal energy storage group. The prediction unit is used to predict the future health decline of each energy storage unit when the difference exceeds a threshold. The calculation unit is used to dynamically amplify the health protection coefficient according to the ratio of the health decline to a set threshold, recalculate the charging and discharging power command, and distribute the resulting power gap to the medium-healthy energy storage group according to the remaining capacity weight. The switching module is used to calculate the global health balance as the ratio of the standard deviation to the mean of the health of all energy storage units, and adaptively switch the scheduling mode and adjust the weight coefficient combination according to the numerical range of the global health balance.

[0006] Secondly, this application provides a method for intelligent regulation of power in a distributed energy storage area, the method comprising: Step S1: Collect the historical sequences of cycle count, internal resistance growth rate, capacity decay rate and charge / discharge depth of each energy storage unit, and calculate the health index of each energy storage unit by weighted fusion. Step S2: Convert the health index into power allocation attention weights, calculate the charging and discharging power commands of each energy storage unit in combination with the state of charge and the load demand of the distribution area, and execute power allocation. Step S3: Monitor the difference between the health change rate of the high-load energy storage group and the health change rate of the normal energy storage group. When the difference exceeds the threshold, predict the future health decline of each energy storage unit. Dynamically amplify the health protection coefficient according to the ratio of the health decline to the set threshold, recalculate the charging and discharging power command, and distribute the resulting power gap to the medium-health energy storage units according to the remaining capacity weight. Step S4: Calculate the global health balance as the ratio of the standard deviation to the mean of the health of all energy storage units, and adaptively switch the scheduling mode and adjust the weight coefficient combination according to the numerical range of the global health balance.

[0007] The technical solution provided in this application comprehensively collects historical data on the cycle count, internal resistance growth rate, capacity decay rate, and charge / discharge depth of each energy storage unit through a data acquisition module. A health index is then calculated through weighted fusion, achieving a shift from single-dimensional health assessment to multi-dimensional fusion assessment. This allows the health index to simultaneously reflect the comprehensive health status of the energy storage unit, including lifespan depletion, performance degradation, and usage stress, avoiding assessment biases caused by relying on a single parameter. The calculation module transforms the health index into a power allocation attention weight design, automatically granting higher power allocation priority to energy storage units with higher health. Combined with state of charge and substation load demand, this achieves intelligent power allocation based on health perception, optimizing the usage strategy of each energy storage unit while meeting load requirements.

[0008] The monitoring module detects new imbalances caused by differentiated power allocation by monitoring the rate of change in health status between high-load and normal energy storage groups. When the difference exceeds a threshold, it triggers the prediction unit to predict future health degradation, shifting health management from passive response to proactive prevention. The calculation unit dynamically amplifies the health protection coefficient and recalculates the charging and discharging power commands based on the prediction results. Power limits are imposed on energy storage units subjected to excessive loads, while the resulting power gap is distributed to medium-healthy energy storage units according to their remaining capacity weights. This avoids accelerated degradation of high-healthy energy storage due to continuous high loads and fully utilizes the remaining capacity of medium-healthy energy storage units, achieving a reasonable redistribution of load. The switching module calculates the global health balance and adaptively switches the scheduling mode according to its numerical range. This allows the system to prioritize economy when the health distribution is balanced, gradually increase the weights of health protection and balance when health differentiation occurs, and enforce a balance strategy in dangerous states of highly differentiated health status. By dynamically adjusting the weight coefficient combination, it achieves refined hierarchical control over different health balance states, avoiding strategy mismatch problems caused by fixed weight configurations. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of one embodiment of the intelligent power regulation device for a distributed energy storage area in this application. Figure 2 This is a schematic diagram illustrating the verification of the accuracy of LSTM network health prediction in the embodiments of this application; Figure 3This is a schematic diagram illustrating the distribution characteristics of energy storage unit health under different scheduling modes in the embodiments of this application. Detailed Implementation

[0011] This application provides a smart power regulation device and method for a distributed energy storage area. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a 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 explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0012] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent power regulation device for distributed energy storage areas in this application includes: The data acquisition module is used to collect the historical sequence of cycle number, internal resistance growth rate, capacity decay rate and charge / discharge depth of each energy storage unit, and to calculate the health index of each energy storage unit by weighted fusion. The data acquisition module synchronously acquires the operating parameters of each energy storage unit through data acquisition equipment with a sampling period of 5 minutes. The cycle count is accumulated and statistically analyzed from historical operating logs. The internal resistance growth rate is measured every 24 hours in a static state using the AC impedance method, which measures the ratio of the current internal resistance value to the initial internal resistance value. The capacity decay rate is calculated by dividing the difference between the rated capacity and the actual usable capacity by the rated capacity. The charge / discharge depth historical sequence records the difference between the maximum and minimum state of charge each day over the past 30 days. The weighted fusion calculation of the health index multiplies the cycle life health, internal resistance health, capacity health, and charge / discharge depth health by weighting coefficients of 0.35, 0.25, 0.25, and 0.15, respectively, and then sums them. The cycle life health is equal to 1 minus the ratio of the current cycle count to the rated cycle life; the internal resistance health is equal to the ratio of the initial internal resistance to the current internal resistance; the capacity health is equal to 1 minus the capacity decay rate; and the charge / discharge depth health is calculated by substituting the historical sequence mean into the exponential decay function.

[0013] The calculation module is used to convert the health index into power allocation attention weights, calculate the charging and discharging power commands of each energy storage unit in combination with the state of charge and the load demand of the distribution area, and execute power allocation. The health assessment subnetwork constructed by the computation module adopts a three-layer fully connected structure. The input layer receives an 8-dimensional feature vector including the number of cycles, internal resistance growth rate, capacity decay rate, mean charge / discharge depth, standard deviation of charge / discharge depth, state of charge, ambient temperature, and load demand of the distribution area. After feature extraction by 16 neurons in the first hidden layer and 8 neurons in the second hidden layer, a corrected health index is output. The transformation process of power allocation attention weights involves substituting the corrected health index into the natural exponential function to obtain the exponential value, and then summing the exponential values ​​of all energy storage units as the denominator for normalization, so that energy storage units with higher health levels receive larger weight values. The attention layer of the scheduling decision subnetwork concatenates the normalized weights with the corrected health index and the current state of charge to form an enhanced feature vector. After processing by two hidden layers, the charging and discharging power commands of each energy storage unit are output. The power commands are verified by upper and lower limits of state of charge and rated power limit constraints.

[0014] The monitoring module includes a monitoring unit, a prediction unit, and a calculation unit. The monitoring unit is used to monitor the difference between the health change rate of the high-load energy storage group and the health change rate of the normal energy storage group. The prediction unit is used to predict the future health decline of each energy storage unit when the difference exceeds a threshold. The calculation unit is used to dynamically amplify the health protection coefficient according to the ratio of the health decline to a set threshold, recalculate the charging and discharging power command, and distribute the resulting power gap to the medium-healthy energy storage group according to the remaining capacity weight. The monitoring module's monitoring unit calculates the average charge / discharge power of each energy storage unit at each sampling point over the past hour, using a 1-hour monitoring cycle. Simultaneously, it accumulates the number of time points where the absolute value of the charge / discharge power exceeds 80% of the rated power, multiplying this by the sampling interval to obtain the high-power operating time. When the average charge / discharge power exceeds 60% of the rated power and the high-power operating time exceeds 0.5 hours, the energy storage unit is classified into the high-load energy storage group. The prediction unit collects the health status values ​​of each energy storage unit over the past 72 hours to form a time series. This data is input into the input layer of a Long Short-Term Memory (LSTM) network, containing 72 time steps. After processing through two hidden layers of 64 LSTM units each, it outputs a health status prediction sequence for the next 24 hours. The average of the 24 values ​​in the prediction sequence is calculated to obtain the average predicted health status. The future health status decrease is obtained by subtracting the average predicted health status from the current health status index. The calculation unit divides the future health decline by a set threshold of 0.08 to obtain a ratio. When the ratio is greater than 1, the basic health protection coefficient of 0.3 is multiplied by the ratio and then 1 is added to obtain the dynamically amplified health protection coefficient. The amplification coefficient is substituted into the multi-objective loss function and the neural network forward propagation is re-executed to calculate the new charging and discharging power command. The difference between the original power command and the new power command is summed on the high-load energy storage group to obtain the power gap. The remaining capacity of each energy storage unit in the medium-health group is obtained by subtracting the current state of charge from the maximum state of charge of 90%. The remaining capacity of all energy storage units in the medium-health group is summed as the denominator, and the remaining capacity of each unit is divided by the sum to obtain the normalized remaining capacity weight. The power gap is multiplied by the remaining capacity weight and then added to the charging and discharging power command of the energy storage units in the medium-health group.

[0015] The switching module is used to calculate the global health balance as the ratio of the standard deviation to the mean of the health of all energy storage units, and adaptively switch the scheduling mode and adjust the weight coefficient combination according to the numerical range of the global health balance.

[0016] Specifically, when the switching module calculates the global health balance, it first calculates the standard deviation and arithmetic mean of the health index of all energy storage units. The standard deviation is divided by the mean to obtain the global health balance value. The balance warning threshold is set to 0.25 and the danger threshold is set to 0.40. When the global health balance is less than 0.15, the weight coefficient combination corresponding to the economic priority mode is selected as follows: economic cost weight 0.6, health degradation rate weight 0.2, and energy storage utilization balance weight 0.2. When the global health balance is between 0.15 and 0.25, the weight coefficient combination corresponding to the health protection mode is selected as follows: 0.5, 0.3, and 0.2. When the global health balance is between 0.25 and 0.40, the weight coefficient combination corresponding to the balance intervention mode is selected as follows: 0.3, 0.35, and 0.35. When the global health balance is greater than or equal to 0.40, the weight coefficient combination corresponding to the forced balance mode is selected as follows: 0, 0.5, and 0.5. The selected weight coefficient combination is substituted into the multi-objective loss function to adjust the charging and discharging power command allocation strategy of each energy storage unit.

[0017] In one specific embodiment, the acquisition module is used for: The current state of charge, cumulative number of charge-discharge cycles, measured internal resistance, and ratio of rated capacity to actual usable capacity of each energy storage unit are collected to obtain the capacity decay rate. Calculate cycle life health, internal resistance health, capacity health, and depth of charge / discharge health respectively; The health index of each energy storage unit is obtained by weighted summation of the cycle life health, internal resistance health, capacity health, and depth of charge / discharge health. A first health threshold and a second health threshold are set. Based on the relationship between the health index and the first and second health thresholds, the energy storage units are divided into a high health group, a medium health group, and a low health group.

[0018] Specifically, the capacity decay rate is calculated by comparing the rated capacity parameters of the energy storage unit at the time of manufacture with the actual usable capacity obtained through a complete charge-discharge cycle test. The difference between the two is divided by the rated capacity and multiplied by 100% to convert it into a percentage. This value directly reflects the degree of energy storage capacity loss caused by chemical performance degradation. Cycle life health is obtained by subtracting the ratio of the current cumulative number of charge-discharge cycles to the rated cycle life of the energy storage unit from 1. Internal resistance health is obtained by dividing the initial internal resistance value of the energy storage unit by the current measured internal resistance value. Capacity health is obtained by subtracting the percentage value of the capacity decay rate from 1 and then dividing by 100. Depth of charge health is obtained by calculating the average of the historical charge-discharge depth series over the past 30 days, substituting it into the exponential decay function with 50 as the denominator, taking the negative natural logarithm, and then taking the exponential value.

[0019] The weighting coefficients in the weighted summation process reflect the degree of influence of each health dimension on the overall health status of the energy storage unit. Cycle count, as the most direct indicator of lifespan depletion, is assigned the highest weight of 0.35; internal resistance growth and capacity decay, as core characteristics of performance degradation, are each assigned a weight of 0.25; and depth of charge / discharge, as a usage stress indicator, is assigned a weight of 0.15. The four sub-items of health are multiplied by their corresponding weights and then summed to obtain a comprehensive health index between 0 and 1. Health status stratification uses a first health threshold of 0.75 and a second health threshold of 0.50 as dividing points. Energy storage units with a health index greater than or equal to 0.75 are classified into the high-health group, indicating good performance; those with a health index less than 0.50 are classified into the low-health group, indicating nearing the end of their lifespan and requiring limited use; and those with a health index between 0.50 and 0.75 are classified into the medium-health group, indicating normal degradation. This three-level stratification provides a basis for subsequent differentiated power allocation strategies.

[0020] In one specific embodiment, the calculation module is used for: Construct a health assessment subnetwork and a scheduling decision subnetwork; The mean and standard deviation of the historical sequence of charge and discharge depth of each energy storage unit are input into the health assessment sub-network, and the corrected health index is output. Substitute the modified health index into the exponential function and normalize it to obtain the power allocation attention weight of each energy storage unit. The power allocation attention weight, the corrected health index, the state of charge, and the load demand of the distribution area are input into the scheduling decision sub-network, and the charging and discharging power commands of each energy storage unit are output.

[0021] Specifically, the health assessment sub-network adopts a three-layer fully connected neural network architecture. The input layer receives multi-dimensional feature vectors including the number of cycles, internal resistance growth rate, capacity decay rate, and the mean and standard deviation of the historical charge / discharge depth sequence. The first hidden layer has 16 neurons that use the ReLU activation function to perform nonlinear transformation and feature extraction on the input features. The second hidden layer has 8 neurons that perform feature dimensionality reduction and deep abstraction. The output layer uses the Sigmoid activation function to map the neuron output values ​​to the 0-1 range to obtain the corrected health index. This correction process calibrates and optimizes the health index calculated based on the formula by learning the complex nonlinear relationship between health and degradation parameters in historical data. The mean of the historical charge / discharge depth sequence reflects the average usage intensity of the energy storage unit, and the standard deviation reflects the degree of fluctuation in usage intensity. Using both the mean and standard deviation as input features allows the network to simultaneously capture the impact of the central trend and dispersion of usage intensity on the health status.

[0022] The calculation of power allocation attention weights involves substituting the modified health index of each energy storage unit into the natural exponential function to obtain the exponential value. Then, the exponential values ​​of all energy storage units are summed as the normalized denominator. The normalized attention weight is obtained by dividing the exponential value of each energy storage unit by the sum. This weight is exponentially positively correlated with the health index, resulting in energy storage units with higher health levels receiving significantly higher power allocation priority. The scheduling decision subnetwork adopts a four-layer structure enhanced by the attention mechanism. The input layer concatenates four types of data—power allocation attention weight, modified health index, current state of charge, and area load demand—to form an enhanced feature vector. The attention layer weights the features according to the weight coefficients, highlighting the scheduling importance of high-health energy storage. The first hidden layer uses 32 neurons and the Tanh activation function to process the weighted features. The second hidden layer uses 16 neurons for deep feature fusion. The output layer directly outputs the charging and discharging power command values ​​of each energy storage unit, which are then corrected by the upper and lower limits of rated power and the safe range of state of charge to ensure that the power command is within the physically feasible domain.

[0023] In one specific embodiment, the monitoring unit is used for: The average charging and discharging power of each energy storage unit in the past period is calculated based on the set monitoring cycle, and the high-power operation time when the absolute value of the charging and discharging power exceeds the set ratio of the rated power is statistically analyzed. High-load energy storage groups were selected based on the average charge / discharge power and the high-power operating time. Calculate the mean health change rate of the high-load energy storage group and the mean health change rate of the normal energy storage group, and obtain the difference in health change rate by subtracting them; When the difference in the rate of change of health status exceeds a threshold, a new imbalance is identified.

[0024] Specifically, the average charge / discharge power is calculated using a 1-hour monitoring cycle. Within this cycle, the charge / discharge power values ​​of each data sampling point for each energy storage unit are summed and divided by the total number of sampling points to obtain the arithmetic mean of past periods. The high-power operation duration is statistically analyzed by iterating through all sampling points within the time period to determine whether the absolute value of the charge / discharge power exceeds 80% of the rated power of the energy storage unit. The number of sampling points that meet the condition is multiplied by the sampling time interval and summed to obtain the duration of high-power operation. The selection of high-load energy storage groups is based on dual criteria: the average charge / discharge power of the energy storage unit must exceed 60% of its rated power and the high-power operation duration must exceed 0.5 hours. Energy storage units that meet both conditions are classified into the high-load energy storage group, while those that do not meet the conditions are classified into the normal energy storage group.

[0025] The health rate of change is calculated by subtracting the health index of each energy storage unit from its historical health index one hour prior, thus determining the decrease in health rate per unit time. The arithmetic mean of the health rate of change for all energy storage units in the high-load energy storage group is obtained, and the same arithmetic mean is calculated for all energy storage units in the normal energy storage group. The difference between the high-load group's average and the normal group's average is then calculated to obtain the health rate of change difference. When this difference is less than -0.02 (meaning the high-load group's health rate of decline is more than 2% faster than the normal group), it is determined that the difference exceeds a set threshold. This identifies a new imbalance phenomenon where the differentiated power allocation strategy causes the high-health energy storage to degrade more rapidly due to bearing a larger load. This monitoring mechanism captures the side effects of the protection strategy in real time, providing trigger signals for subsequent dynamic adjustments.

[0026] In one specific embodiment, the prediction unit is used for: Collect the health time series of each energy storage unit over a set number of hours in the past; The health time series is input into a long short-term memory network to predict the health series for a set number of hours in the future. Calculate the average value of the health status sequence for the specified future hours to obtain the average predicted health status; The difference between the current health index and the average predicted health index is used to obtain the future health decline of each energy storage unit.

[0027] Specifically, the health status time series was collected by backtracking 72 hours with a 1-hour time step. The health index values ​​of each energy storage unit at each hour node were extracted in chronological order to form a time series array containing 72 elements. This sequence completely records the evolution trajectory of the energy storage unit's health status over the past 3 days. The Long Short-Term Memory (LSTM) network adopts a structure with an input layer of 72 time steps, two hidden layers of 64 LSTM units each, and an output layer of 24 time steps. The input layer inputs the 72 historical health status values ​​one by one into the LSTM units. The hidden layers capture the long-term dependencies and short-term fluctuation patterns of health status degradation through a gating mechanism. The output layer predicts the health status values ​​at each hour node for the next 24 hours to form the future health status sequence. The LSTM network is trained using historical health status records from the past 180 days to learn degradation patterns.

[0028] The average predicted health level is calculated by summing 24 predicted health levels over the next 24 hours and dividing by 24 to obtain the arithmetic mean. This average represents the expected health level of the energy storage unit within the next day. The future health decline is calculated by subtracting the average predicted health level from the current health index. A positive difference indicates a predicted decline in future health; a larger difference indicates a faster expected degradation rate. When the difference exceeds a set threshold of 0.08 (i.e., a predicted health decline of more than 8% over the next 24 hours), the energy storage unit is marked as overwork risk and its usage intensity needs to be reduced. This predictive mechanism shifts health management from a passive response to proactive prevention.

[0029] Figure 2 This is a schematic diagram illustrating the verification of the accuracy of LSTM network health prediction in the embodiments of this application. Figure 2 This paper demonstrates the accuracy verification results of the Long Short-Term Memory (LSTM) network in predicting the health status of energy storage units over the next 24 hours. In the figure, black dots represent the actual measured health index, the gray solid line represents the health trend predicted by the LSTM network based on historical data from the past 72 hours, and the light gray shaded area represents the prediction confidence interval. The figure shows that the LSTM prediction curve closely matches the actual health data points, with a mean absolute error (MAE) of only 0.0048, indicating a prediction accuracy of over 99.5%. This figure verifies that the LSTM network used in this application can effectively capture the long-term dependencies and short-term fluctuation patterns of health degradation, accurately predicting the future health decline of each energy storage unit. When the predicted health decline exceeds a set threshold of 0.08, the system can identify energy storage units at risk of overwork in advance and dynamically amplify the health protection coefficient, transforming health management from a passive response to proactive prevention, thus avoiding accelerated degradation of energy storage units due to continuous high-load operation.

[0030] In one specific embodiment, the computing unit is used for: The ratio of the future health decline to the set threshold is calculated. When the ratio is greater than 1, the basic health protection coefficient is multiplied by the ratio and 1 is added to obtain the dynamically amplified health protection coefficient. Substitute the dynamically amplified health protection coefficient into the multi-objective loss function to recalculate the charging and discharging power commands for each energy storage unit. Calculate the difference between the original charge / discharge power command and the recalculated charge / discharge power command for overwork risk energy storage, and sum them to obtain the power gap; The difference between the maximum state of charge and the current state of charge of each energy storage unit in the healthy group is calculated and normalized to obtain the remaining capacity weight. The power gap is multiplied by the remaining capacity weight and then added to the charging and discharging power command of the medium-health group energy storage unit.

[0031] Specifically, the ratio calculation divides the future health decline of each energy storage unit by a set threshold of 0.08 to obtain the degradation rate multiple. When the ratio is greater than 1, it indicates that the expected degradation rate exceeds the normal level. At this time, the basic health protection coefficient of 0.3 is multiplied by the ratio and then 1 is added to obtain the dynamically amplified health protection coefficient. This coefficient increases non-linearly with the increase of the expected degradation rate to achieve enhanced protection for high-risk energy storage. The multi-objective loss function includes three weighted combination parts: economic cost term, health degradation rate term, and energy storage utilization balance term. The dynamically amplified health protection coefficient replaces the original fixed coefficient of 0.3 as the weight of the health degradation rate term, which significantly increases the penalty intensity for high-power charging and discharging of overwork risk energy storage during the loss function optimization process. When the neural network re-executes the forward propagation calculation, it automatically reduces the power allocation of the energy storage unit to obtain a new charging and discharging power command.

[0032] The power gap calculation iterates through all overwork-risk energy storage units. The original charge / discharge power command before adjustment for each unit is subtracted from the recalculated charge / discharge power command to obtain a single gap value. The gap values ​​of all overwork-risk energy storage units are summed to obtain the total power gap. This gap needs to be supplemented by other energy storage units to maintain power balance in the distribution area. The remaining capacity weight calculation is performed for each energy storage unit in the medium-health group. The safe upper limit of the state of charge (90%) is subtracted from the current state of charge value of the energy storage unit to obtain the remaining charging space. The remaining charging space of all energy storage units in the medium-health group is summed as the normalized denominator. The remaining charging space of each energy storage unit is divided by the sum to obtain a remaining capacity weight value between 0 and 1. This weight reflects the ability of each energy storage unit to handle additional power. The total power gap is multiplied by the remaining capacity weight of each energy storage unit to obtain the supplementary power allocated to that energy storage unit. The supplementary power is then added to the original charge / discharge power command of that energy storage unit to complete the load distribution process.

[0033] In one specific embodiment, the switching module is used for: Calculate the standard deviation and mean of the health index of all energy storage units, and then compare the standard deviation with the mean to obtain the global health balance. Set an equilibrium warning threshold and a danger threshold. Based on the relationship between the global health equilibrium and the equilibrium warning threshold and danger threshold, divide the scheduling into four modes: economic priority mode, health protection mode, equilibrium intervention mode and forced equilibrium mode. Select a combination of the corresponding economic cost weight coefficient, health degradation rate weight coefficient, and energy storage utilization balance weight coefficient according to the scheduling mode. The weighted coefficients are substituted into the multi-objective loss function to adjust the charging and discharging power commands of each energy storage unit.

[0034] Specifically, the calculation of global health balance first involves calculating the arithmetic mean of the health indices of all energy storage units to obtain the mean parameter. Then, the square root of the sum of squared deviations of each energy storage unit's health index from the mean is taken by dividing it by the total number of energy storage units to obtain the standard deviation parameter. Dividing the standard deviation by the mean yields the global health balance value, which is a relative measure of the degree of health dispersion; a larger value indicates a more severe differentiation in the health status of each energy storage unit. The scheduling modes are divided using a balance warning threshold of 0.25 and a danger threshold of 0.40 as dividing points. When the global health balance is less than 0.15, it is classified as the economic priority mode; when the global health balance is between 0.15 and 0.25, it is classified as the health protection mode; when the global health balance is between 0.25 and 0.40, it is classified as the balance intervention mode; and when the global health balance is greater than or equal to 0.40, it is classified as the forced balance mode. These four modes correspond to different optimization objectives and focuses.

[0035] The selection of weight coefficient combinations is based on the scheduling mode, which determines the values ​​of the three weight parameters. The economic priority mode uses a combination of economic cost weight of 0.6, health degradation rate weight of 0.2, and energy storage utilization balance weight of 0.2 to focus on reducing operating costs. The health protection mode uses a combination of 0.5, 0.3, and 0.2 to balance economic efficiency and health protection. The balance intervention mode uses a combination of 0.3, 0.35, and 0.35 to strengthen the balance utilization goal. The forced balance mode uses a combination of 0, 0.5, and 0.5 to completely abandon economic optimization and focus on health balance. The multi-objective loss function sums up the economic cost term, the health degradation rate term, and the energy storage utilization balance term after multiplying them by their respective weighting coefficients. The economic cost term is the cumulative product of the electricity price and the purchased power. The health degradation rate term is the cumulative product of the absolute value of the charging and discharging power and the square of the difference in health. The energy storage utilization balance term is the ratio of the standard deviation to the mean of the state of charge of each energy storage unit. The neural network re-optimizes and calculates the charging and discharging power commands of each energy storage unit with the goal of minimizing the weighted loss function, thereby realizing the adaptive switching of the scheduling strategy from economic orientation to health balance orientation.

[0036] Figure 3 This is a schematic diagram illustrating the distribution characteristics of energy storage unit health under different scheduling modes in the embodiments of this application. Figure 3This paper compares the distribution characteristics of the health index of energy storage units under four scheduling strategies: economic priority mode, health protection mode, equilibrium intervention mode, and forced equilibrium mode. The figure shows that the economic priority mode has the largest standard deviation of health index at 0.111, indicating severe differentiation in the health status of each energy storage unit. The health protection mode reduces the standard deviation to 0.069, significantly improving the dispersion of health index. The equilibrium intervention mode further compresses the standard deviation to 0.050. The forced equilibrium mode achieves the smallest standard deviation of 0.027, effectively achieving a high degree of consistency in the health status of each energy storage unit. In the box plot, the boxes represent the quartile intervals of health index, the median line reflects the median level, the whiskers represent the data distribution range, and outliers are marked with circles. This figure fully verifies that this application, through adaptive switching of scheduling modes, can dynamically adjust the weight coefficient combination according to the global health equilibrium, automatically switching to a more balanced scheduling mode when the degree of health index differentiation intensifies, effectively suppressing the polarization of health status among energy storage units.

[0037] The above describes the intelligent power regulation device for the distributed energy storage area in the embodiments of this application. The following describes the intelligent power regulation method for the distributed energy storage area in the embodiments of this application. One embodiment of the intelligent power regulation method for the distributed energy storage area in the embodiments of this application includes: Step S1: Collect the historical sequences of cycle count, internal resistance growth rate, capacity decay rate and charge / discharge depth of each energy storage unit, and calculate the health index of each energy storage unit by weighted fusion. Step S2: Convert the health index into power allocation attention weights, calculate the charging and discharging power commands of each energy storage unit in combination with the state of charge and the load demand of the distribution area, and execute power allocation. Step S3: Monitor the difference between the health change rate of the high-load energy storage group and the health change rate of the normal energy storage group. When the difference exceeds the threshold, predict the future health decline of each energy storage unit. Dynamically amplify the health protection coefficient according to the ratio of the health decline to the set threshold, recalculate the charging and discharging power command, and distribute the resulting power gap to the medium-health energy storage units according to the remaining capacity weight. Step S4: Calculate the global health balance as the ratio of the standard deviation to the mean of the health of all energy storage units, and adaptively switch the scheduling mode and adjust the weight coefficient combination according to the numerical range of the global health balance.

[0038] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart power regulation device for a distributed energy storage area, characterized in that, The device includes: The data acquisition module is used to collect the historical sequence of cycle number, internal resistance growth rate, capacity decay rate and charge / discharge depth of each energy storage unit, and to calculate the health index of each energy storage unit by weighted fusion. The calculation module is used to convert the health index into power allocation attention weights, calculate the charging and discharging power commands of each energy storage unit in combination with the state of charge and the load demand of the distribution area, and execute power allocation. The monitoring module includes a monitoring unit, a prediction unit, and a calculation unit. The monitoring unit is used to monitor the difference between the health change rate of the high-load energy storage group and the health change rate of the normal energy storage group. The prediction unit is used to predict the future health decline of each energy storage unit when the difference exceeds a threshold. The calculation unit is used to dynamically amplify the health protection coefficient according to the ratio of the health decline to a set threshold, recalculate the charging and discharging power command, and distribute the resulting power gap to the medium-healthy energy storage group according to the remaining capacity weight. The switching module is used to calculate the global health balance as the ratio of the standard deviation to the mean of the health of all energy storage units, and adaptively switch the scheduling mode and adjust the weight coefficient combination according to the numerical range of the global health balance.

2. The intelligent power regulation device for distributed energy storage areas according to claim 1, characterized in that, The data acquisition module is used for: The current state of charge, cumulative number of charge-discharge cycles, measured internal resistance, and ratio of rated capacity to actual usable capacity of each energy storage unit are collected to obtain the capacity decay rate. Calculate cycle life health, internal resistance health, capacity health, and depth of charge / discharge health respectively; The health index of each energy storage unit is obtained by weighted summation of the cycle life health, internal resistance health, capacity health, and depth of charge / discharge health. A first health threshold and a second health threshold are set. Based on the relationship between the health index and the first and second health thresholds, the energy storage units are divided into a high health group, a medium health group, and a low health group.

3. The intelligent power regulation device for distributed energy storage areas according to claim 1, characterized in that, The calculation module is used for: Construct a health assessment subnetwork and a scheduling decision subnetwork; The mean and standard deviation of the historical sequence of charge and discharge depth of each energy storage unit are input into the health assessment sub-network, and the corrected health index is output. Substitute the modified health index into the exponential function and normalize it to obtain the power allocation attention weight of each energy storage unit. The power allocation attention weight, the corrected health index, the state of charge, and the load demand of the distribution area are input into the scheduling decision sub-network, and the charging and discharging power commands of each energy storage unit are output.

4. The intelligent power regulation device for distributed energy storage areas according to claim 1, characterized in that, Monitoring unit, used for: The average charging and discharging power of each energy storage unit in the past period is calculated based on the set monitoring cycle, and the high-power operation time when the absolute value of the charging and discharging power exceeds the set ratio of the rated power is statistically analyzed. High-load energy storage groups were selected based on the average charge / discharge power and the high-power operating time. Calculate the mean health change rate of the high-load energy storage group and the mean health change rate of the normal energy storage group, and obtain the difference in health change rate by subtracting them; When the difference in the rate of change of health status exceeds a threshold, a new imbalance is identified.

5. The intelligent power regulation device for distributed energy storage areas according to claim 4, characterized in that, Prediction unit, used for: Collect the health time series of each energy storage unit over a set number of hours in the past; The health time series is input into a long short-term memory network to predict the health series for a set number of hours in the future. Calculate the average value of the health status sequence for the specified future hours to obtain the average predicted health status; The difference between the current health index and the average predicted health index is used to obtain the future health decline of each energy storage unit.

6. The intelligent power regulation device for distributed energy storage areas according to claim 5, characterized in that, Computational unit, used for: The ratio of the future health decline to the set threshold is calculated. When the ratio is greater than 1, the basic health protection coefficient is multiplied by the ratio and 1 is added to obtain the dynamically amplified health protection coefficient. Substitute the dynamically amplified health protection coefficient into the multi-objective loss function to recalculate the charging and discharging power commands for each energy storage unit. Calculate the difference between the original charge / discharge power command and the recalculated charge / discharge power command for overwork risk energy storage, and sum them to obtain the power gap; The difference between the maximum state of charge and the current state of charge of each energy storage unit in the healthy group is calculated and normalized to obtain the remaining capacity weight. The power gap is multiplied by the remaining capacity weight and then added to the charging and discharging power command of the medium-health group energy storage unit.

7. The intelligent power regulation device for distributed energy storage areas according to claim 1, characterized in that, Switching modules, used for: Calculate the standard deviation and mean of the health index of all energy storage units, and then compare the standard deviation with the mean to obtain the global health balance. Set an equilibrium warning threshold and a danger threshold. Based on the relationship between the global health equilibrium and the equilibrium warning threshold and danger threshold, divide the scheduling into four modes: economic priority mode, health protection mode, equilibrium intervention mode and forced equilibrium mode. Select a combination of the corresponding economic cost weight coefficient, health degradation rate weight coefficient, and energy storage utilization balance weight coefficient according to the scheduling mode. The weighted coefficients are substituted into the multi-objective loss function to adjust the charging and discharging power commands of each energy storage unit.

8. A method for intelligent power regulation of a distributed energy storage area based on the intelligent power regulation device of the distributed energy storage area as described in any one of claims 1-7, characterized in that, The methods include: Step S1: Collect the historical sequences of cycle count, internal resistance growth rate, capacity decay rate and charge / discharge depth of each energy storage unit, and calculate the health index of each energy storage unit by weighted fusion. Step S2: Convert the health index into power allocation attention weights, calculate the charging and discharging power commands of each energy storage unit in combination with the state of charge and the load demand of the distribution area, and execute power allocation. Step S3: Monitor the difference between the health change rate of the high-load energy storage group and the health change rate of the normal energy storage group. When the difference exceeds the threshold, predict the future health decline of each energy storage unit. Dynamically amplify the health protection coefficient according to the ratio of the health decline to the set threshold, recalculate the charging and discharging power command, and distribute the resulting power gap to the medium-health energy storage units according to the remaining capacity weight. Step S4: Calculate the global health balance as the ratio of the standard deviation to the mean of the health of all energy storage units, and adaptively switch the scheduling mode and adjust the weight coefficient combination according to the numerical range of the global health balance.

9. The method according to claim 8, characterized in that, Step S1 includes: The current state of charge, cumulative number of charge-discharge cycles, measured internal resistance, and ratio of rated capacity to actual usable capacity of each energy storage unit are collected to obtain the capacity decay rate. Calculate cycle life health, internal resistance health, capacity health, and depth of charge / discharge health respectively; The health index of each energy storage unit is obtained by weighted summation of the cycle life health, internal resistance health, capacity health, and depth of charge / discharge health. A first health threshold and a second health threshold are set. Based on the relationship between the health index and the first and second health thresholds, the energy storage units are divided into a high health group, a medium health group, and a low health group.

10. The method according to claim 8, characterized in that, Step S2 includes: Construct a health assessment subnetwork and a scheduling decision subnetwork; The mean and standard deviation of the historical sequence of charge and discharge depth of each energy storage unit are input into the health assessment sub-network, and the corrected health index is output. Substitute the modified health index into the exponential function and normalize it to obtain the power allocation attention weight of each energy storage unit. The power allocation attention weight, the corrected health index, the state of charge, and the load demand of the distribution area are input into the scheduling decision sub-network, and the charging and discharging power commands of each energy storage unit are output.

Images