A battery state of health based energy storage power plant power distribution and management system

Abstract

The application discloses a kind of energy storage power station power distribution and management system based on battery health state, it is related to electric power energy storage technical field, including perception layer unit, battery health state evaluation and residual life prediction module, double-layer power distribution optimization decision module, edge control execution module, safety control and operation and maintenance management module.This system will SOH as power distribution core decision variable, through multidimensional high-precision SOH online estimation, using SOH positive proportion weighted distribution strategy, let high health degree battery bear more cycle load, aged battery reduced capacity operation, realize full stack battery aging rate balance, can make battery cycle life extension, greatly reduce the full life cycle battery replacement cost;Design double-layer collaborative architecture, realize millisecond level power instruction solution, both meet the requirement of power grid examination, and realize full life cycle benefit optimization.

Description

Technical Field

[0001] This invention relates to the field of power energy storage technology, specifically to a power distribution and management system for energy storage power stations based on battery health status. Background Technology

[0002] With the deepening of the "dual-carbon" strategy and the accelerated construction of new power systems, the high proportion of wind and solar power connected to the grid has exacerbated the contradiction between grid power fluctuations and peak-valley differences. As a core resource for flexible regulation, electrochemical energy storage has experienced explosive growth in installed capacity. The power distribution and energy management system is the core hub of an energy storage power station. Its control strategy directly determines the battery service life, the safety of power station operation, and the economics of the project throughout its entire life cycle. It is a key link in the large-scale commercialization of energy storage technology.

[0003] Current mainstream energy storage management systems all focus on battery state of charge (SOC) balancing and power command tracking as their core control objectives, but they suffer from several insurmountable technical flaws: First, they adopt an indiscriminate power sharing model, ignoring individual differences in battery state of health (SOH), leading to a severe "weakest link" effect, which accelerates the degradation of aging cells, and the actual cycle life of batteries generally only reaches 60% to 70% of the design value; second, the SOH estimation accuracy is low and the real-time performance is poor, so it can only be used as a post-operation and maintenance reference and cannot be embedded in power distribution closed-loop control; third, they cannot take into account both global lifespan optimization and the millisecond-level response requirements of the power grid, resulting in lagging safety management and a disconnect between operation and maintenance, which seriously restricts the long-term safe and economical operation of energy storage power stations. Summary of the Invention

[0004] The purpose of this invention is to provide a power distribution and management system for energy storage power stations based on battery health status, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a power allocation and management system for an energy storage power station based on battery health status, comprising a sensing layer unit, a battery health status assessment and remaining life prediction module, a two-layer power allocation optimization decision module, an edge control execution module, and a safety management and operation and maintenance management module; The sensing layer unit is used to collect full-dimensional operational data from the battery cells to the entire station level of the energy storage power station, and upload the data to the corresponding module after preprocessing. The battery health status assessment and remaining life prediction module is communicatively connected to the sensing layer unit. It is used to construct a multi-dimensional battery health status SOH assessment system based on preprocessed operating data, realize real-time online SOH estimation at the cell level through electrochemical model and AI data-driven fusion algorithm, and complete the prediction of the remaining battery life RUL based on the SOH change trend. The dual-layer power allocation optimization decision module is communicatively connected to the battery health status assessment and remaining life prediction module and the power grid dispatch system, respectively. It includes an upper long-cycle life optimization scheduling layer and a lower real-time power allocation control layer. The upper long-cycle life optimization scheduling layer is used to calculate the power allocation benchmark weight of each battery cell based on the intraday power grid dispatch plan, electricity price signal, new energy output prediction and RUL prediction results of each battery cell, with the dual objectives of minimizing the life-cycle degradation and maximizing the power plant revenue. The lower real-time power allocation control layer is used to calculate the millisecond-level real-time power allocation command of each battery cell based on the power allocation benchmark weight, real-time SOH and SOC data, and real-time power command from the power grid, with SOH as the core decision variable. The edge control execution module is connected to the dual-layer power allocation optimization decision module, the energy storage power station PCS cluster, and the BMS system. It is used to receive and send real-time power allocation instructions to the corresponding PCS and BMS for execution, complete the precise control of charging and discharging power, and feed back the execution status to form a closed-loop control. The safety control and operation and maintenance management module is connected to the perception layer unit, the battery health status assessment and remaining life prediction module, and the edge control execution module, respectively. It is used to realize early warning of battery failure and graded emergency control based on SOH abnormal characteristics, and to generate preventive operation and maintenance plans based on RUL prediction results.

[0006] Preferably, the multi-dimensional SOH assessment system constructed by the battery health status assessment and remaining life prediction module includes capacity decay rate. DC internal resistance growth rate Peak power capability attenuation coefficient Cyclic aging factors Calendar aging factors Historical damage factors There are a total of 6 core characteristic parameters. The SOH comprehensive evaluation formula is as follows: ; Among them, the weighting coefficient to The determination was made by coupling the Analytic Hierarchy Process (AHP) with the entropy weight method. ; , This represents the battery's current maximum usable capacity. This refers to the battery's rated capacity. , This is the current DC internal resistance of the battery. This refers to the battery's factory rated internal resistance. , The current peak charge / discharge power of the battery. This refers to the battery's rated peak power.

[0007] Preferably, the electrochemical model and AI data-driven fusion algorithm is a fusion algorithm combining a second-order Thevenin electrochemical equivalent circuit model with an unscented Kalman filter (UKF) and an LSTM deep neural network. The specific implementation steps are as follows: S1. Construct a second-order Thevenin equivalent circuit model. The state equations are: ; The observation equation is: ; in, This is the battery open-circuit voltage. This refers to the battery terminal voltage. This represents the load current; charging is positive and discharging is negative. The internal resistance of the battery is in ohms. , For electrochemically polarized resistors and capacitors, , For concentration polarization resistors and capacitors, , Polarization voltage, The charge / discharge coulomb efficiency; S2. Use the unscented Kalman filter (UKF) to perform state estimation on the above model to obtain basic estimated values ​​of battery SOC, internal resistance parameters, and maximum usable capacity. S3. Construct an LSTM deep neural network model. The input features include the basic estimated value of UKF output, battery operating temperature, number of cycles, and historical charge / discharge rate data. The output is the SOH correction value. S4. The UKF basic estimation result is fused with the LSTM correction result to obtain the final cell-level SOH estimation value.

[0008] Preferably, the RUL prediction of the battery health status assessment and remaining life prediction module is implemented using the rainflow counting method (RCA) combined with the Gaussian process regression (GPR) algorithm. The specific steps are as follows: S1. Use the rainflow counting method to perform cyclic statistics on the historical charge and discharge current data of the battery, quantify the number of cycles under different depths of charge and discharge (DOD), and calculate the cycle aging damage: ; in, For the first The actual number of loops under this DOD. This represents the number of battery failure cycles under this DOD. S2. Calculate calendar aging damage: ; in, This refers to the actual service life of the battery. This refers to the battery's rated calendar life. S3, Calculate total aging damage ; S4. Using the Gaussian process regression (GPR) algorithm, based on historical SOH trends, total aging damage, and future power scheduling plans, the remaining cycle life and remaining calendar life (RUL) of the battery are predicted.

[0009] Preferably, the upper-level long-cycle lifetime optimization scheduling layer is implemented using the Model Predictive Control (MPC) algorithm, with a prediction time domain of 24 hours, a control time domain of 15 minutes, a rolling optimization step size of 15 minutes, and a bi-objective optimization function: ; in, The total cost of battery aging and damage for the entire station. For the daily operating revenue of the power station, This is a weighting coefficient, with a value ranging from 0.2 to 0.8, which can be adaptively adjusted according to the operating scenario; The formula for calculating the total cost of aging damage is as follows: ; in, For the number of battery clusters, For the first The purchase cost of a single battery cluster For the first The aging damage increment of each battery cluster in the control time domain; The optimization process satisfies four types of rigid constraints: power balance constraint, battery power upper and lower limit constraint, SOC safety constraint, and aging rate equalization constraint.

[0010] Preferably, the lower-level real-time power allocation control layer is implemented by combining improved adaptive droop control with multi-objective particle swarm optimization (MOPSO) algorithm, with an optimization time step of 100ms. The real-time multi-objective optimization function is: ; in, For power tracking error, To achieve real-time aging balance, For the goal of SOC balance, , , These are the weighting coefficients, and + + =1; The power tracking error: ; in For the first Real-time power allocation for each battery cluster Real-time power command for the power grid; The aging equilibrium target: ; The SOC balancing objective is: ; in For the first The real-time state of charge (SOC) of a battery cluster typically ranges from 0 to 1 (or 0% to 100%), representing the percentage of the cluster's current remaining charge relative to its rated total capacity. This represents the average SOC of the entire battery cluster at the station.

[0011] Preferably, the lower-level real-time power allocation control layer incorporates a hierarchical adaptive allocation mode with SOH-SOC dual-dimensional linkage, specifically as follows: Power sharing mode: When the SOH difference among all battery clusters is ≤5%, SOH is ≥80%, and SOC dispersion is ≤3%, the power sharing mode is adopted. ; SOH-weighted adaptive allocation mode: When the SOH difference of battery clusters is greater than 5%, a SOH-proportional weighted allocation is adopted, with power allocation weights... Real-time power allocation And fine-tune it based on the SOC status; Fault unit lockout mode: When the SOH of a battery cluster is less than 60% or an alarm for abnormal voltage, temperature or internal resistance occurs, the power output of that battery cluster is immediately locked out, and its power command is redistributed to other healthy battery clusters.

[0012] Preferably, the system adopts a four-layer collaborative architecture of cloud, edge, and terminal, including an edge terminal layer, a local control layer, a core decision platform layer, and a cloud collaborative management layer. The perception layer unit and the edge control execution module are deployed in the edge terminal layer and the local control layer. The battery health status assessment and remaining life prediction module, the dual-layer power allocation optimization decision module, and the safety management and operation and maintenance management module are deployed in the local control layer and the core decision platform layer. The cloud collaborative management layer is used for centralized monitoring of multiple energy storage power stations, big data analysis, algorithm model training and iterative optimization, and to realize cloud-based model optimization and edge-based updates.

[0013] Preferably, the security control and operation and maintenance management module constructs a three-level early warning and graded emergency control system, specifically as follows: Level 1 warning: Triggered when the battery SOH degradation rate exceeds 0.05% / day or the internal resistance growth rate exceeds 5% / month, and a warning message is pushed to the operation and maintenance personnel; Level 2 warning: Triggered when battery SOH < 70%, voltage dispersion exceeds 100mV, or temperature difference between batteries in the same cluster exceeds 5℃, automatically reducing the upper limit of charging and discharging power of the corresponding battery cluster and generating a maintenance work order; Level 3 warning: Triggered when battery SOH < 60% or when voltage change, temperature rise, or internal resistance jump occurs, the corresponding battery cluster is immediately locked, shutdown protection is executed, and the temperature control and fire protection systems are linked. The safety control and operation and maintenance management module also automatically generates preventive operation and maintenance plans based on RUL prediction results, realizing balanced battery maintenance, accurate fault location, closed-loop management of operation and maintenance work orders, battery tiered utilization assessment and retirement determination.

[0014] Preferably, the system further includes a digital twin simulation module, which is used to construct a 1:1 digital twin of the energy storage power station, mapping the real operating characteristics of the battery, PCS, and electrical system, simulating the battery aging trend and power station revenue under different power distribution strategies and different operating conditions, and realizing offline verification and iterative optimization of optimization strategies.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This system overturns traditional decision-making logic by using State of Health (SOH) as the core decision variable for power allocation. Through multi-dimensional, high-precision online SOH estimation, it adopts a weighted allocation strategy based on SOH proportionally, allowing high-healthy batteries to bear more cycle loads and aging batteries to operate at reduced rates. This achieves a balanced aging rate across the entire battery stack, which can extend battery cycle life and significantly reduce the total life cycle battery replacement cost.

[0016] 2. This system innovatively designs a two-layer collaborative architecture. The upper layer completes the daily global lifetime and revenue dual-objective optimization through the MPC algorithm and generates a weighted benchmark. The lower layer achieves millisecond-level power command solving through improved adaptive droop control and MOPSO algorithm, with a response time of ≤100ms, which not only meets the grid assessment requirements, but also achieves optimal revenue throughout the entire life cycle.

[0017] 3. This system constructs a three-level early warning system based on SOH anomaly characteristics, which can identify potential faults such as lithium deposition and internal micro-short circuits in advance; at the same time, it automatically generates preventive operation and maintenance plans based on RUL prediction results, realizing a closed loop of operation control, safety protection and intelligent operation and maintenance, significantly reducing the risk of safety accidents and improving the return on investment throughout the project life cycle. Attached Figure Description

[0018] Figure 1 This is a flowchart of a power allocation and management system for an energy storage power station based on battery health status, according to the present invention. Figure 2This is a schematic diagram of a two-layer power allocation optimization decision module architecture in a power allocation and management system for an energy storage power station based on battery health status, according to the present invention. Figure 3 This is a diagram of a power distribution and management system for an energy storage power station based on battery health status, according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example: Refer to Figures 1-3 As shown: A power distribution and management system for an energy storage power station based on battery health status, adopting a four-layer collaborative architecture of cloud, edge, and terminal, with modules connected via industrial Ethernet or 5G encrypted communication, specifically including: (i) The sensing layer unit is the data foundation of the system, including the cell-level slave control unit (BMU), the battery cluster master control unit (BCU), the power station-level measurement and control and sensing equipment, and the edge acquisition gateway.

[0021] The BMU collects real-time data on voltage, current, temperature, and internal resistance of individual cells at a frequency of 100Hz, and performs individual cell-level protection and active balancing. The BCU aggregates BMU data, collects cluster total voltage, total current, and insulation resistance parameters, and completes preliminary estimation and protection of cluster-level SOC / SOH. Power plant-level monitoring and control equipment collects PCS operating status, grid connection point voltage / frequency / power, power grid dispatch instructions, ambient temperature and humidity, and fire protection / temperature control system data; The edge acquisition gateway performs data filtering, noise reduction, timestamp alignment, and outlier removal. It uploads standardized data to the upper-layer module through encrypted communication. The data acquisition frequency covers the millisecond level (real-time control) to the second level (status monitoring).

[0022] (ii) Battery health status assessment and remaining life prediction module, deployed on local industrial server and edge computing unit, lightweight inference model deployed at the edge to achieve second-level SOH update, and complete training model deployed on local platform to achieve model iterative optimization.

[0023] (1) Multi-dimensional SOH assessment system: A SOH assessment system containing 6 core characteristic parameters is constructed. The comprehensive assessment formula is as follows: ; in, For capacity decay rate, For DC internal resistance growth rate, Peak power capability attenuation coefficient, For cyclic aging factors, For calendar aging factors, Historical damage factors, , This represents the battery's current maximum usable capacity. This refers to the battery's rated capacity. , This is the current DC internal resistance of the battery. This refers to the battery's factory rated internal resistance. , The current peak charge / discharge power of the battery. The battery's rated peak power, weighting factor to The determination was made by coupling the Analytic Hierarchy Process (AHP) with the entropy weight method. The specific steps are as follows: I. Construct a judgment matrix using the analytic hierarchy process (AHP) and calculate the subjective weights of each feature parameter. ; II. Calculate the objective weights of each characteristic parameter based on historical operating data using the entropy weight method. ; III. Calculate the coupling weights using the multiplicative synthesis normalization method: ; For example, for lithium iron phosphate batteries, the weights are determined as follows: =0.35, =0.25, =0.15, =0.1, =0.08, =0.07, which can be dynamically adjusted according to different battery types such as ternary lithium.

[0024] (2) The SOH estimation algorithm integrating electrochemical model and AI adopts a fusion algorithm combining second-order Thevenin electrochemical equivalent circuit model with unscented Kalman filter (UKF) and LSTM deep neural network. The specific steps are as follows: (1) Construct a second-order Thevenin equivalent circuit model, and the state equation and observation equation as described in claim 3, to accurately describe the dynamic electrical characteristics of the battery; (2) The UKF was used to perform state estimation on the model to obtain basic estimates of battery SOC, ohmic internal resistance, polarization parameters and maximum available capacity; (3) Construct a 3-layer LSTM deep neural network model with an input layer dimension of 8, a hidden layer neuron count of 64, and an output layer with SOH correction value; (4) The UKF basic estimation results are fused with the LSTM correction results to finally achieve a cell-level online SOH estimation error of ≤3%.

[0025] (3) RUL (Remaining Lifetime) prediction is achieved using the Rainflow Counting (RCA) method combined with the Gaussian Process Regression (GPR) algorithm. The RUL prediction of the pool health status assessment and remaining lifetime prediction module is achieved using the Rainflow Counting (RCA) method combined with the Gaussian Process Regression (GPR) algorithm. The specific steps are as follows: S1. Use the rainflow counting method to perform cyclic statistics on the historical charge and discharge current data of the battery, quantify the number of cycles under different depths of charge and discharge (DOD), and calculate the cycle aging damage: ; in, For the first The actual number of loops under this DOD. This represents the number of battery failure cycles under this DOD. S2. Calculate calendar aging damage: ; in, This refers to the actual service life of the battery. This refers to the battery's rated calendar life. S3, Calculate total aging damage ; S4. Using the Gaussian process regression (GPR) algorithm, based on historical SOH trends, total aging damage, and future power scheduling plans, the remaining cycle life and remaining calendar life (RUL) of the battery are predicted.

[0026] Quantify the aging damage of batteries throughout their entire life cycle, accurately predict the remaining cycle life and calendar life, and provide a quantitative basis for upper-level optimization scheduling.

[0027] (III) The two-layer power allocation optimization decision module is the core decision-making center of the system, and its architecture is as follows: Figure 2 As shown, it includes an upper long-cycle lifetime optimization scheduling layer (optimization time scale 15 minutes to 24 hours) and a lower real-time power distribution control layer (optimization time scale 10ms to 1s). The two layers work together to completely resolve the industry contradiction between global optimization and real-time response.

[0028] The upper-level long-cycle lifetime optimization scheduling layer is implemented using the Model Predictive Control (MPC) algorithm. Specifically, the prediction time domain is 24 hours, the control time domain is 15 minutes, the rolling optimization step size is 15 minutes, and the bi-objective optimization function is: ; in, The total cost of battery aging and damage for the entire station. For the daily operating revenue of the power station, This is a weighting coefficient, with a value ranging from 0.2 to 0.8, which can be adaptively adjusted according to the operating scenario; The formula for calculating the total cost of aging damage is: ; in, For the number of battery clusters, For the first The purchase cost of a single battery cluster For the first The aging damage increment of each battery cluster in the control time domain; The optimization process satisfies four types of rigid constraints: power balance constraint, battery power upper and lower limit constraint, SOC safety constraint, and aging rate equalization constraint.

[0029] Based on the intraday grid dispatch plan, peak-valley electricity prices, renewable energy output forecasts, and RUL (Recovery Limiting) forecasts, the power allocation benchmark weights for each battery cell are calculated to achieve optimal global lifetime and profitability. In this embodiment, the weight coefficient λ=0.6 in the frequency regulation scenario prioritizes lifetime balance; in the peak-valley arbitrage scenario, λ=0.4 prioritizes economic profitability.

[0030] The lower-level real-time power allocation control layer is implemented using a combination of improved adaptive droop control and multi-objective particle swarm optimization (MOPSO) algorithm, with an optimization time step of 100ms. The real-time multi-objective optimization function is as follows: ; in, For power tracking error, To achieve real-time aging balance, For the goal of SOC balance, , , These are the weighting coefficients, and + + =1; Power tracking error: ; in For the first Real-time power allocation for each battery cluster Real-time power command for the power grid; Aging balance goals: ; SOC equilibrium objective: ; in For the first The real-time state of charge (SOC) of a battery cluster typically ranges from 0 to 1 (or 0% to 100%), representing the percentage of the cluster's current remaining charge relative to its rated total capacity. This represents the average SOC of the entire battery cluster at the station.

[0031] Based on upper-level benchmark weights, real-time SOH / SOC data, and real-time power commands from the power grid, the system solves for millisecond-level real-time power allocation commands with a solution time ≤50ms and a response time ≤100ms, fully meeting the requirements for rapid grid response. Simultaneously, it incorporates a built-in hierarchical adaptive allocation mode with SOH-SOC dual-dimensional linkage, as described in claim 7, which can adaptively adapt to different battery health state scenarios, balancing aging equalization and SOC safety.

[0032] (iv) Edge control execution module, deployed in the local control room of the power plant, adopts dual-redundant industrial servers, including station-level EMS local controller, PCS cluster control unit, and local safety emergency control unit, supports mainstream industrial communication protocols such as IEC61850, Modbus, and IEC104, and the communication delay with PCS and BMS is ≤20ms.

[0033] The specific workflow is as follows: Receive real-time power allocation instructions from the dual-layer power allocation optimization decision module, break them down into active / reactive power control instructions for each PCS, and send them to the corresponding PCS converters to precisely control the charging and discharging parameters of each battery cluster; at the same time, collect the execution status data of the PCS and BMS in real time and feed them back to the decision module to form a closed-loop control; when a grid fault or equipment abnormality occurs, immediately trigger graded protection to prioritize the safety of the power station.

[0034] The safety control and operation and maintenance management module constructs a three-level early warning and graded emergency control system. The first-level early warning is triggered when the battery SOH decay rate exceeds 0.05% / day or the internal resistance growth rate exceeds 5% / month, and pushes the early warning information to the operation and maintenance personnel. Level 2 warning: Triggered when battery SOH < 70%, voltage dispersion exceeds 100mV, or temperature difference between batteries in the same cluster exceeds 5℃, automatically reducing the upper limit of charging and discharging power of the corresponding battery cluster and generating a maintenance work order; Level 3 warning: Triggered when battery SOH < 60% or when voltage change, temperature rise, or internal resistance jump occurs, the corresponding battery cluster is immediately locked, shutdown protection is executed, and the temperature control and fire protection systems are linked. (v) The safety control and operation and maintenance management module also automatically generates preventive operation and maintenance plans based on RUL prediction results, realizing balanced battery maintenance, accurate fault location, closed-loop management of operation and maintenance work orders, battery cascade utilization assessment and retirement determination.

[0035] (vi) Based on the characteristics of abnormal SOH decay, sudden change in internal resistance, and capacity drop, potential faults such as lithium deposition and internal micro short circuits in the battery can be identified in advance, with an early warning accuracy of ≥95%, realizing full-process emergency control from power derating and module removal to linkage fire protection.

[0036] Meanwhile, based on RUL prediction results, preventive maintenance plans are automatically generated, including battery equalization maintenance, internal resistance detection, capacity calibration, and spare parts scheduling, to achieve closed-loop management of maintenance work orders; it also supports battery cascade utilization assessment and retirement determination, completing the closed-loop management of the entire battery life cycle.

[0037] (vi) Cloud-based collaborative management layer and digital twin simulation module: The cloud-based collaborative management layer is deployed on the industry's private cloud, supporting centralized monitoring and collaborative scheduling of dozens of energy storage power stations; The digital twin simulation module constructs a 1:1 digital twin of the energy storage power station, simulating battery aging trends and power station revenue under different strategies, and realizing offline verification of optimization strategies; Through the historical big data of the entire station, the algorithm model is continuously trained and optimized, and then sent to the edge for updates, realizing continuous iterative improvement of system performance.

[0038] This system uses battery state of health (SOH) as the core decision variable and adopts a four-layer collaborative architecture of cloud, edge, and device to build a dual closed-loop management and control system for the entire life cycle of "perception-assessment-decision-execution-feedback-iteration". This system addresses the industry pain point that traditional energy storage systems focus only on SOC balancing and ignore differences in battery aging when allocating power.

[0039] The system first collects real-time operational data from all battery cells across the entire station, including voltage, current, temperature, and internal resistance, through the cell-level acquisition terminal at the sensing layer. After preprocessing, the data is synchronously uploaded to the core evaluation module. Employing a fusion algorithm of "electrochemical equivalent circuit model + AI data-driven" approach, the system constructs a multi-dimensional SOH evaluation system covering capacity decay, internal resistance growth, and aging damage. This achieves an online SOH estimation error of ≤3% at the cell level and simultaneously predicts the remaining battery life (RUL), transforming SOH from an operational reference indicator into a core variable for real-time control.

[0040] The core decision-making layer adopts a two-layer collaborative architecture: the upper layer, based on the scheduling plan, electricity price signal and RUL result, takes minimizing the life-cycle degradation and maximizing the power plant revenue as the dual objectives, and generates the power allocation benchmark weight through the MPC algorithm; the lower layer combines real-time SOH and SOC data, and solves the power command of each battery cell in milliseconds through improved adaptive droop control and multi-objective optimization algorithm, and achieves the equalization of the aging rate of the entire stack of batteries through hierarchical adaptive mode.

[0041] Ultimately, the edge side executes commands and provides real-time feedback on operational data, forming a millisecond-level real-time control loop; the cloud uses big data to iteratively optimize the algorithm model, combining it with intelligent operation and maintenance data to form a long-term optimization loop, and simultaneously embeds a three-level safety early warning system to achieve synergistic optimization of power plant safety, lifespan, and revenue.

[0042] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power distribution and management system for an energy storage power station based on battery health status, characterized in that, It includes a perception layer unit, a battery health status assessment and remaining life prediction module, a two-layer power allocation optimization decision module, an edge control execution module, and a safety management and operation and maintenance management module; The sensing layer unit is used to collect full-dimensional operational data from the battery cells to the entire station level of the energy storage power station, and upload the data to the corresponding module after preprocessing. The battery health status assessment and remaining life prediction module is communicatively connected to the sensing layer unit. It is used to construct a multi-dimensional battery health status SOH assessment system based on preprocessed operating data, realize real-time online SOH estimation at the cell level through electrochemical model and AI data-driven fusion algorithm, and complete the prediction of the remaining battery life RUL based on the SOH change trend. The dual-layer power allocation optimization decision module is communicatively connected to the battery health status assessment and remaining life prediction module and the power grid dispatch system, respectively. It includes an upper long-cycle life optimization scheduling layer and a lower real-time power allocation control layer. The upper long-cycle life optimization scheduling layer is used to solve for the power allocation benchmark weight of each battery cell based on the intraday power grid dispatch plan, electricity price signal, new energy output prediction and RUL prediction results of each battery cell, with the dual objectives of minimizing the life decay of the entire life cycle and maximizing the power plant revenue. The lower-level real-time power allocation control layer is used to solve for the millisecond-level real-time power allocation command of each battery cell based on the power allocation benchmark weight, real-time SOH and SOC data, and real-time power command from the power grid, with SOH as the core decision variable. The edge control execution module is connected to the dual-layer power allocation optimization decision module, the energy storage power station PCS cluster, and the BMS system. It is used to receive and send real-time power allocation instructions to the corresponding PCS and BMS for execution, complete the precise control of charging and discharging power, and feed back the execution status to form a closed-loop control. The safety control and operation and maintenance management module is connected to the perception layer unit, the battery health status assessment and remaining life prediction module, and the edge control execution module, respectively. It is used to realize early warning of battery failure and graded emergency control based on SOH abnormal characteristics, and to generate preventive operation and maintenance plans based on RUL prediction results.

2. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The battery health status assessment and remaining life prediction module constructs a multi-dimensional SOH assessment system, which includes capacity decay rate. DC internal resistance growth rate Peak power capability attenuation coefficient Cyclic aging factors Calendar aging factors Historical damage factors There are a total of 6 core characteristic parameters. The SOH comprehensive evaluation formula is as follows: ； Among them, the weighting coefficient to The determination was made by coupling the Analytic Hierarchy Process (AHP) with the entropy weight method. ; , This represents the battery's current maximum usable capacity. This refers to the battery's rated capacity. , This is the current DC internal resistance of the battery. This refers to the battery's factory rated internal resistance. , The current peak charge / discharge power of the battery. This refers to the battery's rated peak power.

3. The power distribution and management system for an energy storage power station based on battery health status according to claim 2, characterized in that, The electrochemical model and AI data-driven fusion algorithm is a fusion algorithm combining a second-order Thevenin electrochemical equivalent circuit model with an unscented Kalman filter (UKF) and an LSTM deep neural network. The specific implementation steps are as follows: S1. Construct a second-order Thevenin equivalent circuit model. The state equations are: ； The observation equation is: ； in, This is the battery open-circuit voltage. This refers to the battery terminal voltage. This represents the load current; charging is positive and discharging is negative. The internal resistance of the battery is in ohms. , For electrochemically polarized resistors and capacitors, , For concentration polarization resistors and capacitors, , Polarization voltage, The charge / discharge coulomb efficiency; S2. Use the unscented Kalman filter (UKF) to perform state estimation on the above model to obtain basic estimated values ​​of battery SOC, internal resistance parameters, and maximum usable capacity. S3. Construct an LSTM deep neural network model. The input features include the basic estimated value of UKF output, battery operating temperature, number of cycles, and historical charge / discharge rate data. The output is the SOH correction value. S4. The UKF basic estimation result is fused with the LSTM correction result to obtain the final cell-level SOH estimation value.

4. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The RUL prediction of the battery health status assessment and remaining life prediction module is implemented using the rainflow counting method (RCA) combined with the Gaussian process regression (GPR) algorithm. The specific steps are as follows: S1. Use the rainflow counting method to perform cyclic statistics on the historical charge and discharge current data of the battery, quantify the number of cycles under different depths of charge and discharge (DOD), and calculate the cycle aging damage: ； in, For the first The actual number of loops under this DOD. This represents the number of battery failure cycles under this DOD. S2. Calculate calendar aging damage: ； in, This refers to the actual service life of the battery. This refers to the battery's rated calendar life. S3, Calculate total aging damage ; S4. Using the Gaussian process regression (GPR) algorithm, based on historical SOH trends, total aging damage, and future power scheduling plans, the remaining cycle life and remaining calendar life (RUL) of the battery are predicted.

5. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The upper-level long-cycle lifetime optimization scheduling layer is implemented using the Model Predictive Control (MPC) algorithm, with a prediction time domain of 24 hours, a control time domain of 15 minutes, a rolling optimization step size of 15 minutes, and a bi-objective optimization function: ； in, The total cost of battery aging and damage for the entire station. For the daily operating revenue of the power station, This is a weighting coefficient, with a value ranging from 0.2 to 0.8, which can be adaptively adjusted according to the operating scenario; The formula for calculating the total cost of aging damage is as follows: ； in, For the number of battery clusters, For the first The purchase cost of a single battery cluster For the first The aging damage increment of each battery cluster in the control time domain; The optimization process satisfies four types of rigid constraints: power balance constraint, battery power upper and lower limit constraint, SOC safety constraint, and aging rate equalization constraint.

6. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The lower-level real-time power allocation control layer is implemented by combining improved adaptive droop control with multi-objective particle swarm optimization (MOPSO) algorithm, with an optimization time step of 100ms. The real-time multi-objective optimization function is: ； in, For power tracking error, To achieve real-time aging balance, For the goal of SOC balance, , , These are the weighting coefficients, and + + =1; The power tracking error: ； in For the first Real-time power allocation for each battery cluster Real-time power command for the power grid; The aging equilibrium target: ； The SOC balancing objective is: ； in For the first The real-time state of charge (SOC) of a battery cluster typically ranges from 0 to 1 (or 0% to 100%), representing the percentage of the cluster's current remaining charge relative to its rated total capacity. This represents the average SOC of the entire battery cluster at the station.

7. The power distribution and management system for an energy storage power station based on battery health status according to claim 6, characterized in that, The lower-level real-time power allocation control layer incorporates a hierarchical adaptive allocation mode with SOH-SOC dual-dimensional linkage, specifically as follows: Power sharing mode: When the SOH difference among all battery clusters is ≤5%, SOH is ≥80%, and SOC dispersion is ≤3%, the power sharing mode is adopted. ; SOH-weighted adaptive allocation mode: When the SOH difference of battery clusters is greater than 5%, a SOH-proportional weighted allocation is adopted, with power allocation weights... Real-time power allocation And fine-tune it based on the SOC status; Fault cell lockout mode: When the SOH of a battery cluster is less than 60% or an alarm for abnormal voltage, temperature or internal resistance occurs, the power output of that battery cluster is immediately locked out and its power command is redistributed to other healthy battery clusters.

8. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The system adopts a four-layer collaborative architecture of cloud, edge, and terminal, including an edge terminal layer, a local control layer, a core decision platform layer, and a cloud-based collaborative management layer. The perception layer unit and the edge control execution module are deployed in the edge terminal layer and the local control layer. The battery health status assessment and remaining life prediction module, the dual-layer power allocation optimization decision module, and the safety management and operation and maintenance management module are deployed in the local control layer and the core decision platform layer. The cloud-based collaborative management layer is used for centralized monitoring of multiple energy storage power stations, big data analysis, algorithm model training and iterative optimization, and to realize cloud-based model optimization and edge-based updates.

9. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The security control and operation and maintenance management module constructs a three-level early warning and graded emergency control system, specifically as follows: Level 1 warning: Triggered when the battery SOH degradation rate exceeds 0.05% / day or the internal resistance growth rate exceeds 5% / month, and a warning message is pushed to the operation and maintenance personnel; Level 2 warning: Triggered when battery SOH < 70%, voltage dispersion exceeds 100mV, or temperature difference between batteries in the same cluster exceeds 5℃, automatically reducing the upper limit of charging and discharging power of the corresponding battery cluster and generating a maintenance work order; Level 3 warning: Triggered when battery SOH < 60% or when voltage change, temperature rise, or internal resistance jump occurs, the corresponding battery cluster is immediately locked, shutdown protection is executed, and the temperature control and fire protection systems are linked. The safety control and operation and maintenance management module also automatically generates preventive operation and maintenance plans based on RUL prediction results, realizing balanced battery maintenance, accurate fault location, closed-loop management of operation and maintenance work orders, battery tiered utilization assessment and retirement determination.

10. The power distribution and management system for an energy storage power station based on battery health status according to claim 1, characterized in that, The system also includes a digital twin simulation module, which is used to construct a 1:1 digital twin of the energy storage power station, mapping the real operating characteristics of the battery, PCS, and electrical system, simulating the battery aging trend and power station revenue under different power distribution strategies and different operating conditions, and realizing offline verification and iterative optimization of optimization strategies.

Images