Long and short term dynamic management and control method and system of modules in a battery energy storage station

Abstract

The long and short term dynamic management and control method and system of the module in the battery energy storage station will automatically model the power generation control as a leader agent and model a battery module as a follower agent. The leader agent determines the aging state parameters of each follower agent in the day-ahead and intra-day stages in the dispatching period. Under the whole link constraint, the day-ahead scheduling scheme is solved based on the robust Pareto optimization objective. Based on the day-ahead scheduling scheme, different model predictive control strategies are adopted according to the different time scales of the intra-day scheduling instruction interval, and the intra-day scheduling scheme is solved based on the MPC optimization objective. The day-ahead scheduling scheme and the intra-day scheduling scheme constitute the state matrix of the leader agent, each follower agent calculates the local control instruction through the distributed fixed time sliding mode controller, and each battery module executes the operation according to the local control instruction. The adaptive source and load uncertainty, the optimal balance between calculation efficiency and control accuracy are solved.

Description

Technical Field

[0001] This invention belongs to the field of energy storage system scheduling, specifically relating to a method and system for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station. Background Technology

[0002] Reclaiming retired power batteries and reusing them in the energy storage field can reduce the construction cost of energy storage systems and achieve carbon emission reduction throughout their entire life cycle. However, the core challenge of repurposed battery energy storage stations lies in the deep heterogeneity and inconsistent state of the battery modules. Retired batteries come from diverse sources, and their historical operating conditions, remaining lifespan, internal resistance characteristics, and capacity decay trajectories vary greatly.

[0003] Existing technologies for managing heterogeneous energy storage stations generally suffer from the following limitations: The control model is too coarse, lacking refined module-level management. Current energy storage management systems typically treat the entire station as a uniform "black box" aggregation model, ignoring the parameter dispersion among battery modules within the station. Adopting a uniform charging and discharging strategy forces poorly functioning modules to overcharge or over-discharge, accelerating their aging process and limiting the station's available capacity and charging / discharging power due to the "bottleneck effect," potentially even inducing thermal runaway safety risks. Long-cycle scheduling strategies lack robustness to uncertainty. Traditional day-ahead scheduling is largely based on deterministic load and renewable energy output forecasts for optimization; however, power fluctuations on both the source and load sides are highly uncertain in actual operation. Existing deterministic optimization methods often fail when faced with prediction errors, causing the energy storage system to frequently deviate from optimal operating conditions, making it difficult to balance economy and system safety. Short-cycle real-time control struggles to balance computational efficiency and optimization accuracy. To meet the demands of second-level ancillary services such as grid frequency regulation, control systems require extremely high response speeds. However, real-time solutions using mixed-integer programming, which incorporates refined aging models (such as rainflow counting and electrochemical mechanism models), incur extremely heavy computational loads, making it difficult to meet the timeliness requirements of online real-time control. Conversely, simplifying the model with linear programming fails to accurately quantify the nonlinear degradation costs of secondary batteries under complex operating conditions, resulting in control strategies that achieve only suboptimal rather than optimal outcomes. Therefore, developing a management method that adapts to the high heterogeneity of secondary batteries, balances long-cycle robust programming with short-cycle real-time precise response, and achieves the optimal balance between computational efficiency and control accuracy is a critical bottleneck that urgently needs to be overcome in the current energy storage technology field. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for long-term and short-term dynamic management and control of modules within a cascaded battery energy storage station. It deeply integrates a multi-agent collaborative architecture, adaptive robust optimization, and event-triggered model predictive control to achieve refined proactive management and control of the state of modules within the cascaded battery energy storage station across multiple time scales. This solves the problem of dynamic scheduling throughout the entire lifecycle, which adapts to source-load uncertainty and achieves the optimal balance between computational efficiency and control accuracy.

[0005] The present invention adopts the following technical solution.

[0006] This invention proposes a method for long-term and short-term dynamic management and control of modules in a tiered battery energy storage station, comprising: Automatic power generation control is modeled as a leader agent, and a battery module is modeled as a follower agent; the leader agent acquires the day-ahead forecast data of the power grid and the real-time operating status parameters of all battery modules; The leader agent determines the aging state parameters of each follower agent in the day-ahead and intraday phases within the scheduling cycle based on the real-time operating state parameters of all battery modules. The leader agent establishes full-link constraints based on the electrical topology of the cascaded battery energy storage station, including: state constraints of battery modules, power constraints of DC / DC converters, power constraints of DC / AC inverters, and power constraints of transformers. Based on the actual scheduling data and day-ahead grid forecast data from the previous scheduling cycle, the leader agent solves for the day-ahead scheduling scheme for each battery module in the current scheduling cycle using a robust Pareto optimization objective under full-link constraints. Based on the day-ahead scheduling scheme, the leader agent employs different model predictive control strategies according to different time scales of intraday scheduling command intervals, and solves for the intraday scheduling scheme for each battery module in the current scheduling cycle using an MPC optimization objective. The leader agent's state matrix is ​​composed of the day-ahead scheduling scheme and the intraday scheduling scheme. Each follower agent calculates local control commands through a distributed fixed-time sliding mode controller based on the leader agent's state matrix, its own state information, and the state information of neighboring follower agents. Each battery module performs power adjustment operations according to the corresponding local control commands.

[0007] Preferably, the aging state parameters of each follower agent in the day-ahead phase include: cumulative calendar aging increment and cumulative cyclic aging degree; the aging state parameters of each follower agent in the day-intraday phase include: cumulative cyclic aging sub-gradient.

[0008] Preferably, during the day-ahead phase of the current scheduling cycle, based on the SOC and operating temperature of all battery modules at the end of the previous scheduling cycle, the cumulative calendar aging increment of the battery modules is determined using the capacity decay of the battery modules caused by calendar aging based on a semi-empirical exponential model and an incremental integration strategy. During the day-ahead phase of the current scheduling cycle, the cumulative cycle aging degree of the battery module is determined by utilizing the battery cycle aging state based on power time-domain integration and the rainflow counting strategy. During the intraday phase of the current scheduling cycle, based on the sub-gradient strategy, the cumulative cyclic aging sub-gradient of the battery module is determined, including: the cumulative cyclic aging sub-gradient of the battery module under the condition of no switching between charge and discharge states, and the cumulative cyclic aging sub-gradient of the battery module under the condition of switching between charge and discharge states.

[0009] Preferably, during the day-ahead phase of the current scheduling cycle, after determining the health status of the battery module using the cumulative calendar aging increment and cumulative cycle aging degree, the actual available capacity of the battery module in the current scheduling cycle is determined in conjunction with the nominal rated capacity of the battery module.

[0010] Preferably, the discharge cycle depth of the battery module is updated based on the actual available capacity and power of the battery module in the current scheduling cycle; the cumulative cycle aging sub-gradient of the battery module under the condition of no switching between charge and discharge states is calculated using the power gradient of the battery module based on the discharge cycle depth, and a linear expression of the cumulative cycle aging sub-gradient of the battery module is obtained by linearizing the power gradient.

[0011] Preferably, within the constraints of the cumulative calendar aging increment and cumulative cycle aging degree of the battery module, the actual available capacity of the battery module during the scheduling cycle, the full-link constraints, and the constraints of the day-ahead grid-side power fluctuation set, a robust Pareto optimization objective is established, as shown in the following formula:

[0012] In the formula, For deterministic decision variables, For deterministic decision space, For grid-side power, The day-ahead grid-side power fluctuation set is constructed based on the actual dispatch data and the day-ahead forecast data of the power grid in the previous dispatch cycle. For the uncertain set of power The uncertainty error extracted from it, To optimize the objective function.

[0013] Preferably, an optimization objective function is established based on command power deviation penalty, battery module aging penalty, inter-module SOC deviation penalty, energy storage station power loss penalty, and inter-module circulating current penalty. As shown in the following formula: = + + + +

[0014] In the formula, , , , , These are the Pareto undetermined coefficients for the following penalties: command power deviation penalty, battery module day-ahead aging penalty, inter-module SOC deviation penalty, energy storage station power loss penalty, and inter-module circulating current penalty. For cost electricity price, , , , , These are, respectively, power command deviation penalty, battery module day-ahead aging penalty, battery module SOC deviation penalty, energy storage power station power loss penalty, and battery module circulating current penalty.

[0015] Preferably, the command power deviation penalty is as shown in the following formula: =

[0016] In the formula, For the predicted time period The current grid-side power, For the time period Total power of the low-voltage side of the transformer in the energy storage power station = , Battery cabinet The transformer during the time period Input power, For scheduling duration, For the scheduling period, This refers to the number of battery cabinets.

[0017] Preferably, the daily aging penalty of the battery module is as shown in the following formula: =

[0018] In the formula, These are the weighting coefficients. The current scheduling cycle under the day-ahead scheduling. Total aging cost of energy storage power stations For scheduling duration, The scheduling period; Among them, the total aging cost of energy storage power stations under the current dispatch is: =

[0019] In the formula, Battery cabinet Battery module in In the previous scheduling cycle -1 calendar aging cost, Battery cabinet Battery module in During the period The current cycle aging cost, This refers to the number of battery cabinets. This refers to the number of battery modules; Under the current scheduling, considering only the cumulative cycle aging gradient under the condition that the charge / discharge state of the battery module has not switched, the current cycle aging cost is as follows: = + +

[0020] In the formula, Weighted by the degree of historical degradation. Battery cabinet Battery module in During the scheduling period The actual available capacity The construction cost allocated to each battery module per cycle. =1 indicates a lithium iron phosphate battery module. =2 indicates a ternary lithium battery module. Battery cabinet Battery module in In the previous scheduling cycle -1 represents the cumulative aging level of the battery module. , These are the half-cycle coefficients under discharge and charge conditions, respectively. , Based on the charging depth in the day-ahead scheduling and depth of discharge The linear expression for the cumulative cyclic aging subgradient under the condition where the charge and discharge states are not switched.

[0021] Preferably, the SOC deviation penalty between battery modules is as shown in the following formula: =

[0022] The inconsistency in SOC between battery modules within an energy storage power station is as follows: = -

[0023] In the formula: For the time period Differences in lumped SOC of energy storage power stations For the time period Average SOC of energy storage power station For scheduling duration, For the scheduling period, This refers to the number of battery cabinets. For the number of battery modules, Battery cabinet Battery module in During the period The current stage of SOC.

[0024] Preferably, the power loss penalty of the energy storage power station is as shown in the following formula: =

[0025] The total lumped losses of converters and transformers in an energy storage power station are: = +

[0026] In the formula, For the time period Total losses of energy storage power stations Battery cabinet The transformer during the time period Lumped power loss, Battery cabinet DC / AC inverters during time periods Power loss, For battery module DC / DC converter in time period Power loss, For scheduling duration, For the scheduling period, This refers to the number of battery cabinets. This represents the number of battery modules.

[0027] Preferably, the circulating current penalty between battery modules is as follows: =

[0028] Auxiliary variables The following is used to quantify the circulating current intensity between battery modules: = { , } In the formula, For the time period Auxiliary variables, , Battery cabinet Battery module in DC / DC converter in time period The output power and input power, For scheduling duration, For the scheduling period, This refers to the number of battery cabinets. This represents the number of battery modules.

[0029] Preferably, when the time scale of the intraday scheduling instruction interval is less than 5 minutes, a linear model predictive control strategy is adopted to establish an MPC model based on quadratic programming. In the MPC model based on quadratic programming, the charging and discharging states of each battery module in the daytime scheduling scheme are adopted, and the power of each battery module at the intraday scheduling time is used as the control input to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of less than 5 minutes are obtained. When the time scale of the intraday scheduling instruction interval is not less than 5 minutes, a mixed integer model predictive control strategy is adopted, and an MPC model based on mixed integer quadratic programming is established. In the MPC model based on mixed integer quadratic programming, the power and charge / discharge status of each battery module at the intraday scheduling time are used as control inputs to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of not less than 5 minutes are obtained. Intraday scheduling instructions with a time interval of less than 5 minutes and intraday scheduling instructions with a time interval of not less than 5 minutes constitute an intraday scheduling scheme.

[0030] Preferably, within the constraints imposed by the cumulative cyclic aging subgradient of the battery module, the full-link constraints, and the charge state space of the battery module, the MPC cost function is used. Minimize the MPC optimization objective, and the MPC cost function is shown in the following equation: = + + + + + +

[0031] In the formula, , , , , These are the Pareto undetermined coefficients for command power deviation penalty, battery module intraday aging penalty, battery module SOC deviation penalty, energy storage station power loss penalty, and battery module circulating current penalty, respectively. The weight for penalizing deviations in energy storage command response. The weighting of the battery module SOC deviation between the intraday phase and the previous day phase; The cost price of electricity; , , , , , , These are, respectively, power command deviation penalty, energy storage response command deviation penalty, battery module intraday aging penalty, battery module SOC deviation penalty, intraday and day-ahead battery module SOC deviation penalty, energy storage power station power loss penalty, and battery module circulating current penalty.

[0032] Preferably, the penalty for deviation in energy storage response command is as follows: = +

[0033] In the formula, , They are respectively in the time period Reverse power deviation and forward power deviation of energy storage power station in response to ancillary service commands.

[0034] Preferably, the penalty for the deviation of the battery module's SOC between the intraday and the previous day stage is as follows: =

[0035] In the formula, , Battery cabinet Battery module in During the period SOC for the day-end and intraday phases, This refers to the number of battery cabinets. This represents the number of battery modules.

[0036] Preferably, the daily aging penalty for the battery module is as shown in the following formula: =

[0037] In the formula, These are the weighting coefficients. For intraday scheduling, the current scheduling cycle Total aging cost of energy storage power stations For scheduling duration, The scheduling period; Among them, the total aging cost of energy storage power stations under intraday dispatch is: =

[0038] In the formula, Battery cabinet Battery module in In the previous scheduling cycle -1 calendar aging cost, Battery cabinet Battery module in During the period The daily cycle aging cost, This refers to the number of battery cabinets. This refers to the number of battery modules; Under intraday scheduling, the cumulative cycle aging sub-gradient is selected based on the battery module's operating condition between the state of charge / discharge not switching and the operating condition at the moment of switching. The intraday cycle aging cost is shown in the following formula: = + +

[0039] In the formula, Weighted by the degree of historical degradation. Battery cabinet Battery module in During the scheduling period The actual available capacity The construction cost allocated to each battery module per cycle. =1 indicates a lithium iron phosphate battery module. =2 indicates a ternary lithium battery module. Battery cabinet Battery module in In the previous scheduling cycle -1 represents the cumulative aging level of the battery module. , These are the half-cycle coefficients under discharge and charge conditions, respectively. , Based on the charging depth during intraday scheduling and depth of discharge The linear expression for the cumulative cyclic aging subgradient.

[0040] This invention also proposes a long-term and short-term dynamic management and control system for modules in a cascaded battery energy storage station, comprising: The automatic power generation control is modeled as a leader agent, and a battery module is modeled as a follower agent. The leader agent is used to acquire day-ahead forecast data of the power grid and real-time operating status parameters of all battery modules. Based on the real-time operating status parameters of all battery modules, the aging status parameters of each follower agent in the day-ahead and intraday phases within the scheduling cycle are determined. Based on the electrical topology of the cascaded battery energy storage station, full-link constraints are established, including: state constraints of battery modules, power constraints of DC / DC converters, power constraints of DC / AC inverters, and power constraints of transformers. Based on the actual scheduling data and day-ahead forecast data of the power grid in the previous scheduling cycle, under full-link constraints, the day-ahead scheduling scheme of each battery module in the current scheduling cycle is solved based on the robust Pareto optimization objective. Based on the day-ahead scheduling scheme, different model predictive control strategies are adopted according to different time scales of intraday scheduling command intervals, and the intraday scheduling scheme of each battery module in the current scheduling cycle is solved based on the MPC optimization objective. The leader agent's state matrix is ​​composed of the day-ahead scheduling scheme and the intraday scheduling scheme. Each follower agent calculates local control commands through a distributed fixed-time sliding mode controller based on the leader agent's state matrix, its own state information, and the state information of neighboring follower agents. Each battery module performs power adjustment operations according to the corresponding local control commands.

[0041] The present invention is also a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to perform operations according to the instructions to execute the steps of the method.

[0042] The present invention is also a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0043] The beneficial effects of this invention are as follows, compared with the prior art, at least including: the method proposed in this invention constructs a hierarchical distributed multi-agent system, mapping physically heterogeneous cascaded battery modules into intelligent agent nodes with autonomous perception and decision-making capabilities, breaking the limitations of the traditional "black box" aggregation model; at the day-ahead long-period scale, this invention introduces a data-driven adaptive uncertainty set construction mechanism, and formulates a global benchmark that takes into account aging equilibrium and economic benefits through multi-objective Pareto optimization; at the intraday short-period scale, a model prediction rolling optimization strategy based on an event-triggered mechanism is designed, introducing transient marginal aging costs and dynamically switching the solution algorithm to formulate the optimal tracking trajectory for grid ancillary service commands; at the transient response scale, a distributed energy storage module power tracking strategy based on a fixed-time sliding mode controller is designed to achieve millisecond-level accurate response to grid ancillary service commands. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for long-term and short-term dynamic management and control of modules in a battery energy storage station based on a multi-agent system, as proposed in this invention. Figure 2 This is the modular battery energy storage system topology in this invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0046] To adapt to the high heterogeneity of batteries used in cascaded applications, balance robust long-cycle planning with accurate real-time short-cycle response, and achieve the optimal balance between computational efficiency and control precision, this invention proposes a long- and short-term dynamic management method for modules within a cascaded battery energy storage station. Figure 1 As shown, the method includes the following steps: Step S10: Based on the battery module topology, the automatic power generation control is modeled as a leader agent and a battery module is modeled as a follower agent to construct a multi-agent system.

[0047] This invention first constructs a hierarchical distributed multi-agent system, mapping physically heterogeneous cascaded battery modules into intelligent agent nodes with autonomous perception and decision-making capabilities, breaking the limitations of the traditional "black box" aggregation model.

[0048] Specifically, such as Figure 2 As shown, for a DC-side parallel-type modular cascaded battery energy storage station, a battery cabinet contains N heterogeneous battery modules with different battery states, such as health state and state of charge. The output of each battery module is regulated by a DC / DC converter and then connected in parallel to the common DC bus in the battery cabinet. A cascaded battery energy storage station contains M battery cabinets. One or more battery cabinets are connected in series / parallel and then a DC / AC inverter converts the DC power into low-voltage three-phase AC power. The low-voltage three-phase AC power output from the inverter is stepped up by a transformer and then connected to the main grid. In the above topology, each battery module is modeled as a follower agent. When the energy storage station is connected to the grid, the working state of the energy storage station is issued by Automatic Generation Control (AGC) commands. Therefore, the AGC commands are modeled as leader agents, and the changes in the AGC command state are modeled as exogenous systems. This constructs a multi-agent system in a "leader-follower" model.

[0049] This invention proposes a hierarchical distributed multi-agent system for intelligent collaborative control of cascaded battery modules in automatic power generation control scenarios. Each module acts as a follower agent, with its internal state (SOC, SOH, etc.) being externally visible or obtainable through communication. Follower agents can autonomously adjust their response strategies based on their own state. The failure of a single agent does not affect the overall system operation. Furthermore, each agent, while tracking the global objective, also considers multiple objectives such as inter-module balance, lifetime loss optimization, and efficiency maximization. This surpasses the traditional "black box" aggregation model and fully reflects the heterogeneity and autonomy of each battery module.

[0050] Step S20: The leader agent acquires the day-ahead forecast data of the power grid and the real-time operating status parameters of all battery modules.

[0051] Under the centralized control framework, steps S20 to S60 are executed by the leader agent.

[0052] Specifically, through an external data communication interface, the system acquires the grid-side day-ahead power forecast data sequence and grid time-of-use electricity price curve for the next scheduling cycle (24 hours). Simultaneously, through real-time communication with the underlying Battery Management System (BMS), it collects real-time operating status parameters of all battery modules within the cascaded battery energy storage power station, including but not limited to: battery cabinets. Battery module in During the period State of charge Health status Rated capacity Current scheduling cycle Actual available capacity Current scheduling cycle Operating temperature Cumulative deployment time , , The heterogeneous state data collected above will be mapped as the initial boundary conditions and state constraint benchmarks for the subsequent multi-timescale robust optimization scheduling model.

[0053] Step S30: The leader agent determines the aging state parameters of each follower agent in the day-ahead and intraday phases within the scheduling cycle based on the real-time operating status parameters of all battery modules.

[0054] Among them, the aging state parameters of each follower agent in the day-ahead phase include: cumulative calendar aging increment and cumulative cyclic aging degree; the aging state parameters of each follower agent in the intraday phase include: cumulative cyclic aging sub-gradient.

[0055] Specifically, step S30 includes: Step S301, in the current scheduling cycle In the day-ahead phase, based on the SOC and operating temperature of all battery modules at the end of the previous scheduling cycle, the cumulative calendar aging increment of the battery modules is determined using the capacity decay of the battery modules caused by calendar aging based on a semi-empirical exponential model and an incremental integration strategy.

[0056] Previous scheduling cycle At the end of -1, record the SOC and operating temperature of all battery modules; and expand the complete SOC sequence of all battery modules in all scheduling cycles.

[0057] The capacity degradation of the battery module caused by calendar aging is quantitatively described by a semi-empirical exponential model based on the Arrhenius law, as shown in the following formula: = × × (1) In the formula, For the time period The amount of battery module capacity degradation caused by calendar aging. This represents the mapping relationship between the state of charge of the battery module and calendar aging. Mapping relationship between battery module deployment time and calendar aging. Let be the ideal gas constant. The operating temperature of the battery module. To calculate the cumulative deployment time, This is the calendar aging time coefficient; In particular, for lithium iron phosphate battery modules and ternary lithium battery modules, the capacity degradation caused by calendar aging takes the following specific forms: = × × × (2) = × × (3) In the formula, , They are respectively in the time period Capacity decay of lithium iron phosphate battery modules and ternary lithium battery modules due to calendar aging , , , , , , , All are fitting coefficients, and SOC is the state of charge of the battery; In this embodiment, the fitting coefficient , , and Typical values ​​are: 0.0006, 0.7437, -0.9030, and 0.48; similarly, the fit coefficients... , , and Typical values ​​are 0.0006, 0.0005, -0.7633 and 0.49.

[0058] In this embodiment, the incremental integral strategy refers to describing the calendar aging of the battery module after a complete scheduling cycle based on the linearization of the first-order Taylor series. For lithium iron phosphate and ternary lithium battery modules, the cumulative calendar aging increments are as follows: = ×[1+ × + ×( - (4) = ×[1+ + ×( - (5) = × × (6) = × × (7) In the formula, , They are respectively in the time period Lithium iron phosphate and ternary lithium battery cabinets due to calendar aging Battery module in The cumulative calendar aging increment, This is the reference value for the battery's state of charge. This serves as the baseline for cumulative deployment time.

[0059] Step S302, in the current scheduling cycle In the current stage, the cumulative cycle aging degree of the battery module is determined by using the battery cycle aging state based on power time domain integration and the rain flow counting strategy. The cycle aging effect of a battery module is strongly correlated with its operating conditions. This invention describes the battery cycle aging state through an equivalent cycle depth strategy characterized by power time-domain integration, as shown in the following equation: (8) In the formula, This is an increment for the cyclic aging of the battery module. , All are undetermined coefficients. The depth of each loop; depth of each loop With 100% cycle depth The equivalent number of iterations is given by the following formula: × = × (9) In the formula, 100% loop depth For 100% depth loop count, This represents the actual depth of discharge. Based on equation (9), the coefficients Defined as: = (10) Based on the rainflow counting strategy, for each scheduling cycle, the battery module's cyclic aging state consists of a half-cycle of discharge cyclic aging and a half-cycle of charging cyclic aging. The cumulative cyclic aging degree is shown in the following formula: = + (11) In the formula, Battery cabinet Battery module in During the period Based on loop depth The cumulative degree of cyclic aging, , These represent the depth of discharge and the depth of charge cycle, respectively. , They are respectively in the time period The collection of discharge / charge cycles in a discharge / charge cycle. , These are the half-cycle coefficients under discharge and charge conditions, respectively. This is the increment for discharge cycle aging. This is the increment for charging cycle aging.

[0060] Step S303, in the current scheduling cycle In the current period, after determining the health status of the battery module by using the cumulative calendar aging increment and cumulative cycle aging degree, the actual available capacity of the battery module in the current scheduling cycle is determined in combination with the nominal rated capacity of the battery module. Battery cabinet Battery module in During the scheduling period The cumulative aging degree of the battery module It is modeled as a linear combination of cumulative calendar aging increment and cumulative cycle aging degree, satisfying = + , , Battery cabinet Battery module in During the scheduling period The cumulative calendar aging increment and the cumulative cycle aging degree.

[0061] Therefore, battery cabinet Battery module in During the scheduling period Health status is defined as =1-( + Based on this, the battery module's scheduling cycle is calculated. The actual available capacity is shown in the following formula: = × (12) In the formula, Battery cabinet Battery module in During the scheduling period The actual available capacity Battery cabinet Battery module in During the scheduling period A health status of -1 This refers to the nominal rated capacity of the battery module.

[0062] This invention combines a battery aging mechanism model with day-ahead power system scheduling. It decouples, models, and recouples calendar aging and cyclic aging mechanisms, resulting in clear physical meaning and higher accuracy than single empirical models. At each day-ahead stage, based on the latest accumulated aging, the battery's health status and actual available capacity are updated in real time. This ensures that scheduling decisions are based on current real-time capabilities, avoiding overcharging / over-discharging risks. Day-ahead scheduling allocates power based on the updated capacity, forming a closed-loop feedback loop where aging affects scheduling, and scheduling in turn affects aging, achieving a lifespan-optimal scheduling strategy. Furthermore, considering the potential lag in day-ahead updates, subsequent steps incorporate intraday rolling optimization, introducing more frequent aging status updates.

[0063] Step S304, in the current scheduling cycle During the intraday phase, the cumulative cycle aging sub-gradient of the battery module is determined based on the sub-gradient strategy.

[0064] In the day-ahead phase of the scheduling cycle, this invention first calculates the cycle depth using equation (8). The incremental cyclic aging is calculated using Equation (11), which then calculates the cumulative cyclic aging degree. However, this method suffers from significant nonlinearity, and the solver cannot obtain accurate results. Therefore, the cumulative cyclic aging degree calculated in the current phase cannot be used as the basic data for the fine-grained control of each battery module. Furthermore, since the charge / discharge state is not allowed to change or only changes at the initial moment in the current phase, there is no dynamic extreme point. Accurate aging calculation requires considering the extreme point of each battery charge / discharge state switch as a dynamic variable. To address this, this invention quantifies the cumulative cyclic aging sub-gradient of the battery module during short-term scheduling based on a sub-gradient strategy, including: 1) Cumulative cycle aging gradient of the battery module under conditions where the charge / discharge state is not switched: Specifically, when the charging and discharging operating state of the battery module does not switch, that is, it is always in a discharging or charging state, the cumulative cycle aging sub-gradient of the battery module satisfies the following relationship: × (13) In the formula, Battery cabinet Battery module in The cumulative cyclic aging gradient, Battery cabinet Battery module in During the period power, For the time period Based on power, the cumulative cycle aging degree of the battery module. For scheduling duration, Battery cabinet Battery module in During the period Based on discharge cycle depth The power gradient; Specifically, the discharge cycle depth of the battery module is updated based on the actual available capacity and power of the battery module in the current scheduling cycle. As shown in the following formula: =| - |+ × (14) In the formula, Battery cabinet Battery module in During the period A charge state of -1 Battery cabinet Battery module in Extreme points of charge / discharge state switching -1, For extreme point index, Battery cabinet Battery module in During the period power, Battery cabinet Battery module in During the period The actual available capacity For scheduling duration; 2) Cumulative cycle aging gradient of the battery module under charge / discharge switching conditions: For the charge / discharge switching moment, the cumulative cycle aging subgradient is given by the following formula: × (15) In the formula, It is a positive real number, and in the example, it is taken as 12.8.

[0065] Based on the fundamental objective of accumulating cyclic aging subgradients to eliminate nonlinear problems, it is clear that | - Since the constraint is nonlinear, an auxiliary variable is introduced. This can be transformed into an equivalent linear constraint, as shown in the following equation: ≤ - + × (16) or ≤- + + × (17) Using auxiliary variables in ultra-short and short-short optimization timescales The locally linearized power gradient is linearized and expanded using piecewise linearization and Taylor series linearization respectively, resulting in a linear expression for the cumulative cycle aging subgradient of the battery module under the condition of no switching between charge and discharge states, as shown in the following equation: = + (18) In the formula, Battery cabinet Battery module in During the period Auxiliary variables, For the time period A linear expression for the cumulative cyclic aging subgradient of the battery module based on auxiliary variables. , All are battery cabinets Battery module in linearization coefficients, Battery cabinet Battery module in During the period The baseline value for the cycle depth, =| - |

[0066] The obtained linear expression for the cumulative cyclic aging subgradient can be used as the basis for subsequent fine-grained control.

[0067] Step S305: At the end of the current scheduling cycle, update the aging status parameters of the battery module.

[0068] Based on the state of charge (SOC) of the battery modules and the actual available capacity of the current scheduling cycle, determine the charging and discharging power of each battery module during the intraday phase of the current scheduling cycle. Based on the sum of the cumulative aging state parameters of the battery modules at the end of the previous scheduling cycle and the increment of the aging state parameters of the battery modules at the end of the current scheduling cycle, determine the cumulative aging state parameters of the battery modules at the end of the current scheduling cycle. After one complete scheduling cycle, record the SOC sequence of all battery modules for today. } and the complete SOC sequence for all scheduled days { Operating temperature }, and based on { }and{ Update the cumulative calendar aging increment of today's battery module, based on { Update the cumulative cycle aging level of the battery module, and calculate the actual available capacity of the battery module for the next scheduling cycle based on step S303.

[0069] Step S40: Based on the electrical topology of the cascaded battery energy storage station, the leader agent establishes full-link constraints, including: state constraints of the battery module, power constraints of the DC / DC converter, power constraints of the DC / AC inverter, and power constraints of the transformer.

[0070] In this invention, each battery module needs to be regulated by a DC / DC converter during charging / discharging. All battery modules in the same battery cabinet convert DC power to AC power through the same DC / AC inverter, and finally connect to the grid through a transformer for isolation. Therefore, the load losses of the DC / DC converter, DC / AC inverter and transformer need to be considered.

[0071] Specifically, step S40 includes: Step S401: Establish power constraints for the DC / DC converters of each battery module; The power of the DC / DC converter under battery module charging and discharging conditions are shown in the following formulas: = - , >0 (19) = + , ≤0 (20) In the formula, Battery cabinet Battery module in DC / DC converter in time period power, , Battery cabinet Battery module in During the period Discharge power and charging power Battery cabinet Battery module in DC / DC converter in time period Power loss.

[0072] The aforementioned "if-else" logic block is difficult to directly parse in commercial solvers such as Gurobi or CPLEX. Therefore, the "Big M method" is used to equivalently transform the above linear constraints to achieve mutual exclusion constraints for the converter in the charging / discharging state, as shown in the following equation: ≤ × (twenty one) ≤ ×(1- )(twenty two) = - (twenty three) = + (twenty four) = - (25) In the formula, Battery cabinet Battery module in During the period The binary variable represents that any battery module can only be in one state of charging or discharging at any given time. Battery cabinet Battery module in The maximum power of the DC / DC converter, , Battery cabinet Battery module in DC / DC converter in time period The output power and input power, , Battery cabinet Battery module in DC / DC converter in time period The power loss during discharge and charging is also constrained by two variables.

[0073] The power loss of a DC / DC converter consists of two parts: conduction loss and switching loss, which are defined by the following formula: = + (26) = × (27) = ×DCR (28) = (29) = × × (30) In the formula, , Battery cabinet Battery module in DC / DC converter in time period The conduction loss and switching loss, , Battery cabinet Battery module in DC / DC converter in time period The conduction losses and inductor equivalent resistance losses of switching elements such as IGBTs. , Battery cabinet Battery module in During the period The root mean square current and current amplitude, DCR is the on-resistance of the switching transistor and the equivalent DC resistance of the inductor. Battery cabinet Battery module in During the period The amplitude of the ripple current. This refers to the parasitic capacitance value of switching elements such as IGBTs. This is the DC bus voltage. This refers to the switching frequency.

[0074] The equality constraint for the root-mean-square current of the battery module can be transformed into an equivalent form of the convex second-order cone constraint: ≥ + (31) In the formula, This represents the root mean square current of the battery module. For time period The current amplitude, This represents the amplitude of the ripple current.

[0075] Step S402: Establish power constraints for the DC / AC inverter; For DC / AC inverter losses, the calculation methods for conduction, switching, and filter losses differ significantly depending on the inverter topology. Therefore, this invention uses a more typical empirical formula for secondary power calculation, as shown in the following equation: = + × + × (32) In the formula, For the power loss of the DC / AC inverter, This refers to the DC-side input power of the DC / AC inverter. , and These are the equivalent fitting coefficients obtained for efficiencies at different power points.

[0076] Similarly, the inverter power loss can be expressed in linear form using a first-order Taylor series expansion: = + × + ×

[0077] +( +2 × )×( - (33) In the formula, For DC / AC inverters during time periods Power loss, The rated power of the DC / AC inverter. For DC / AC inverters during time periods DC-side input power; For DC / AC inverters, the following power constraints exist: = - (34) ≥0, ≥0 (35) <0, <0 (36) In the formula: , Battery cabinet DC / AC inverters during time periods AC side power and power loss, For time period The network-side load, for any scheduling time This value is a known constant.

[0078] Step S403: Establish power constraints for the transformer; Unlike converters, transformer losses consist of no-load losses that are independent of the load. and load-related load losses Composition. Let the power exchanged between the transformer and the grid side be... Under the charging / discharging conditions of the battery module, the transformer loss is given by the following formula: = - (37) In the formula, Battery cabinet The transformer during the time period Input power, Battery cabinet The transformer during the time period Lumped power loss.

[0079] No-load loss The no-load loss is caused by hysteresis and eddy current effects resulting from the alternating voltage applied to the transformer core. Since the grid voltage and frequency are approximately constant during normal operation, the no-load loss is a fixed value, and this data can be obtained from the transformer datasheet.

[0080] Load loss Ohmic losses are caused by the load current flowing through the internal resistance of the primary and secondary windings of the transformer. Therefore, load losses are proportional to the square of the current flowing through them, i.e.: = × (38) In the formula, Battery cabinet The transformer during the time period Load loss, The rated apparent power of the transformer. This is the rated power load loss of the transformer, which can be obtained from the transformer datasheet.

[0081] Similar to the power inequality constraint in DC / AC inverters, transformers also need to satisfy the following conditions during operation: The nonlinear constraint with the same sign is naturally satisfied by equations (35) and (36).

[0082] Step S404: Establish state constraints for the battery module; After accurately modeling the losses, the battery module's charging and discharging processes must also meet the following state constraints: 1) Battery module SOC dynamic constraints: = + × - × (39) 2) Battery module SOC limit constraints: ≤ ≤ (40) In the formula, , These represent the lower and upper limits of the state of charge of the battery module, respectively.

[0083] 3) Battery module charging power constraints: 0≤ ≤ (41) 0≤ ≤ (42) In the formula: , Battery cabinet Battery module in The upper limit of charging and discharging power.

[0084] 4) Battery module charging state mutual exclusion constraints: ≤ × (43) ≤ ×(1- (44) In step S50, the leader agent, based on the actual scheduling data and the day-ahead forecast data of the power grid in the previous scheduling cycle, solves the day-ahead scheduling scheme of each battery module in the current scheduling cycle under the constraints of the entire link and the robust Pareto optimization objective.

[0085] Specifically, step S50 includes: Step S501: Based on the actual scheduling data and the day-ahead forecast data of the power grid in the previous scheduling cycle, construct the day-ahead power fluctuation set on the grid side; To address the power uncertainty faced by battery energy storage modules during grid dispatch, this invention proposes a data-driven adaptive robust optimization framework, aiming to construct a dynamically evolving power fluctuation set. Characterizing the power of the grid side Its fluctuation characteristics.

[0086] Define the day-ahead grid-side power fluctuation set for: ={ | = + , (45) In the formula, For the predicted time period The current network side uncertain power set, For the predicted time period The current grid-side power collection, For the uncertain set of power, = × , For the normalized perturbation term, [-1,1].

[0087] The time-varying uncertainty fluctuations are adaptively updated using the maximum absolute error (MaxAE) within the sliding time window; therefore, the normalized disturbance term is as follows: = × {| - |}(46) In the formula, For safety margin coefficient, For time period The current grid-side power, For the predicted time period The current grid-side power.

[0088] Through the aforementioned adaptive update mechanism for uncertainty fluctuations, when the power prediction model is accurate on day d-1 (i.e., MaxAE is small), Automatic contraction within d days reduces the model's conservatism and improves the economy of day-ahead scheduling; conversely, if power fluctuations are severe within d-1 days, Adaptive expansion is used to ensure the robustness and feasibility of the day-ahead scheduling strategy.

[0089] This invention overcomes the limitations of simplified traditional battery energy storage models in the day-ahead scheduling phase by establishing a refined battery degradation model that includes calendar aging and cyclic aging by integrating Arrhenius's law and rainflow counting. Based on this refined model, a data-driven adaptive robust optimization framework is further proposed, dynamically updating the uncertainty set using the maximum absolute error within a sliding time window. This mechanism automatically adjusts the "shrinkage" and "expansion" of the uncertainty set according to the accuracy of grid-side power prediction, effectively overcoming the drawback of traditional robust optimization sacrificing economic efficiency due to excessive conservatism. It achieves a Pareto optimal balance between system economy and robustness under source-load uncertainty.

[0090] Step S502: Within the constraints of the cumulative calendar aging increment and cumulative cycle aging degree of the battery module, the actual available capacity of the battery module in the scheduling cycle, the full-link constraints, and the constraints of the day-ahead grid-side power fluctuation set, establish a robust Pareto optimization objective. The robust Pareto optimization objective is shown in the following equation: (47) The constraints of the robust Pareto optimization objective include: the cumulative calendar aging increment of the battery module shown in Equations (4) to (5), the actual available capacity of the battery module in the scheduling cycle shown in Equation (11), the cumulative cyclic aging degree of the battery module shown in Equation (12), the full-link constraints shown in Equations (19) to (44), and the day-ahead network-side power fluctuation set shown in Equation (45). In the formula, For deterministic decision variables, For deterministic decision space, For grid-side power, This is a collection of current grid-side power fluctuations. For the uncertain set of power The uncertainty error extracted from it, To optimize the objective function; Among them, an optimization objective function is established based on command power deviation penalty, battery module aging penalty, SOC deviation penalty between battery modules, power loss penalty of energy storage station, and circulating current penalty between battery modules. As shown in the following formula: = + + + + (48) In the formula, , , , , These are the Pareto undetermined coefficients for the following penalties: command power deviation penalty, battery module day-ahead aging penalty, inter-module SOC deviation penalty, energy storage station power loss penalty, and inter-module circulating current penalty. For cost electricity price, , , , , These are, respectively, power command deviation penalty, battery module day-ahead aging penalty, battery module SOC deviation penalty, energy storage station power loss penalty, and battery module circulating current penalty; 1) Command power deviation penalty, as shown in the following formula: = (49) In the formula, For the predicted time period The current grid-side power, For the time period Total power of the low-voltage side of the transformer in the energy storage power station = , Battery cabinet The transformer during the time period The input power; 2) The daily aging penalty of the battery module is shown in the following formula: = (50) In the formula, These are the weighting coefficients. The current scheduling cycle under the day-ahead scheduling. Total aging cost of energy storage power stations; The total aging cost of energy storage power stations under the current dispatch strategy is: = (51) In the formula, For the time period Total aging cost of energy storage power stations Battery cabinet Battery module in In the previous scheduling cycle -1 calendar aging cost, Battery cabinet Battery module in During the period The current day cycle aging cost; In the day-ahead scheduling scenario, only the cumulative cycle aging sub-gradient under the condition where the charge / discharge state of the battery module has not been switched is considered. Therefore, the day-ahead cycle aging cost is as follows: = + + (52) In the formula, Weighted by the degree of historical degradation. Battery cabinet Battery module in During the scheduling period The actual available capacity The construction cost allocated to each battery module per cycle. =1 indicates a lithium iron phosphate battery module. =2 indicates a ternary lithium battery module. Battery cabinet Battery module in In the previous scheduling cycle -1 represents the cumulative aging level of the battery module. , These are the half-cycle coefficients under discharge and charge conditions, respectively. , Based on the charging depth in the day-ahead scheduling and depth of discharge A linear expression for the cumulative cyclic aging subgradient under the condition of no switching between charge and discharge states is used to improve the accuracy of the calculation results.

[0091] 3) Penalty for SOC deviation between battery modules, as shown in the following formula: = (53) Under the current dispatch strategy, the SOC inconsistency between battery modules within an energy storage power station is as follows: = - (54) In the formula: For the time period Differences in lumped SOC of energy storage power stations For the time period Average SOC of energy storage power station Battery cabinet Battery module in During the period The current stage of SOC.

[0092] 4) The power loss penalty of the energy storage power station is shown in the following formula: = (55) The total lumped losses of converters and transformers in an energy storage power station are: = + (56) In the formula, For the time period Total losses of energy storage power stations Battery cabinet The transformer during the time period Lumped power loss, Battery cabinet DC / AC inverters during time periods Power loss, For battery module DC / DC converter in time period Power loss.

[0093] 5) The circulating current penalty between battery modules is shown in the following formula: = (57) Under the current scheduling strategy, circulating current between different battery modules within the same inverter is allowed to achieve better SOC and aging balance. Therefore, this invention constructs an auxiliary variable... To quantify the circulating current intensity between battery modules: = { , (58) In the formula, For the time period Auxiliary variables.

[0094] This invention not only establishes a full-link energy efficiency constraint model covering DC / DC converters, DC / AC inverters, and transformers, but also introduces binary variables through the Big M method to accurately characterize converter switching losses, conduction losses, and transformer no-load / load losses. This full-link perspective corrects the scheduling bias of traditional methods that only consider battery efficiency. Furthermore, based on a multi-agent system architecture, this invention can proactively manage the heterogeneous characteristics (SOH, SOC differences) of secondary-use battery modules. Moreover, by introducing a SOC consistency penalty term and an inter-module circulating current penalty into the objective function to allow controlled circulating current, it strategically guides modules in different health states to undertake differentiated power tasks. This not only achieves state equilibrium within the secondary-use battery cluster but also significantly extends the overall service life of the energy storage system by mitigating the bottleneck effect.

[0095] Step S503: Using equivalent transformation of second-order cone constraints, the robust Pareto optimization objective is transformed into a mixed-integer second-order cone programming (MISOCP) problem for day-ahead scheduling; solve the mixed-integer second-order cone programming problem to obtain the day-ahead scheduling scheme for each battery module over a long time scale; In the optimization objective, it is obvious , , , Depends only on deterministic decision variables ,and Simultaneously dependent on deterministic decision variables and uncertainty error This makes it a robust objective term. However, the two-level min-max optimization problem described by equation (47) is difficult to solve using commercial solvers such as Gurobi. Therefore, this invention first introduces an auxiliary variable. = - , For the predicted time period The current grid-side uncertain power For the predicted time period Given the current grid-side power, the two-layer optimization problem in equation (47) is transformed into an equivalent single-layer minimization problem: ( + (59) ≥ , (60) ≥ (61) The constraints of the single-layer minimization problem include: the cumulative calendar aging increment of the battery module shown in Equations (4) to (5), the actual available capacity of the battery module in the scheduling cycle shown in Equation (11), the cumulative cyclic aging degree of the battery module shown in Equation (12), the full-link constraints shown in Equations (19) to (44), and the semi-infinite constraints shown in Equations (61) to (63). In the formula, for , for , for , for , For time period The equivalent robust cost during optimization solution, The robustness error index is defined in the original sense. For uncertain error, As an auxiliary variable, For time period Second-order cone constraint auxiliary variables; Furthermore, for the semi-infinite constraint problem in equation (61), an equivalent transformation of the second-order cone constraint can be introduced, and it can be solved efficiently in commercial solvers such as Gurobi, specifically: ≥ , (62) ≥ , (63) In the formula, Let be the prediction bias at time t.

[0096] By adjusting the Pareto coefficient , , , , Furthermore, the mixed-integer second-order cone programming problem is solved to obtain Pareto front spaces with different emphases, including but not limited to: the optimal charge / discharge state soft-switching sequence of each battery module in the energy storage power station over a long time scale. This sequence is used as the day-ahead scheduling scheme to provide a power benchmark for subsequent short-term dynamic optimization.

[0097] In step S60, the leader agent, based on the day-ahead scheduling scheme, adopts different model predictive control strategies according to different time scales of the intraday scheduling instruction interval, and solves the intraday scheduling scheme of each battery module in the current scheduling cycle based on the MPC optimization objective.

[0098] Specifically, when the time scale of the intraday scheduling instruction interval is less than 5 minutes, a linear model predictive control strategy is adopted to establish an MPC model (QP-MPC) based on quadratic programming. In the QP-MPC model, the charging and discharging states of each battery module in the day-ahead scheduling scheme are adopted, and the power of each battery module at the intraday scheduling time is used as the control input to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of less than 5 minutes are obtained. Specifically, when the time scale of the intraday scheduling instruction interval is not less than 5 minutes, a mixed integer model predictive control strategy is adopted, and an MPC model based on mixed integer quadratic programming (MIQP-MPC) is established. In the MIQP-MPC model, the power and charge / discharge status of each battery module at the intraday scheduling time are used as control inputs to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of not less than 5 minutes are obtained. Intraday scheduling instructions with a time interval of less than 5 minutes and intraday scheduling instructions with a time interval of not less than 5 minutes constitute an intraday scheduling scheme.

[0099] In this embodiment, based on the intraday real-time power curve and ancillary service type, different time scales for intraday scheduling instruction intervals are determined, specifically as follows: When the ancillary service type is primary frequency regulation local control or secondary frequency regulation AGC control, the time scale of the intraday dispatch command interval is less than 5 minutes; when the ancillary service type is frequency regulation or peak regulation, the time scale of the intraday dispatch command interval is not less than 5 minutes; if the intraday real-time power curve is fluctuating, the time scale of the intraday dispatch command interval is less than 5 minutes; if the intraday real-time power curve is abrupt, the time scale of the intraday dispatch command interval is not less than 5 minutes.

[0100] During the intraday scheduling phase, as the timescale shortens, the load prediction accuracy improves accordingly, allowing for real-time adjustments to the day-ahead scheduling scheme using ultra-short-term forecast data. Model Predictive Control (MPC), as a flexible rolling optimization control strategy, can effectively handle multiple constraints and dynamically adjust control decisions based on the actual system state. Therefore, this invention uses an MPC strategy to perform real-time rolling adjustments and optimizations to the intraday scheduling scheme. Specifically, for MPC tasks with instruction intervals less than 5 minutes, this invention transforms them into a quadratic programming task for rapid solution; while for MPC tasks with instruction intervals greater than 5 minutes, this invention constructs a mixed-integer quadratic programming task.

[0101] Specifically, step S60 includes: Step S601: When the time scale of the intraday scheduling instruction interval is less than 5 minutes, a linear model predictive control strategy is adopted to establish an MPC model based on quadratic programming. In the MPC model based on quadratic programming, the charging and discharging states of each battery module in the daytime scheduling scheme are adopted, and the power of each battery module at the intraday scheduling time is used as the control input to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of less than 5 minutes are obtained. Based on the daily scheduling scheme of each battery module in the current scheduling cycle, and using a linear model predictive control strategy, a sub-optimal problem based on quadratic programming is established and solved to obtain intraday scheduling instructions with time intervals of less than 5 minutes. Specifically, in the sub-optimal problem based on quadratic programming, the charging and discharging state of each battery module is the same as that in the daily scheduling scheme based on robust Pareto optimization objective, with the power of each battery module at the intraday scheduling time as the basis. and To control the input, the charge state-space equations of the battery module are constructed as follows: = + (64) In the formula, In the suboptimal problem, the first The predicted charged state space at the next iteration. In the suboptimal problem, the first The charged state space at the next iteration. = , For the first Battery cabinet during the next iteration middle The state-of-charge vector of each battery module = , For the first Battery cabinet during the next iteration Mid-cell battery module The state of charge, In the suboptimal problem, the first Power space at the next iteration = , For the first Battery cabinet during the next iteration middle The power vector of each battery module = , For the first Battery cabinet during the next iteration Mid-cell battery module The power set, = , and For battery modules in different time periods The charging power and discharging power; , These are all coefficient matrices in a suboptimal optimization problem. = , In the suboptimal problem, the first Battery module during the next iteration The coefficient matrix, = , It is a 1×1 identity matrix. = , In the suboptimal problem, the first Battery cabinet during the next iteration The coefficient matrix, = , In the suboptimal problem, the first Battery cabinet during the next iteration Battery Module The set of coefficients, = , For the first Battery cabinet during the next iteration Battery Module The actual available capacity.

[0102] Step S602: When the time scale of the intraday scheduling instruction interval is not less than 5 minutes, a mixed integer model predictive control strategy is adopted to establish an MPC model based on mixed integer quadratic programming. In the MPC model based on mixed integer quadratic programming, the power and charge / discharge status of each battery module at the intraday scheduling time are used as control inputs to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of not less than 5 minutes are obtained. Based on the daily scheduling scheme of each battery module in the current scheduling cycle, and using a mixed-integer model predictive control strategy, an optimization problem based on mixed-integer quadratic programming is established and solved to obtain intraday scheduling instructions with a time interval of no less than 5 minutes. Similarly, in the optimization problem based on mixed-integer quadratic programming, the charging and discharging state of each battery module is regarded as the variable to be optimized, with the power of each battery module at the intraday scheduling time as the factor. , and charge / discharge state variables , To control the input, the charge state-space equations of the battery module are constructed as follows: = + (65) In the formula, For the optimization problem The predicted charged state space at the next iteration. For the optimization problem The charged state space at the next iteration. = , For the first Battery cabinet during the next iteration middle The state-of-charge vector of each battery module = , For the first Battery cabinet during the next iteration Mid-cell battery module The set of charged states, = , For the first Battery cabinet during the next iteration Mid-cell battery module The state of charge, For the optimization problem Power space at the next iteration = , For the first Battery cabinet during the next iteration middle The power vector of each battery module = , For the first Battery cabinet during the next iteration Mid-cell battery module The power set, = , and For battery modules in different time periods The charging power and discharging power, and For battery modules in different time periods The charging and discharging state variables, , Both are coefficient matrices in optimization problems. = , For the optimization problem Battery module during the next iteration The coefficient matrix, = , = , = , For the optimization problem Battery cabinet during the next iteration The coefficient matrix, = , For the optimization problem Battery cabinet during the next iteration Battery Module The set of coefficients, = , For the first Battery cabinet during the next iteration Battery Module The actual available capacity.

[0103] Step S603: Within the constraints imposed by the cumulative cyclic aging subgradient of the battery module, the full-link constraints, and the charge state space of the battery module, the MPC cost function is used. The minimum is the MPC optimization target.

[0104] Combining the two MPC optimization strategies mentioned above, the objective function for day-ahead scheduling is... Based on this, additional penalties are added for energy storage response command deviation, intraday and day-ahead battery module SOC deviation, and intraday battery module aging penalty is used to replace day-ahead battery module aging penalty, thus obtaining the MPC cost function. As shown in the following formula: = + + + + + + (67) In the formula, , , , , These are the Pareto undetermined coefficients for command power deviation penalty, battery module intraday aging penalty, battery module SOC deviation penalty, energy storage station power loss penalty, and battery module circulating current penalty, respectively. The weight for penalizing deviations in energy storage command response. The weighting of the battery module SOC deviation between the intraday phase and the previous day phase; The cost price of electricity; , , , , , , These are, respectively, power command deviation penalty, energy storage response command deviation penalty, battery module intraday aging penalty, battery module SOC deviation penalty, intraday and day-ahead battery module SOC deviation penalty, energy storage power station power loss penalty, and battery module circulating current penalty. The penalty for deviation in energy storage response command is shown in the following formula: = + (69) In the formula, , They are respectively in the time period The reverse power deviation and forward power deviation of the energy storage power station in response to ancillary service commands. and Through binary variables Form mutual exclusion constraints; The daily aging penalty for the battery module is shown in the following formula: =

[0105] In the formula, These are the weighting coefficients. For intraday scheduling, the current scheduling cycle Total aging cost of energy storage power stations For scheduling duration, The scheduling period; Among them, the total aging cost of energy storage power stations under intraday dispatch is: =

[0106] In the formula, Battery cabinet Battery module in In the previous scheduling cycle -1 calendar aging cost, Battery cabinet Battery module in During the period The daily cycle aging cost, This refers to the number of battery cabinets. This refers to the number of battery modules; Under intraday scheduling, the cumulative cycle aging sub-gradient is selected based on the battery module's operating condition between the state of charge / discharge not switching and the operating condition at the moment of switching. The intraday cycle aging cost is shown in the following formula: = + +

[0107] In the formula, Weighted by the degree of historical degradation. Battery cabinet Battery module in During the scheduling period The actual available capacity The construction cost allocated to each battery module per cycle. =1 indicates a lithium iron phosphate battery module. =2 indicates a ternary lithium battery module. Battery cabinet Battery module in In the previous scheduling cycle -1 represents the cumulative aging level of the battery module. , These are the half-cycle coefficients under discharge and charge conditions, respectively. , Based on the charging depth during intraday scheduling and depth of discharge The linear expression for the cumulative cyclic aging subgradient needs to be selected based on the charging and discharging state of the battery module, between the non-switching state and the switching state.

[0108] The penalty for the deviation of battery module SOC between intraday and daytime stages is shown in the following formula: = (70) In the formula, , Battery cabinet Battery module in During the period The SOC for the day-ahead and intraday phases.

[0109] With MPC cost function The minimum is the MPC optimization target, and the constraints of the MPC optimization target include: the cumulative cyclic aging subgradient of the battery module shown in Equations (13) to (15), the full-link constraint shown in Equations (19) to (44), and the charge state space of the battery module shown in Equation (64) or (65).

[0110] Step S604: Establish an intraday rolling optimization trigger mechanism based on the dynamic characteristics of the MPC optimization target and system stability constraints.

[0111] Both quadratic programming and mixed-integer quadratic programming problems have high computational complexity and require significant computing resources. Under conditions of smooth power fluctuations or steady-state system operation, fixed-period time-domain rolling optimization leads to a large amount of redundant computation, resulting in unnecessary waste of computing power. To improve computational efficiency, this invention is based on optimizing the objective function... Based on the dynamic characteristics, an event triggering mechanism for MPC optimization tasks was designed, as shown in the following equation: =inf{ ≥ +H· | (71) In the formula, , The first +1st time and the first The next MPC optimization event trigger time point, where H represents the MPC prediction field of view. and These are all event triggering functions, and are defined as follows: = ≥ (72) = ≤0 (73) In the formula, For time period The MPC cost function value, The deviation coefficient, For the first The trigger time of the next MPC optimization event The MPC cost function value. In this embodiment, It is 1.15.

[0112] The first triggering event is when the MPC cost function value in the current time period is not less than the MPC cost function value when the previous optimization event was triggered, that is, when it is found that the MPC optimization target cannot be met, intraday rolling optimization is triggered. For time period The current grid-side power, For the time period The total power on the low-voltage side of the energy storage power station transformer, if the product of the two is negative, indicates that the power flow direction on the grid side and the power flow direction on the transformer are inconsistent. This is not allowed for the normal and stable operation of the system. Therefore, it is necessary to trigger intraday rolling optimization in a timely manner as the second triggering event.

[0113] This invention addresses the dual requirements of real-time performance and computational efficiency in intraday scheduling. It designs a multi-timescale coupled model predictive control architecture and adaptively switches between quadratic programming and mixed-integer quadratic programming solution strategies based on differences in instruction time intervals, achieving a dynamic match between control accuracy and solution speed. Furthermore, this invention proposes an event-triggered mechanism based on the dynamic characteristics of the objective function. By monitoring state deviations and the rate of change of the objective function within the prediction field, optimization calculations are triggered only when the system state deviates significantly or operating conditions fluctuate drastically. This design significantly reduces redundant computational load under conditions of smooth power or steady-state operation, greatly improves the real-time response capability of the system under limited embedded computing resources, and ensures computational stability during long-term operation.

[0114] In step S70, the leader agent's state matrix is ​​constructed using the day-ahead scheduling scheme and the intraday scheduling scheme. Each follower agent calculates local control commands through a distributed fixed-time sliding mode controller based on the leader agent's state matrix, its own state information, and the state information of neighboring follower agents. Each battery module performs power adjustment operations according to the corresponding local control commands.

[0115] Unlike steps S20 to S60, which are all based on a centralized control framework (i.e. calculated by the leader), step S70 is based on a distributed control framework, where each follower agent follows the leader's state matrix through a distributed fixed-time sliding mode controller based on the leader's state and neighboring agent information.

[0116] Specifically, step S70 includes: Step S701: Based on the exogenous system, construct a dynamic model of the tracking error of the follower agent.

[0117] For the Each battery module defines its local state variables as follows: The dynamic equation is described as follows: = + (74) In the formula, For the first Each battery module during the period The actual state For the time period The control input to be designed For the time period Lumped disturbance term (including model uncertainty and external disturbance).

[0118] Definition of the first The tracking error of each battery module to the reference trajectory and its derivative: = - (75) = - = + - (76) In the formula, For the first Each battery module during the period Tracking error of the reference trajectory, The leader is in a certain state.

[0119] Step S702: Design a non-singular fixed-time integral sliding surface and a distributed controller based on a fixed-time reaching law.

[0120] To ensure that the tracking error converges within a fixed time and to avoid singularities, an integral sliding surface of the following form is constructed: = + (77) In the formula, For the first Each battery module during the period Integral sliding surface, , All are constants greater than 0. , All are constants greater than 0. .

[0121] In this embodiment, , , , The values ​​are 1.5, 0.5, 15, and 15 respectively.

[0122] To ensure that the system state reaches the sliding surface within a fixed time, the following double power fixed-time approach law is adopted: =- - (78) In the formula, , All are constants greater than 0. , All are constants greater than 0; In this embodiment, , , , The values ​​are 1.5, 0.5, 5.5, and 5.5, respectively.

[0123] = - - - -

[0124] In the formula, For perturbation switching gain, This is a symbolic function. In the example, It is 25.

[0125] Under the control of the aforementioned controller, the total time required for the system state tracking error to converge to zero is... There exists an upper bound that is independent of the initial state. : ≤ , = + (79) ≤ + + + (80) The proposed method for long-term and short-term dynamic management and control of battery energy storage modules based on a multi-agent system was validated using a case study based on power auxiliary service monitoring data from a certain region (time span of 30 days, starting from October 1, 2023). The number of lithium iron phosphate battery modules was 4 (LPF0, LPF1, LPF2, and LPF3); the initial state of charge were 0.59, 0.54, 0.51, and 0.49; the initial health status was 0.96, 0.92, 0.88, and 0.84; and the cumulative service time was 56, 238, 533, and 1008 days, respectively. The number of ternary lithium battery modules is 6 (namely: NMC0, NMC1, NMC2, NMC3, NMC4, and NMC5); the initial state of charge are 0.55, 0.54, 0.52, 0.5, 0.48, and 0.46, respectively; the initial state of health are 0.95, 0.92, 0.89, 0.86, 0.83, and 0.8, respectively; and the cumulative service time is 101, 263, 505, 826, 1228, and 1668 days, respectively.

[0126] All results are based on equilibrium trade-offs in the Pareto optimal front. To address the nonlinear coupling and competition among multi-scale scheduling objectives in BESS, this example selects Pareto weighting coefficients that consider both the system's external characteristics and internal state. In the current phase... , and They are 10, 5, and 5 respectively; during the intraday period, , and They are 40, 20, and 10 respectively; , , They are 1e4, 20, and 5 respectively.

[0127] The calculation results show that the strategy of the present invention can implement differentiated power allocation based on the real-time SOC and SOH of each module and the marginal operating cost, so that the energy throughput is transferred in an orderly manner within the cluster: LPF0, which has a better state and lower cost, is prioritized as the main energy throughput theme, with a cumulative positive output of 3706.60 kWh, accounting for 3.70%; while NMC5, which has a worse state and higher cost, is strategically protected by load reduction, with a cumulative positive output of 1758.71 kWh, accounting for 6.55%, which is about 52.19% lower than that of LPF0, effectively suppressing its excessive aging.

[0128] The SOC active balancing mechanism of this invention can effectively address the state dispersion caused by differences in internal resistance, capacity, and aging rate between modules. The SOC range was further reduced from 11.08% at 07:00 on October 12th to 1.26% at 19:00 on October 15th. This process demonstrates that the strategy can achieve continuous correction of SOC from dynamic deviation to high-precision synchronization without sacrificing external power tracking, thereby improving the stability of long-term collaborative operation.

[0129] Corresponding to SOC convergence, the module SOH decay exhibits a depolarization equilibrium trend within the scheduling cycle. The strategy adaptively adjusts power weights through SOH feedback to promote consistency. LPF0, with a higher SOH, initially bears a higher load, with a relative aging rate as high as 237%. As SOH gradually approaches the cluster mean, the strategy reduces its priority, causing the relative aging rate to drop to 223% in the final stage. Conversely, the relative aging rate of modules with lower SOH increases slowly. Overall, the standard deviation of the module relative aging rate converges from 0.619 to 0.563, and the SOH decay of LPF0 and NMC5 are 0.574% and 0.154%, respectively, indicating that the strategy effectively achieves dynamic SOH equilibrium while maintaining response accuracy.

[0130] In terms of operational efficiency, the average operating efficiency of BESS during the 30-day cycle scheduling was 84.41%. Due to the randomness and intermittency of frequency regulation commands, the system inevitably experiences light load conditions. At this time, the proportion of fixed losses in the converter increases, leading to a drop in efficiency. For example, at 17:20 on October 23, the grid-side demand was only 1.56 kW, and the efficiency briefly dropped to 45.10%. However, overall, the strategy maintains cycle economy by optimizing power weights to concentrate the main throughput in a higher efficiency range.

[0131] Regarding command tracking, the average response rate across the entire cycle reached 98.45%, indicating that accurate tracking of frequency modulation commands can be achieved in the vast majority of operating windows. A few outliers were mainly triggered by battery physical boundaries: for example, at 03:00 on October 30th, after continuous high-power discharge, the average SOC dropped to 10.15% and triggered the lower limit protection, resulting in a decrease in response rate due to discharge being locked out. Except for extreme moments affected by rigid constraints such as upper / lower SOC limits, the strategy demonstrated stable reliability and accuracy.

[0132] Compared with traditional benchmark methods, the strategy of this invention outperforms the traditional methods in both response accuracy and efficiency: the frequency modulation command response rate is 98.45%, which is 5.52% and 7.20% higher than maximum power scheduling and remaining capacity scheduling, respectively; the average operating efficiency is 84.41%, which is also better than the benchmark strategies of 82.28% and 82.99%. Traditional rigid scheduling lacks fine-grained perception and active constraints on the health status of heterogeneous modules, which leads to a deterioration of SOC dispersion to 4.48% under the maximum power strategy, making it more likely to trigger the "barrel effect" caused by the first arrival of a single module and force the system to derating. The strategy of this invention suppresses state divergence through robust multi-timescale coordinated allocation, thereby improving the operational robustness and life-cycle economy of heterogeneous BESS.

[0133] This invention also proposes a long-term and short-term dynamic management and control system for modules in a cascaded battery energy storage station, comprising: The automatic power generation control is modeled as a leader agent, and a battery module is modeled as a follower agent. The leader agent is used to acquire day-ahead forecast data of the power grid and real-time operating status parameters of all battery modules. Based on the real-time operating status parameters of all battery modules, the aging status parameters of each follower agent in the day-ahead and intraday phases within the scheduling cycle are determined. Based on the electrical topology of the cascaded battery energy storage station, full-link constraints are established, including: state constraints of battery modules, power constraints of DC / DC converters, power constraints of DC / AC inverters, and power constraints of transformers. Based on the actual scheduling data and day-ahead forecast data of the power grid in the previous scheduling cycle, under full-link constraints, the day-ahead scheduling scheme of each battery module in the current scheduling cycle is solved based on the robust Pareto optimization objective. Based on the day-ahead scheduling scheme, different model predictive control strategies are adopted according to different time scales of intraday scheduling command intervals, and the intraday scheduling scheme of each battery module in the current scheduling cycle is solved based on the MPC optimization objective. The leader agent's state matrix is ​​composed of the day-ahead scheduling scheme and the intraday scheduling scheme. Each follower agent calculates local control commands through a distributed fixed-time sliding mode controller based on the leader agent's state matrix, its own state information, and the state information of neighboring follower agents. Each battery module performs power adjustment operations according to the corresponding local control commands.

[0134] To address the challenge of dynamic tracking of upper-level optimization commands during lower-level execution, this invention constructs a distributed fixed-time sliding mode control strategy based on the exogenous system state matrix at the lower-level control stage. By designing a non-singular fixed-time integral sliding mode surface and a double power-law approach, this invention theoretically proves the global upper bound of the system tracking error convergence time. This control law ensures that the distributed energy storage unit can track the power commands issued by the MPC without steady-state error within a preset fixed time, regardless of its initial state, and exhibits strong robustness to model parameter perturbations and external disturbances. This hierarchical architecture of upper-level optimization scheduling and lower-level fixed-time control achieves comprehensive decoupling and efficient coordination from the time scale (day-to-day-real-time) to the control level (planning-tracking).

[0135] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0136] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0137] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0138] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station, characterized in that, include: Automatic power generation control is modeled as a leader agent, and a battery module is modeled as a follower agent; the leader agent acquires the day-ahead forecast data of the power grid and the real-time operating status parameters of all battery modules; The leader agent determines the aging state parameters of each follower agent in the day-ahead and intraday phases within the scheduling cycle based on the real-time operating status parameters of all battery modules. The leader agent establishes full-link constraints based on the electrical topology of the tiered battery energy storage station, including: state constraints of battery modules, power constraints of DC / DC converters, power constraints of DC / AC inverters, and power constraints of transformers. Based on the actual scheduling data and the day-ahead forecast data of the power grid in the previous scheduling cycle, the leader agent solves the day-ahead scheduling scheme of each battery module in the current scheduling cycle under the full-link constraints and based on the robust Pareto optimization objective. Among them, a robust Pareto optimization objective function is established based on command power deviation penalty, battery module aging penalty, inter-module SOC deviation penalty, energy storage power station power loss penalty, and inter-module circulating current penalty. As shown in the following formula: = + + + + Within the constraints imposed by the cumulative cyclic aging subgradient of the battery module, the full-link constraints, and the charge state space of the battery module, the MPC cost function is used. Minimize the MPC optimization objective, and the MPC cost function is shown in the following equation: = + + + + + + In the formula, , , , , These are the Pareto undetermined coefficients for command power deviation penalty, battery module intraday aging penalty, battery module SOC deviation penalty, energy storage station power loss penalty, and battery module circulating current penalty, respectively. The weight for penalizing energy storage command response deviation. The weighting of the battery module SOC deviation between the intraday phase and the previous day phase; The cost price of electricity; , , , , , , , These are: power command deviation penalty, energy storage response command deviation penalty, battery module day-ahead aging penalty, battery module intraday aging penalty, battery module SOC deviation penalty, intraday and day-ahead stage battery module SOC deviation penalty, energy storage power station power loss penalty, and battery module circulating current penalty. Based on the day-ahead scheduling scheme, the leader agent adopts different model predictive control strategies according to different time scales of intraday scheduling instruction intervals, and solves the intraday scheduling scheme of each battery module in the current scheduling cycle based on the MPC optimization objective. When the time scale of the intraday scheduling instruction interval is less than 5 minutes, a linear model predictive control strategy is adopted to establish an MPC model based on quadratic programming. In the MPC model based on quadratic programming, the charging and discharging states of each battery module in the daytime scheduling scheme are adopted, and the power of each battery module at the intraday scheduling time is used as the control input to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of less than 5 minutes are obtained. When the time scale of the intraday scheduling instruction interval is not less than 5 minutes, a mixed integer model predictive control strategy is adopted, and an MPC model based on mixed integer quadratic programming is established. In the MPC model based on mixed integer quadratic programming, the power and charge / discharge status of each battery module at the intraday scheduling time are used as control inputs to establish the charge state space equation of the battery module. Based on the MPC optimization objective, intraday scheduling instructions with a time interval of not less than 5 minutes are obtained. Intraday scheduling instructions with a time interval of less than 5 minutes and intraday scheduling instructions with a time interval of not less than 5 minutes constitute an intraday scheduling scheme; The leader agent's state matrix is ​​composed of the day-ahead scheduling scheme and the intraday scheduling scheme. Each follower agent calculates local control commands through a distributed fixed-time sliding mode controller based on the leader agent's state matrix, its own state information, and the state information of neighboring follower agents. Each battery module performs power adjustment operations according to the corresponding local control commands.

2. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The aging state parameters of each follower agent in the day-ahead phase include: cumulative calendar aging increment and cumulative cyclic aging degree; the aging state parameters of each follower agent in the intraday phase include: cumulative cyclic aging sub-gradient.

3. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 2, characterized in that, During the day-ahead phase of the current scheduling cycle, based on the SOC and operating temperature of all battery modules at the end of the previous scheduling cycle, the cumulative calendar aging increment of the battery modules is determined using the capacity decay of the battery modules caused by calendar aging based on a semi-empirical exponential model and an incremental integration strategy. During the day-ahead phase of the current scheduling cycle, the cumulative cycle aging degree of the battery module is determined by utilizing the battery cycle aging state based on power time-domain integration and the rainflow counting strategy. During the intraday phase of the current scheduling cycle, based on the sub-gradient strategy, the cumulative cyclic aging sub-gradient of the battery module is determined, including: the cumulative cyclic aging sub-gradient of the battery module under the condition of no switching between charge and discharge states, and the cumulative cyclic aging sub-gradient of the battery module under the condition of switching between charge and discharge states.

4. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 3, characterized in that, During the day-ahead phase of the current scheduling cycle, the health status of the battery module is determined by using the cumulative calendar aging increment and cumulative cycle aging degree of the battery module. Then, the actual available capacity of the battery module in the current scheduling cycle is determined by combining the nominal rated capacity of the battery module.

5. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 4, characterized in that, Based on the actual available capacity and power of the battery module in the current scheduling cycle, the discharge cycle depth of the battery module is updated; the cumulative cycle aging sub-gradient of the battery module under the condition of no switching between charge and discharge states is calculated using the power gradient of the battery module based on the discharge cycle depth, and the linear expression of the cumulative cycle aging sub-gradient of the battery module is obtained by linearizing the power gradient.

6. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 5, characterized in that, Within the constraints of the cumulative calendar aging increment and cumulative cycle aging degree of the battery module, the actual available capacity of the battery module during the scheduling cycle, the full-link constraints, and the day-ahead power fluctuation set of the network side, a robust Pareto optimization objective is established, as shown in the following formula: In the formula, For deterministic decision variables, For deterministic decision space, For grid-side power, The day-ahead grid-side power fluctuation set is constructed based on the actual dispatch data and the day-ahead forecast data of the power grid in the previous dispatch cycle. For the uncertain set of power The uncertainty error extracted from it, To optimize the objective function.

7. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The penalty for command power deviation is shown in the following formula: = In the formula, For the predicted time period The current grid-side power, For the time period Total power of the low-voltage side of the transformer in the energy storage power station = , Battery cabinet The transformer during the time period Input power, For scheduling duration, For the scheduling period, This refers to the number of battery cabinets.

8. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The current aging penalty of the battery module is shown in the following formula: = In the formula, These are the weighting coefficients. The current scheduling cycle under the day-ahead scheduling. Total aging cost of energy storage power stations For scheduling duration, The scheduling period; Among them, the total aging cost of energy storage power stations under the current dispatch is: = In the formula, Battery cabinet Battery module in In the previous scheduling cycle -1 calendar aging cost, Battery cabinet Battery module in During the period The current cycle aging cost, This refers to the number of battery cabinets. This refers to the number of battery modules; Under the current scheduling, considering only the cumulative cycle aging gradient under the condition that the charge / discharge state of the battery module has not switched, the current cycle aging cost is as follows: = + + In the formula, Weighted by the degree of historical degradation. Battery cabinet Battery module in During the scheduling period The actual available capacity The construction cost allocated to each battery module per cycle. =1 indicates a lithium iron phosphate battery module. =2 indicates a ternary lithium battery module. Battery cabinet Battery module in In the previous scheduling cycle -1 represents the cumulative aging level of the battery module. , These are the half-cycle coefficients under discharge and charge conditions, respectively. , Based on the charging depth in the day-ahead scheduling and depth of discharge The linear expression for the cumulative cyclic aging subgradient under the condition where the charge and discharge states are not switched.

9. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The SOC deviation penalty between battery modules is shown in the following formula: = The inconsistency in SOC between battery modules within an energy storage power station is as follows: = - In the formula: For the time period Differences in lumped SOC of energy storage power stations For the time period Average SOC of energy storage power station For scheduling duration, For the scheduling period, This refers to the number of battery cabinets. For the number of battery modules, Battery cabinet Battery module in During the period The current stage of SOC.

10. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The power loss penalty for energy storage power stations is shown in the following formula: = The total lumped losses of converters and transformers in an energy storage power station are: = + In the formula, For the time period Total losses of energy storage power stations Battery cabinet The transformer during the time period Lumped power loss, Battery cabinet DC / AC inverters during time periods Power loss, For battery module DC / DC converter in time period Power loss, For scheduling duration, For the scheduling period, This refers to the number of battery cabinets. This represents the number of battery modules.

11. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The inter-module circulating current penalty is shown in the following formula: = Auxiliary variables The following is used to quantify the circulating current intensity between battery modules: = { ， } In the formula, For the time period Auxiliary variables, , Battery cabinet Battery module in DC / DC converter in time period The output power and input power, For scheduling duration, For the scheduling period, This refers to the number of battery cabinets. This represents the number of battery modules.

12. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The penalty for energy storage response command deviation is shown in the following formula: = + In the formula, , They are respectively in the time period Reverse power deviation and forward power deviation of energy storage power station in response to ancillary service commands.

13. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The penalty for the deviation of battery module SOC between intraday and daytime stages is shown in the following formula: = In the formula, , Battery cabinet Battery module in During the period SOC for the day-end and intraday phases, This refers to the number of battery cabinets. This represents the number of battery modules.

14. The method for long-term and short-term dynamic management and control of modules in a cascaded battery energy storage station according to claim 1, characterized in that, The daily aging penalty for the battery module is shown in the following formula: = In the formula, These are the weighting coefficients. For intraday scheduling, the current scheduling cycle Total aging cost of energy storage power stations For scheduling duration, The scheduling period; Among them, the total aging cost of energy storage power stations under intraday dispatch is: = In the formula, Battery cabinet Battery module in In the previous scheduling cycle -1 calendar aging cost, Battery cabinet Battery module in During the period The daily cycle aging cost, This refers to the number of battery cabinets. This refers to the number of battery modules; Under intraday scheduling, the cumulative cycle aging sub-gradient is selected based on the battery module's operating condition between the state of charge / discharge not switching and the operating condition at the moment of switching. The intraday cycle aging cost is shown in the following formula: = + + In the formula, Weighted by the degree of historical degradation. Battery cabinet Battery module in During the scheduling period The actual available capacity The construction cost allocated to each battery module per cycle. =1 indicates a lithium iron phosphate battery module. =2 indicates a ternary lithium battery module. Battery cabinet Battery module in In the previous scheduling cycle -1 represents the cumulative aging level of the battery module. , These are the half-cycle coefficients under discharge and charge conditions, respectively. , Based on the charging depth during intraday scheduling and depth of discharge The linear expression for the cumulative cyclic aging subgradient.

15. A long-term and short-term dynamic management and control system for modules in a cascaded battery energy storage station, used to implement the long-term and short-term dynamic management and control method for modules in a cascaded battery energy storage station as described in any one of claims 1 to 14, characterized in that, include: The automatic power generation control is modeled as a leader agent, and a battery module is modeled as a follower agent. The leader agent is used to acquire the day-ahead forecast data of the power grid and the real-time operating status parameters of all battery modules; based on the real-time operating status parameters of all battery modules, the aging status parameters of each follower agent in the day-ahead and intraday phases within the scheduling cycle are determined. Based on the electrical topology of the cascaded battery energy storage station, full-link constraints are established, including: state constraints of battery modules, power constraints of DC / DC converters, power constraints of DC / AC inverters, and power constraints of transformers. Based on the actual scheduling data and day-ahead forecast data of the power grid from the previous scheduling cycle, and under full-link constraints, the day-ahead scheduling scheme for each battery module in the current scheduling cycle is obtained using a robust Pareto optimization objective. Based on the day-ahead scheduling scheme, and according to different time scales of intraday scheduling command intervals, different model predictive control strategies are adopted, and the intraday scheduling scheme for each battery module in the current scheduling cycle is obtained using an MPC optimization objective. The leader agent's state matrix is ​​composed of the day-ahead scheduling scheme and the intraday scheduling scheme. Each follower agent calculates local control commands through a distributed fixed-time sliding mode controller based on the leader agent's state matrix, its own state information, and the state information of neighboring follower agents. Each battery module performs power adjustment operations according to the corresponding local control commands.

16. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-14.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-14.

Images