A microgrid energy management method based on dynamic perception of energy storage battery health state

Abstract

The present invention relates to a microgrid energy management method based on dynamic perception of the health state of energy storage batteries, belonging to the field of microgrid energy management. It includes constructing a dynamic perception model of the battery health state, with the original data uploaded by the battery management system supporting the energy storage battery cluster as the input; constructing a planning layer, establishing an objective function with the minimum operating cost of the energy storage system throughout its life cycle, and the optimized output being the optimal rated capacity and power level of the energy storage system; constructing a real-time layer, taking the output of the planning layer as the initial boundary, adopting a collaborative architecture of model predictive control and particle swarm optimization, and the MPC real-time calling the driven dynamic constraint set to generate the optimal energy storage power command within the current rolling window, so as to solve the technical problems that due to the lack of dynamic perception and closed-loop feedback mechanism for the health state of energy storage batteries, the scheduling plan is seriously mismatched with the actual available capacity, thereby leading to a decline in the economic efficiency of system operation and an accelerated decay of battery life.

Description

Technical Field

[0001] This invention relates to a microgrid energy management method based on dynamic sensing of the health status of energy storage batteries, belonging to the field of microgrid energy management. Background Technology

[0002] As the global energy structure transitions towards cleaner and lower-carbon energy, the penetration rate of renewable energy sources, such as distributed photovoltaic (PV) and wind power, in the power system continues to increase. However, due to the significant intermittency, volatility, and uncontrollability of solar and wind power, large-scale grid integration can easily lead to problems such as voltage exceeding limits, frequency deviations, and power imbalances, seriously threatening the safe and stable operation of the distribution network. To improve the absorption capacity of renewable energy and the reliability of power supply, microgrid technology has emerged and developed rapidly. A microgrid is a small-scale autonomous power system that integrates distributed generation systems (DERs), loads, energy storage systems (ESS), and control devices. It can flexibly switch between grid-connected and islanded modes, achieving local energy self-balancing and highly resilient power supply.

[0003] In microgrid systems, energy storage systems play a crucial role. Their core functions include smoothing out fluctuations in renewable energy output, peak shaving and valley filling, providing backup power, and participating in frequency regulation and voltage support. Currently, lithium-ion batteries, due to their high energy density, long cycle life, and fast response characteristics, have become the mainstream energy storage technology in microgrids. However, batteries inevitably age during long-term charge-discharge processes, leading to capacity decay, increased internal resistance, and decreased thermal stability, which in turn affects the economy, safety, and operational efficiency of the entire microgrid system. Therefore, how to accurately perceive the battery's State of Health (SOH) and integrate it into the microgrid energy management system (EMS) has become a key scientific problem and technological bottleneck of common concern to both academia and industry.

[0004] I. Current Status of Microgrid Energy Management Technology The fundamental goal of a microgrid energy management system is to optimize the scheduling of various distributed resources to achieve optimal overall system performance, while meeting user electricity demand and comprehensively considering factors such as economy, environmental protection, reliability, and equipment lifespan. A typical EMS architecture usually comprises three layers: a prediction layer (e.g., photovoltaic power generation prediction, load prediction), an optimization layer (formulating day-ahead / real-time scheduling plans), and an execution layer (controller issuing commands). In recent years, with the development of artificial intelligence, big data, and advanced control theory, microgrid EMS has evolved from static rule-based control to dynamic optimization decision-making.

[0005] Early energy management often employed fixed schedules or simple threshold logic for control, such as "charging during the day and discharging at night" or "starting diesel generators when SOC is below 20%". While simple to implement, these methods lacked adaptability and could not cope with complex and changing operating environments. Subsequently, researchers introduced mathematical optimization methods such as Model Predictive Control (MPC), Dynamic Programming (DP), and Stochastic Optimization. By establishing system models and solving for the optimal scheduling strategy in the rolling time domain, they significantly improved control accuracy and economy. For example, Zhang et al. proposed a microgrid scheduling framework based on two-stage stochastic MPC, effectively handling the uncertainties of photovoltaic power and electricity prices; Liu et al. constructed a mixed-integer linear programming (MILP) model, achieving synergistic optimization of multiple types of energy storage and controllable loads.

[0006] While the aforementioned methods have achieved good theoretical results, they still face many challenges in practical applications. First, most optimization models assume energy storage devices are ideal components, neglecting the impact of their aging process on available capacity, efficiency, and safety boundaries, leading to scheduling results deviating from actual operating conditions. Second, traditional EMS often separates "energy scheduling" from "health management," lacking explicit modeling and feedback control of battery degradation mechanisms, resulting in overuse or resource waste. Finally, existing systems generally rely on offline calibration parameters, have weak online update capabilities, and struggle to adapt to the nonlinear characteristics of battery performance evolution over time.

[0007] II. Research Progress in Battery State of Health (SOH) Estimation Technology To achieve refined energy management, it is essential to accurately grasp the health status of energy storage batteries. State of Health (SOH) is typically defined as the ratio of current maximum usable capacity to its initial nominal capacity, but can also be extended to include indicators such as internal resistance growth rate and power retention rate. SOH is not only the basis for measuring the remaining useful life (RUL) of a battery, but also a crucial basis for developing reasonable charge and discharge strategies.

[0008] Currently, SOH estimation methods can be mainly divided into three categories: empirical model-based methods, equivalent circuit model-based methods, and data-driven methods.

[0009] 2.1 Empirical Model-Based Approach These methods utilize laboratory aging test data to fit empirical formulas for capacity decay, commonly including Arrhenius models, power-law models, and double-exponential models. For example, Dubarry et al., based on extensive aging tests of commercial lithium batteries, proposed a fusion temperature, C...rate A semi-empirical decay model based on the deep discharge (DOD) factor can be used to predict capacity loss trends under different operating conditions. Similarly, Yang et al. constructed a multi-stress coupled aging model that quantifies the influence weights of factors such as cycle number, average SOC, and temperature rise on SOH.

[0010] Although empirical models have clear physical meaning and are computationally simple, their generalization ability is poor, and they are only applicable to specific battery models and preset operating conditions. Once operating conditions change (such as the addition of fast charging scenarios), the model prediction error increases significantly, making it difficult to meet the needs of online monitoring.

[0011] 2.2 Method based on equivalent circuit model (ECM) The equivalent circuit model simulates the dynamic voltage response characteristics of the battery through a resistor-capacitor network. Combined with state estimation algorithms such as Kalman filter (KF) or particle filter (PF), it achieves joint estimation of SOC and SOH. Typical examples include the Thevenin model, the RC parallel model, and the PNGV model. Plett first proposed the Extended Kalman Filter (EKF) for lithium-ion battery SOC estimation and further developed it into a dual-filter structure to simultaneously track SOH changes. Subsequently, many scholars improved this framework; for example, Wang et al. used the Unscented Kalman Filter (UKF) to improve the convergence speed and accuracy of nonlinear systems.

[0012] However, the ECM method relies heavily on accurate model parameter identification, and these parameters themselves slowly evolve during the aging process, leading to the "parameter drift" problem. Furthermore, the filtering algorithm is sensitive to initial values ​​and susceptible to noise interference, exhibiting insufficient stability, especially under complex operating conditions such as frequent microgrid start-ups and shutdowns and sudden load changes.

[0013] 2.3 Data-Driven Approach In recent years, with the rise of machine learning and deep learning technologies, data-driven methods have become the new mainstream for SOH estimation. These methods do not require building complex physical models; instead, they directly learn the input-output mapping relationship from historical operational data. Commonly used algorithms include Support Vector Regression (SVR), Artificial Neural Networks (ANN), Long Short-Term Memory Networks (LSTM), and Graph Neural Networks (GNN).

[0014] For example, Li et al. used LSTM networks to extract time-series features from charge-discharge curves, achieving high-precision prediction of single-cell state of health (SOH). Severson et al.'s research published in Nature Energy showed that a random forest model accurately predicting full-lifecycle capacity degradation can be trained using only the data from the first 100 cycles, demonstrating the enormous potential of data-driven methods. Further research has attempted to integrate multiple sensor signals (voltage, current, temperature, and expansion force) for multimodal health assessment, further enhancing robustness.

[0015] While data-driven methods exhibit strong fitting capabilities, they also suffer from problems such as strong "black box" characteristics, poor interpretability, and high dependence on training samples. Especially in application scenarios like microgrids with small sample sizes and long cycles, a lack of sufficient fault data can lead to model overfitting or extrapolation failure.

[0016] III. Major Defects and Shortcomings of Existing Technologies Despite the significant research achievements made both domestically and internationally in microgrid energy management and battery health management, the existing technological system still faces the following two challenges: (1) The SOH estimation model is difficult to adapt to the complex operating conditions of microgrids. Energy storage systems in microgrids typically face diverse operational tasks, including intraday high-frequency regulation, seasonal peak shaving and valley filling, and emergency backup power supply, resulting in complex current profiles and temperature fluctuations in the batteries. Traditional SOH estimation models are mostly trained based on constant current or standard cycle test data, making it difficult to accurately characterize the multi-scale, non-stationary aging behavior in real microgrid environments.

[0017] For example, some ECM-based methods perform well under laboratory conditions, but parameter mismatches occur during field deployment due to changes in contact impedance and sensor drift. Furthermore, data-driven models trained using only historical data from a single power plant are difficult to transfer to microgrid projects with different geographical locations or climates, posing a serious risk of overfitting. In addition, existing methods generally neglect the inconsistency evolution between individual cells within a battery cluster, providing only the average SOH value for the entire cluster, which cannot guide equalization control and fault early warning.

[0018] (2) Energy management and health management are disconnected and lack a closed-loop coordination mechanism. The current management of microgrid energy storage systems suffers from a systemic flaw: the lack of a real-time, precise, and closed-loop coordination mechanism between the energy management system (EMS)'s optimization decisions and the actual, dynamically changing state of health (SOH) of the batteries. This results in the system operating in a "blind" state. Specifically, when formulating scheduling plans, the EMS typically treats energy storage as a "black box" device with a fixed rated capacity. Its optimization model fails to incorporate SOH, which reflects battery aging, as a core dynamic constraint variable, and also lacks the ability to perceive and provide feedback on the real-time aging rate of the batteries. This directly leads to a severe mismatch between "planning" and "capacity": as batteries age, their actual usable capacity continuously declines, but the EMS still schedules based on initial, excessively large capacity parameters. This not only leads to increased curtailment rates and operational plan failures but also, due to inappropriate charge and discharge commands (such as deep cycling of aged battery packs), unintentionally accelerates the battery aging process itself.

[0019] In summary, the open-loop management model of "inaccurate sensing, unknown decision-making, and uncoordinated execution" in the current microgrid environment prevents the microgrid from building an intelligent operation and maintenance closed loop of "sensing status → optimizing decision-making → executing verification → learning and updating". Ultimately, it falls into a vicious cycle of "blind use accelerates aging, and accelerated aging leads to a decline in operational efficiency", which seriously restricts the economy, reliability and full life cycle value of energy storage systems.

[0020] For example, patent CN120601494B describes a dynamic charging and discharging strategy collaborative optimization method for extending the lifespan of energy storage systems. The method includes: inputting historical electricity price time series data based on a pre-defined long short-term memory network model to generate an electricity price prediction sequence; constructing a state vector, which includes the internal state of the energy storage system at the current time t and the electricity price prediction sequence; inputting the state vector into a pre-trained deep reinforcement learning agent, which outputs a specific action selected from a predefined set of discrete charging and discharging actions based on the input state vector; and converting the output specific action into charging and discharging power control commands and sending them to the energy storage converter for execution. This invention, through deep reinforcement learning combined with accurate electricity price prediction and dynamic battery health modeling, achieves collaborative optimization of charging and discharging strategies for energy storage systems in complex market environments. The following problems exist: three key dimensions are missing: First, there is a lack of a multi-stress coupling aging model, which fails to fully characterize the nonlinear impact of factors such as charge / discharge depth, temperature, and rate on battery degradation; second, there is a lack of a "sensing-decision" closed-loop mechanism, which fails to embed real-time SOH as a dynamic constraint into energy scheduling, resulting in a disconnect between planned and actual capabilities; third, there is a lack of full life-cycle economic considerations, with optimization objectives focusing only on short-term operating costs and failing to introduce degradation costs and failure risk costs derived from SOH, making it difficult to support the full life-cycle value management of energy storage assets. Summary of the Invention

[0021] To address this problem, this invention proposes a microgrid energy management method based on dynamic sensing of the health status of energy storage batteries. This method solves the technical problem that the lack of dynamic sensing and closed-loop feedback mechanisms for the health status of energy storage batteries leads to a severe mismatch between scheduling plans and actual available capacity, resulting in decreased system operating economy and accelerated battery life degradation.

[0022] The technical solution adopted by the present invention to solve the above problems is as follows: The present invention includes the following steps: S1. Construct a dynamic sensing model of battery state of health (SOH) that integrates the depth of charge and discharge (DOD) factor of energy storage lithium batteries, and take the raw data uploaded by the battery management system (BMS) that is matched with the energy storage battery cluster as input; S2. Construct the planning layer. Establish an objective function with the goal of minimizing the operating cost of the energy storage system throughout its entire life cycle. Solve the function using the Robust Mixed Integer Linear Programming (RMILP) method. Inputs include historical load and photovoltaic power output sequences, time-of-use electricity price curves, and future SOH prediction curves generated by the SOH dynamic model. The optimized output is the optimal rated capacity and power level of the energy storage system. S3. Construct a real-time layer, using the output of the planning layer as the initial boundary and continuously incorporating the latest measured data. Employ a collaborative architecture of Model Predictive Control (MPC) and Particle Swarm Optimization (PSO). MPC calls the dynamic constraint set driven in real time, using the daily operating cost of the system as the optimization objective, to generate the optimal energy storage power command sequence within the current rolling window.

[0023] Furthermore, the dynamic sensing model for battery state of health (SOH) is as follows:

[0024]

[0025]

[0026] in , are the coefficients to be fitted, and the function satisfies: when No loop. (Minimum value), when Full-depth loop, (Maximum value); These are the coefficients to be fitted. Represents irreversible lithium deposition. This represents the loss of active lithium due to cracks. Loss of active lithium due to surface SEI growth.

[0027] Furthermore, S1 includes the following steps: S11. Perform data filtering and outlier removal within a sliding time window (15 minutes in length). Use the 3σ criterion to identify and remove temperature, C-rate, and DOD samples that significantly deviate from the normal operating range, and complete the time alignment of multi-source signals. S12. Based on the capacity loss model of the energy storage lithium battery established in this invention, the SOH information of the energy storage unit is updated every 15 minutes, where SOH = ( C rate - C loss ) / C rate .

[0028] Among them, among them, C rate Rated battery capacity; C loss Battery capacity loss.

[0029] Furthermore, S2 aims to minimize the operating cost of the energy storage system throughout its entire lifecycle. The objective function is:

[0030] In the formula, Present value of total life cycle cost, Where y is the initial investment cost, y is the year the energy storage equipment is put into operation, and Y is the total operating years. For the degradation cost in year y, For the failure cost in year y, denoted as , where is the annual fixed maintenance cost, and r is the discount rate.

[0031] Furthermore, the planning layer operates on a 24-hour timescale.

[0032] Furthermore, the real-time layer operates within a rolling time window, typically 15 minutes.

[0033] Furthermore, the latest measured data includes the current SOC, grid interconnection power, real-time photovoltaic power output, and the latest values ​​pushed by the SOH dynamic sensing module, as well as short-term forecast information for the next 15 minutes.

[0034] Furthermore, the dynamic constraint set includes the available capacity that shrinks as SOH decays, the maximum charge and discharge power limit that is reduced proportionally, and the upper and lower limits of safe operation of SOC that are adaptively adjusted according to the degree of aging.

[0035] Furthermore, S3 includes the following steps: S31. Objective function: Minimize the total daily operating cost.

[0036] in,

[0037] In the formula, The total operating cost for the day is (in yuan), and t is the time period index, dividing the day into N time periods (24 hours). The grid electricity price for time period t is (yuan / kWh, with a positive value indicating the purchase price). It is the power exchanged with the grid during time period t (kW, positive values ​​represent electricity purchased from the grid, and negative values ​​represent electricity sold to the grid). The daily degradation cost (in yuan) of energy storage equipment. Cost of daily failure risk for energy storage equipment (RMB); S32. Constructing power balance constraints for microgrid systems:

[0038] In the formula, For photovoltaic power generation, Power generation of micro gas turbines, For energy storage discharge power, To purchase power from the grid, For load power, For energy storage charging power, The power output sold to the power grid; S33. Construct an energy storage model that includes energy storage lifetime and fault conditions.

[0039] Specifically, the battery's maximum capacity parameter is transformed into a function that varies with time: Define dynamic rated capacity:

[0040] Define dynamic maximum power:

[0041]

[0042] Charge and discharge power constraints:

[0043]

[0044] Charge and discharge are mutually exclusive: , SOC range:

[0045] DOD calculation: .

[0046] in, Rated energy; This is the rated maximum charging power; This represents the actual maximum discharge power. Maximum allowable charging power; The charging power at time t; for ; Maximum permissible discharge power; The beneficial effects of this invention are: 1. This invention is the first to embed a dynamic evolution model of battery state of health (SOH) into the entire process of energy storage capacity optimization, fully considering the impact of key stress factors such as depth of charge / discharge (DOD), cycle count, and temperature on capacity decay and internal resistance growth. Compared with traditional static configuration methods that ignore aging characteristics, it can more accurately predict the actual service life of the energy storage system, avoid premature retirement due to overuse, and significantly improve the lifespan synergy among power sources, loads, and energy storage in a microgrid.

[0047] 2. This invention breaks through the limitations of traditional energy storage capacity configuration that prioritizes economy over lifespan and static performance over dynamic performance. It constructs a complete technical system that integrates aging modeling, intelligent control, and multi-objective optimization. This not only improves the operational reliability and economy of microgrid systems, but also provides an advanced solution for the full life-cycle management of energy storage assets in new power systems.

[0048] 3. This invention embeds the acceleration mechanism of DOD on active lithium loss (LLI) and active material loss (LAM) into the estimation framework in an explicit mathematical form, breaking through the limitation of traditional models that are only applicable to standard cyclic conditions.

[0049] 4. This invention embeds real-time dynamically updated SOH information as a core constraint into the energy management optimization model. By linking the battery's "current usable capacity" with the scheduling's "planned usage" in real time, it completely changes the traditional EMS's extensive mode of treating energy storage as a fixed-capacity "black box". It solves problems such as plan execution failure and increased curtailment rate caused by aging, and realizes a refined management closed loop of "perceptible status and optimizable decision-making".

[0050] 5. This invention adopts a two-stage optimization architecture of "planning layer + real-time layer". It considers the total cost of the entire life cycle in the day-ahead planning and dynamically balances the revenue from purchasing and selling electricity with the cost of battery aging in the real-time scheduling. This avoids excessive damage to battery life in pursuit of short-term economic benefits, thereby maximizing the economic benefits of energy storage assets throughout their entire life cycle while ensuring system safety. Attached Figure Description

[0051] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0052] Specific implementation method one: Combining Figure 1 This embodiment describes S1, which involves constructing a dynamic sensing model of battery state of health (SOH) that integrates the depth of charge and discharge (DOD) factor of energy storage lithium batteries, and using the raw data uploaded by the battery management system (BMS) that accompanies the energy storage battery cluster as input. First, a dynamic sensing model for State of Health (SOH) is constructed. This model considers the full-lifetime loss mechanism of lithium iron phosphate batteries, including the loss of active lithium ions and active materials. Furthermore, the model incorporates battery operating temperature, charge / discharge current rate (r-rate), and depth of charge / discharge (DOD) during its development, making it more adaptable to the complex operating conditions of microgrids compared to existing models. The model uses raw data uploaded by the BMS as input, first performing data filtering and outlier removal within a sliding time window (15 minutes). The 3σ criterion is used to identify and remove temperatures and values ​​that significantly deviate from the normal operating range. rate The DOD samples were compared with those obtained in this paper, and time alignment of the multi-source signals was completed. Subsequently, the SOH information of the energy storage unit was updated every 15 minutes based on the capacity loss model of the energy storage lithium battery established in this invention. Wherein, SOH = ( C rate - C loss ) / C rate .

[0053] The capacity loss estimation model for lithium-ion batteries adapted to the complex operating conditions of microgrids is as follows: Establish a capacity loss model for energy storage lithium batteries (1) in: This represents irreversible lithium deposition (related to stress). This represents the loss of active lithium due to cracks. Loss of active lithium due to surface SEI growth.

[0054] In formula (1) LAM liNE and LLI a,crk The mathematical model is as follows: (2) (3) Considering that different real-world microgrid scenarios often lead to energy storage batteries operating in different charge / discharge depths (DODs), it is necessary to incorporate the impact of battery DOD into the above model. The mechanism of capacity loss due to DOD in energy storage lithium batteries is as follows: (1) A larger DOD will cause more lithium ions to be inserted into the negative electrode during the charging process, making the negative electrode potential more "negative". The lower negative electrode potential significantly enhances the driving force of solvent reduction reaction, thereby increasing the effective volume concentration of the solid / liquid interface reaction on the surface of the negative electrode active ions, amplifying the lithium ion concentration gradient that promotes the formation of SEI film, and finally the LLI increases nonlinearly with the increase of DOD.

[0055] (2) The larger the DOD (Depth of Discharge), the greater the change in lithium-ion concentration during charging and discharging. The larger the concentration fluctuation, the more likely it is to cause a series of physicochemical effects, ultimately leading to accelerated battery aging. According to the generalized Hooke's law, the strain of the material... With lithium concentration Proportional, that is stress And with strain Proportional, that is The uneven distribution of lithium concentration within the particles leads to a mismatch in stress distribution, especially radial stress. With tangential stress The increased difference in stress creates a larger stress gradient. This significantly promotes the initiation of fatigue cracks and exacerbates the problem. When cracks form, new unpassivated active material is exposed on the electrode surface. These fresh surfaces react with the electrolyte to form a new solid electrolyte interphase (SEI) film, leading to a further increase in irreversible capacity loss (LLI). Therefore, considering that DOD directly affects processes such as lithium concentration changes, stress evolution, crack initiation, and SEI regeneration, ultimately... It increases nonlinearly with the increase of DOD.

[0056] A capacity loss estimation model considering the depth of charge and discharge (DOD) of energy storage lithium batteries is proposed.

[0057] First, we define a comprehensive correction function with DOD as the variable to describe the nonlinear effect of DOD on battery capacity loss: (4) in , are the coefficients to be fitted.

[0058] This function satisfies: when No loop. (Minimum value) when Full-depth loop, (Maximum value) With the addition of the DOD factor, the cracking caused by active material particles... , The corrections for LLI on the surface of active particles are as follows: (5) (6) Thus, the capacity loss model applicable to lithium-ion batteries with different charge / discharge rates, temperatures, and DODs is completed, namely formulas (4), (5), and (6), where... These are the coefficients to be fitted.

[0059] The implementation of energy management is divided into two closely coupled phases.

[0060] The first stage is the planning layer, operating on a 24-hour timescale. Its inputs include historical load and photovoltaic power output sequences, time-of-use electricity price curves, and future SOH prediction curves generated by the SOH dynamic model. This layer employs a robust mixed-integer linear programming (RMILP) method for solving the problem. The objective function explicitly incorporates a full lifecycle degradation cost term on top of the conventional operating cost. This cost is determined by the current SOH state, planned charge / discharge power, and corresponding DOD level. All constraints are dynamically adjusted based on the SOH evolution pattern, such as the maximum allowable number of cycles and the daily equivalent cycle depth threshold. The optimized output is the optimal rated capacity and power level of the energy storage system.

[0061] In the worst-case scenario considering the uncertainty of wind and solar load, a mixed-integer linear programming problem is solved to minimize the total investment cost, determine the optimal energy storage capacity, and output a robust day-ahead scheduling plan.

[0062] With the goal of minimizing the total lifecycle operating cost of the energy storage system, the objective function is: (7) In the formula, Present value of total life cycle cost, Where y is the initial investment cost, y is the year the energy storage equipment is put into operation, and Y is the total operating years. For the degradation cost in year y, For the failure cost in year y, denoted as , where is the annual fixed maintenance cost, and r is the discount rate.

[0063] The second stage is the real-time layer, operating within a rolling time window (typically 15 minutes). It uses the planning layer output as its initial boundary and continuously receives the latest measured data, including the current SOC, grid tie-line power, real-time photovoltaic output, and the latest values ​​pushed by the SOH dynamic sensing module, as well as short-term forecasts for the next 15 minutes. This layer employs a collaborative architecture of Model Predictive Control (MPC) and Particle Swarm Optimization (PSO): the MPC's internal prediction model no longer uses fixed rated parameters but instead invokes a dynamic set of constraints in real-time—including available capacity shrinking with SOH decay, proportionally reduced maximum charge / discharge power limits, and adaptively adjusted upper and lower limits for safe SOC operation based on aging. The PSO acts as an outer optimizer, dynamically adjusting the MPC's objective function weight matrix to generate the optimal energy storage power command sequence within the current rolling window, while satisfying power balance and equipment constraints, with the system's daily operating cost as the optimization objective.

[0064] Building upon the robust plan and energy storage configuration provided in the first phase, model predictive control and real-time electricity price / output signals are introduced for rolling optimization to dynamically correct scheduling instructions. The second phase employs a particle swarm optimization (PSO) algorithm for global search, combined with a CPLEX solver to handle subproblems, improving convergence efficiency.

[0065] First, with the goal of minimizing the total daily operating cost, the objective function is: (8) in,

[0066] In the formula, The total operating cost for the day is (in yuan), and t is the time period index, dividing the day into N time periods (24 hours). The grid electricity price for time period t is (yuan / kWh, with a positive value indicating the purchase price). It is the power exchanged with the grid during time period t (kW, positive values ​​represent electricity purchased from the grid, and negative values ​​represent electricity sold to the grid). The daily degradation cost (in yuan) of energy storage equipment. Cost of daily failure risk for energy storage equipment (in yuan).

[0067] The power balance constraints of a microgrid system are: (9) In the formula, For photovoltaic power generation, Power generation of micro gas turbines, For energy storage discharge power, To purchase power from the grid, For load power, For energy storage charging power, The amount of electricity sold to the power grid.

[0068] The innovation of this patent is the establishment of an energy storage model that takes into account energy storage lifespan and fault conditions: (10) Specifically, the battery's maximum capacity parameter is transformed into a function that varies with time: Define dynamic rated capacity:

[0069] Define dynamic maximum power:

[0070]

[0071] Power constraints:

[0072]

[0073] Charge and discharge are mutually exclusive:

[0074] SOC range:

[0075] DOD calculation:

[0076] The above two-stage information interaction (planning results provide the hardware foundation for operation, and the economic potential feedback from operation can guide planning in reverse) achieves the goal of "maximizing economic benefits while ensuring system safety in the worst-case scenario." Utilizing the proposed real-time SOH estimation method for energy storage batteries, embedded in the energy storage battery capacity planning and implementation scheduling, the economic benefits obtained compared to existing technical solutions are more realistic, and it can more effectively guide the operation of microgrid systems.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent substitutions, and improvements made to the above embodiments without departing from the scope of the present invention, based on the technical essence of the present invention and within the spirit and principles of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A microgrid energy management method based on dynamic sensing of the health status of energy storage batteries, characterized in that, Includes the following steps: S1. Construct a dynamic sensing model of battery health status (SOH) that integrates the depth of charge and discharge (DOD) factor of energy storage lithium batteries, and use the raw data uploaded by the battery management system (BMS) that is matched with the energy storage battery cluster as input. S2. Construct the planning layer. Establish the objective function with the goal of minimizing the operating cost of the energy storage system throughout its entire life cycle. Solve the objective function using the robust mixed-integer linear programming (RMILP) method. Inputs include historical load and photovoltaic power output sequences, time-of-use electricity price curves, and future SOH prediction curves generated by the SOH dynamic model. The optimal output is the optimal rated capacity and power level of the energy storage system. S3. Construct a real-time layer, using the output of the planning layer as the initial boundary and continuously incorporating the latest measured data. Employ a collaborative architecture of Model Predictive Control (MPC) and Particle Swarm Optimization (PSO). MPC calls the dynamic constraint set in real time, using the daily operating cost of the system as the optimization objective, to generate the optimal energy storage power command sequence.

2. The microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 1, characterized in that, The battery health status (SOH) dynamic sensing model is as follows: in , are the coefficients to be fitted, and the function satisfies: when No loop. Minimum value, when Full-depth loop, Maximum value; These are the coefficients to be fitted. Represents irreversible lithium deposition. This represents the loss of active lithium due to cracks. Loss of active lithium due to surface SEI growth.

3. The microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 2, characterized in that, S1 includes the following steps: S11. Implement a sliding time window of 15 minutes for data filtering and outlier removal. Use the 3σ criterion to identify and remove temperatures and values ​​that significantly deviate from the normal operating range. rate With DOD samples, and complete time alignment of multi-source signals; S12. Based on the capacity loss model of the energy storage lithium battery established in this invention, the SOH information of the energy storage unit is updated every 15 minutes, where SOH = ( C rate - C loss ) / C rate ,in, C rate Rated battery capacity; C loss Battery capacity loss.

4. The microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 3, characterized in that, In S2, the objective is to minimize the operating cost of the energy storage system throughout its entire lifecycle. The objective function is: In the formula, Present value of total life cycle cost, Where y is the initial investment cost, y is the year the energy storage equipment is put into operation, and Y is the total operating years. For the degradation cost in year y, For the failure cost in year y, denoted as , where is the annual fixed maintenance cost, and r is the discount rate.

5. A microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 4, characterized in that, The planning layer operates on a 24-hour timescale.

6. The microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 1, characterized in that, The real-time layer operates within a rolling time window, typically 15 minutes.

7. The microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 1, characterized in that, The latest measured data includes the current SOC, grid interconnection power, real-time photovoltaic power output, and the latest values ​​pushed by the SOH dynamic sensing module, as well as short-term forecast information for the next 15 minutes.

8. The microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 1, characterized in that, The dynamic constraint set includes the available capacity that shrinks as SOH decays, the maximum charge and discharge power limit that is reduced proportionally, and the upper and lower limits of safe SOC operation that are adaptively adjusted according to the degree of aging.

9. A microgrid energy management method based on dynamic sensing of the health status of energy storage batteries according to claim 1, characterized in that, S3 includes the following steps: S31. With the objective of minimizing the total daily operating cost, the objective function is: in, In the formula, The total operating cost for the day is given by t, which is the time period index. The day is divided into N time periods. Let t be the grid electricity price. It refers to the interaction power with the power grid during time period t. The daily degradation cost of energy storage equipment, Cost of daily failure risk for energy storage equipment; S32. Constructing power balance constraints for microgrid systems: In the formula, For photovoltaic power generation, Power generation of micro gas turbines, For energy storage discharge power, To purchase power from the grid, For load power, For energy storage charging power, The power output sold to the power grid; S33. Construct an energy storage model that includes energy storage lifetime and fault conditions. Specifically, the battery's maximum capacity parameter is transformed into a function that varies with time: Define dynamic rated capacity: Define dynamic maximum power: Power constraints: Charge and discharge are mutually exclusive: This indicates that the battery cannot be in both "charging" and "discharging" states at the same time. Battery fully charged If the battery is not charging, Battery fully discharged The battery is not discharging; These are the battery charging and discharging status indicators; SOC range: DOD calculation: ; Rated energy; This is the rated maximum charging power; This refers to the actual maximum charging power. Let be the charging power at time t; for ; Maximum permissible discharge power; This represents the actual maximum discharge power.

Images