An energy storage charging and discharging strategy optimization evaluation method based on big data analysis

Abstract

This invention relates to the field of big data analytics, and discloses a method for optimizing and evaluating energy storage charging and discharging strategies based on big data analytics. The method includes: constructing a quantitative rule base that maps the current state of the energy storage system to opportunity costs; defining a behavioral primitive library containing standardized charging, discharging, and stationary actions; acquiring the system state in real time and calculating the opportunity cost of each behavioral primitive. This invention avoids reliance on complex prediction models through the collaborative mapping of behavioral primitives and opportunity costs, enabling real-time decision-making in high-frequency market environments; combining probabilistic selection mechanisms and state awareness to naturally guide system behavior towards safety boundaries and optimize long-term benefits; and utilizing the synergy of historical data quantile analysis and diagnostic behavioral primitives to achieve hardware function reuse and implicit state awareness, significantly improving the accuracy of decision-making throughout the entire lifecycle.

Description

Technical Field

[0001] This invention relates to a method for optimizing and evaluating energy storage charging and discharging strategies based on big data analysis, belonging to the field of big data analysis technology. Background Technology

[0002] Current mainstream solutions rely on a combination of electricity price and load forecasting models with multi-objective optimization algorithms to generate charging and discharging plans by establishing battery degradation models. This approach is effective to a certain extent in stable market environments. However, in the high-frequency fluctuation scenario of the electricity spot market, its deep-seated contradictions gradually emerge: when electricity prices are refreshed every 5-15 minutes and exhibit irregular peaks / troughs, traditional forecasting-optimization methods cannot respond to market changes in real time due to computational delays, causing energy storage systems to miss arbitrage opportunities. More fundamentally, this approach forcibly couples real-time financial costs and lagging physical loss costs to the same optimization function. Due to the differences in their quantitative dimensions and timeliness, the decision-making logic is systematically fragile.

[0003] To improve real-time performance, the industry has attempted to simplify prediction models or adopt reinforcement learning, but the bias in training data has exacerbated decision-making risks. Another approach introduces complex battery health models, which in turn leads to a surge in computational burden. Existing technologies are still plagued by three core contradictions: 1. Computational latency causes decisions to lag behind market pace; 2. It is difficult to scientifically balance financial costs and physical losses in a unified optimization function; 3. Refined models drive up computing power requirements, making it difficult to match with the deployment of affordable hardware.

[0004] With increasing volatility in the electricity spot market and the expansion of energy storage, the aforementioned contradictions have evolved into a common bottleneck restricting the industry's development. Therefore, how to construct a decision-making mechanism that does not require a predictive model, can respond to market fluctuations in real time, and endogenously coordinates economic arbitrage and battery safety has become the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a method for optimizing and evaluating energy storage charging and discharging strategies based on big data analysis. Its main purpose is to solve the problems of real-time decision-making delay, economic efficiency and battery safety in high-frequency fluctuation scenarios of the electricity spot market.

[0006] To achieve the above objectives, this invention provides a method for optimizing and evaluating energy storage charging and discharging strategies based on big data analysis, comprising the following steps:

[0007] Step a, construct a behavior primitive library, which contains standard action units for charging, discharging and stationary scenarios of energy storage systems;

[0008] Step b: Construct an opportunity cost quantification rule base. The rule base establishes quantitative rules that map the current operating state of the energy storage system to the real-time opportunity cost of the corresponding behavioral primitives based on historical electricity price data and historical system state of charge data. The current operating state includes the historical percentile level of the current real-time electricity price of the power grid and the corresponding interval of the current system state of charge.

[0009] Step c: Obtain the current operating status data of the energy storage system in real time;

[0010] Step d: For each executable behavioral primitive in the behavioral primitive library, calculate the current opportunity cost of each executable behavioral primitive based on the current running state data and using the opportunity cost quantification rule library;

[0011] Step e: Based on the current opportunity cost of each executable action primitive, select one action primitive as the execution action for the next period in a probabilistic manner, wherein the probability of any action primitive being selected is inversely proportional to its current opportunity cost.

[0012] Step f involves periodically executing diagnostic behavioral primitives to obtain parameters characterizing the internal health state of the battery, and adaptively calibrating the opportunity cost quantification rule base based on the health state parameters.

[0013] Preferably, in the opportunity cost quantification rule base, when the current system state of charge is in an extremely high or extremely low range determined based on battery safety characteristics, a penalty factor is applied to the opportunity cost of charging and discharging behavior primitives. The penalty factor is determined according to the degree to which the system state of charge deviates from the safe range, so as to guide the behavior of the energy storage system towards the battery safe operating range.

[0014] Preferably, the probabilistic selection method is: calculate the first element in the behavioral primitive library. The probability of a behavior primitive being selected probability The calculation formula is: ,in, Indicates the first The current opportunity cost of an individual behavioral element. This represents the total number of executable behavioral primitives in the behavioral primitive library.

[0015] Preferably, the historical percentile level of the current real-time electricity price is determined by comparing the current real-time electricity price with multiple percentiles obtained from the statistics of historical electricity price data. These multiple percentiles are used to characterize the position of the electricity price in the historical distribution.

[0016] Preferably, in step f, the diagnostic behavior unit includes a standardized micro-current charge-discharge process, which is executed by the power electronics of the energy storage system itself, and the current decay curve of the battery is recorded with high precision during the process; the internal health status parameters of the battery are obtained by fitting the current decay curve to derive the time constant characterizing the electrochemical impedance of the battery, so as to realize online monitoring of the battery health status.

[0017] Preferably, after step d and before step e, the method further includes: determining whether the energy storage system has entered a dormant state, wherein the dormant state refers to the time during which the dormant behavior primitive is continuously selected exceeds a set duration threshold; when the energy storage system has entered a dormant state, forcibly selecting and executing a wake-up probe primitive in the behavior primitive library with a preset low probability, wherein the wake-up probe primitive is used to change the system state of charge of the energy storage system with a small power, so as to prompt the energy storage system to evaluate a new set of opportunity costs for behavior primitives in the next decision cycle.

[0018] Preferably, the wake-up detection element includes charging or discharging operations with a preset very small current.

[0019] Preferably, the method further includes: after selecting the behavior primitive as the execution action for the next time period in step e, before issuing the execution command to the power control unit of the energy storage system, performing physical limit boundary arbitration; the physical limit boundary arbitration performs virtual simulation of the behavior primitive to be executed based on the fixed absolute safety boundary rule library, and determines whether the estimated voltage and estimated temperature rise of the battery after executing the behavior primitive will touch the absolute safety boundary; when it is determined that the absolute safety boundary has been touched, the original execution command is rejected and replaced with a static primitive in the behavior primitive library.

[0020] Preferably, the physical limit boundary arbitration further includes: during the execution of the behavioral primitive or at the moment of completion of the execution, using the battery voltage and current change data measured before and after the execution of the behavioral primitive, estimating the macroscopic dynamic internal resistance of the battery by the voltage drop method; and performing virtual simulation to calculate the predicted battery voltage and predicted temperature rise based on the macroscopic dynamic internal resistance.

[0021] Preferably, before step d is executed, the method further includes: evaluating the real-time effectiveness of the opportunity cost quantification rule base through a physical characteristic deviation preprocessing gateway. The physical characteristic deviation preprocessing gateway compares the real-time aging drift characteristics of the battery with the reference internal resistance value when the opportunity cost quantification rule base was established. When the percentage deviation between the aging drift characteristics and the reference internal resistance value exceeds a set threshold of 20%, a penalty factor is applied to the opportunity cost of all high-power charging and discharging behavior primitives in the opportunity cost quantification rule base. The magnitude of the penalty factor is positively correlated with the percentage deviation.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. This invention abandons the need for predictive modeling of complex future states, and instead decouples energy storage operations into finite, standardized behavioral primitives such as charging, discharging, and resting. By constructing a rule base that maps current definite states such as state of charge intervals and real-time electricity price levels to opportunity costs, at each decision moment, the opportunity cost of each primitive can be evaluated in parallel based only on the most readily available real-time data. This mechanism makes the decision-making process no longer constrained by the accuracy and computational delay of the predictive model, and naturally endows it with the ability to handle high-frequency fluctuations in the electricity market. The decision response speed is only constrained by basic state perception and simple rule matching, and the system robustness is significantly enhanced.

[0024] 2. Based on opportunity cost assessment, a probabilistic selection mechanism is adopted, which not only smooths out the risk of overreaction that may be caused by a single optimal solution, but also automatically applies a penalty factor to the charging / discharging primitives through a rule base, such as when the state of charge is in a preset extreme high or low range. This allows the system behavior to spontaneously converge to the safe operating range of the battery without the constraints of a complex physical model. This guidance method based on economic incentives rather than physical coercion, combined with the exploratory nature of probabilistic selection, enables the system to avoid physical risks while retaining the flexibility to capture market arbitrage opportunities within the safety margin, which helps to achieve a balance between asset protection and economic benefits.

[0025] 3. This invention utilizes statistical quantiles of historical electricity price data, such as high, medium, and low intervals, as the basis for quantifying opportunity cost, avoiding complex model fitting. Furthermore, by introducing specific diagnostic behavioral primitives, such as standardized microcurrent charge-discharge sequences, which are triggered under low system load or specific preset conditions, the core value of these primitives is not directly profiting from charge-discharge, but rather that the electrochemical response during execution, such as the current decay curve during the constant voltage phase, can be used to invert implicit health state parameters of the battery, such as the time constant. This inversion result is dynamically fed back to the opportunity cost rule base in real time, adjusting the penalty intensity. This ensures that the decision-making process is not only based on macroscopic power consumption but also incorporates microscopic health information. This mechanism fully utilizes the potential of existing power converters, achieving cross-boundary function reuse with execution-awareness, improving the accuracy of full lifecycle decision-making and asset protection capabilities, and forming asymmetric synergy with the core opportunity cost decision-making mechanism by adding a physical anchoring layer. At the input end, the physical characteristic deviation gateway dynamically judges the effectiveness of the rule base by estimating the battery's macroscopic dynamic internal resistance in real time and comparing it with the benchmark value when the rule base is built. When significant aging deviation is detected, the gateway will take action. In case of a problem, the gateway automatically applies dynamic penalties to the opportunity cost of high-power primitives, gently guiding the system to adapt to changes in the battery's physical state and avoiding economic degradation or risks caused by rule failure. At the output end, the physical limit boundary decision arbitrator performs a simplified virtual deduction of the primitive instructions to be executed based on fixed absolute safety boundary rules such as voltage and temperature limits. Once it is predicted that the boundary will be touched, the veto power is exercised and replaced with a conservative safety action. This two-layer mechanism provides a physical world anchor for the core high-cognitive decision-making logic, ensuring the reliability and industrial applicability of the solution throughout the battery's entire life cycle and reducing computational overhead.

[0026] 4. To address decision-making deadlocks caused by prolonged stable markets, this invention introduces a dormant-to-wake opportunity detection mechanism. When continuous inactivity exceeds a threshold, the system enters a dormant state and temporarily reconstructs the opportunity cost logic, introducing an information value decay dimension. The mechanism will trigger a low-probability, extremely low-cost wake-up detection primitive, such as a short-term charge with a tiny current, with a preset low probability. Although this micro-perturbation generates a small deterministic cost, it aims to change the state of charge, prompting the system to unlock and evaluate a new set of opportunity costs for behavioral primitives corresponding to the updated state in the next decision-making cycle. This creates the possibility of escaping local decision-making deadlocks and entering a potentially better state space. This mechanism transforms the concept of inactivity loss, traditionally considered a negative asset, into an active information maintenance cost to maintain system activity and opportunity sensitivity, enabling the system to reverse entropy increase and proactively seek global optimization in a mediocre market. Attached Figure Description

[0027] Figure 1 This is a flowchart of the optimization and evaluation process for the energy storage charging and discharging strategy of the present invention;

[0028] Figure 2 This is a probability analysis diagram of the charging and discharging strategy selection in this invention;

[0029] Figure 3 This is a flowchart of the decision-making and execution process of the energy storage system of the present invention.

[0030] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in further detail below. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0032] This invention provides a method for optimizing and evaluating energy storage charging and discharging strategies based on big data analysis. The system operates as a closed-loop adaptive decision-making process. This process begins with the construction of a standardized set of operational instructions, i.e., a behavioral primitive library, and is driven by a decision-making process based on an opportunity cost quantification rule library established from historical data. This rule library maps the real-time acquired current operating status data of the energy storage system to the current opportunity cost of each executable behavioral primitive. Subsequently, the system selects and issues instructions from all feasible behavioral primitives in a probabilistic manner, and obtains deep state parameters by periodically executing specific diagnostic behavioral primitives, thereby providing feedback calibration to the opportunity cost quantification rule library itself, thus forming a continuously optimizing cycle. In the face of power... When faced with the challenge of high-frequency and irregular fluctuations in spot market prices, traditional decision-making methods relying on predictive models are unable to respond to market changes in real time due to their inherent computational delays. To avoid this dependence on complex predictive models, this solution decouples all operations of the energy storage system into a finite set of standardized behavioral primitives, collectively forming a behavioral primitive library. These behavioral primitives are not broad instructions, but rather precisely quantified standard action units, such as charging at a 0.5C rate for 15 minutes, discharging at a 1C rate for 15 minutes, or remaining stationary for 15 minutes. This discretization and standardization lays the foundation for subsequent quantitative assessment and rapid decision-making. Based on this, the system constructs an opportunity cost quantification rule base, the core function of which is to build... This rule base establishes a direct mapping relationship between the current definite state and potential future gains or losses. Its data structure is essentially a multi-dimensional mapping table. Its input dimensions include the historical percentile level of the current real-time electricity price and the preset interval of the current system state of charge. The procedure for determining the historical percentile level is as follows: the system first retrieves electricity price data from the past statistical period, such as 90 days, and calculates a set of key percentiles, such as 10%, 25%, 50%, 75%, and 90%. The current real-time electricity price is then compared with these percentiles to uniquely determine its historical high / low level interval. The system state of charge is divided into several non-overlapping intervals, such as 0-15%, 15-30%, and so on. 85-100%, etc.; When the system obtains the current state in real time, such as when the electricity price is above the 90th percentile and the state of charge is in the 70-85% range, the rule base immediately calculates the current opportunity cost for each behavioral element. For example, for the charging element, its opportunity cost is high because it means buying electricity at a high price, while for the discharging element, its opportunity cost is extremely low because it represents seizing the opportunity to profit from high-price arbitrage. For the idle element, its opportunity cost reflects the loss of missing this arbitrage opportunity. By locking the decision-making basis to the current observable state in this way, the response speed of the entire decision-making process is only limited by the efficiency of data collection and rule matching, thereby realizing the real-time adaptability to the high-frequency market environment.

[0033] Furthermore, simply choosing the action with the lowest opportunity cost may lead to oscillations or excessive profit-seeking at the decision boundary. To smooth decision outputs and endogenously coordinate economic benefits and physical security, this scheme, after determining the current opportunity cost of each action, adopts a probabilistic approach to select the action to be executed in the next time period. Specifically, in the system, the first... The probability of a behavior primitive being selected It is not a deterministic Boolean value, but is calculated using an inverse proportional formula, namely... ,in, Indicates the first The current opportunity cost of each behavioral primitive, and This represents the total number of currently executable action primitives. This mechanism ensures that action primitives with lower opportunity costs have a higher probability of being selected, while primitives with slightly higher costs still have a certain probability of being executed, thus increasing the system's exploratory nature. More importantly, this mechanism works in synergy with the dynamic adjustment of opportunity costs. When the system's state of charge enters an extremely high or low range determined based on battery safety characteristics, such as below 15% or above 95%, the opportunity cost quantification rule base automatically applies a penalty factor to the opportunity cost of charging and discharging action primitives. The specific value of this penalty factor is set to be positively correlated with the degree to which the system's state of charge deviates from the safe range, for example, increasing quadratically. Thus, the closer the state of charge is to the physical limit, the more the economic cost of charging or discharging is artificially increased, and the probability of its selection drops sharply. Therefore, without introducing complex physical models and hard constraints, the system behavior is naturally guided to converge within the battery's safe operating range through economic incentives, achieving an inherent balance between asset protection and market arbitrage flexibility.

[0034] To address the drift in physical characteristics caused by aging throughout the battery's lifecycle and ensure the long-term effectiveness of the decision-making model, this solution integrates an adaptive calibration and physical boundary arbitration mechanism based on hardware function reuse. On one hand, it achieves online sensing of the battery's internal health status by periodically executing a specific diagnostic behavior primitive. This diagnostic behavior primitive is designed as a standardized micro-current charge-discharge process. Its execution is not for energy arbitrage, but rather to utilize the energy storage system's own power electronics to accurately record the current decay curve during the constant-voltage charging phase. The system then uses nonlinear fitting of this current decay curve to deduce key parameters characterizing the battery's electrochemical impedance properties, such as the time constant. The time constant The changes directly reflect the degree of battery health degradation, and the value will be fed back to adaptively calibrate the opportunity cost quantification rule base, such as dynamically adjusting cost penalty items related to battery aging. On the other hand, to ensure absolute operational safety, this solution adds two physical anchoring mechanisms. The first is a physical characteristic deviation preprocessing gateway set up before opportunity cost calculation. This gateway estimates the battery's macroscopic dynamic internal resistance using the measured voltage and current fluctuation data before and after each behavioral primitive execution, and compares it with the baseline internal resistance value at the time of database construction. When the deviation percentage exceeds a set threshold, such as 20%, the rule base is adjusted. First, a penalty factor positively correlated with the percentage of deviation is applied to the opportunity cost of all high-power charging and discharging units. Second, a physical limit boundary arbitration module is set up before the final instruction is issued. This module, based on fixed absolute safety boundary rules, performs a rapid virtual simulation on the selected unit to be executed, and calculates the predicted voltage and temperature rise after execution based on the latest estimated macroscopic dynamic internal resistance. If the predicted result will touch the absolute safety boundary, the module will exercise its veto power and force the original execution instruction to be replaced by a stationary unit. The synergistic effect of these two mechanisms ensures that the decision-making system can adapt to the slow aging of the battery while avoiding instantaneous physical risks. Finally, to solve the system... To address the potential decision-making deadlock in prolonged stable market environments—specifically, the lack of significant electricity price differentials leading to the system continuously selecting inactive behavior elements and losing sensitivity to market changes—this solution introduces a dormant-to-awaken opportunity detection mechanism. This mechanism is triggered by determining whether the energy storage system has entered a dormant state. The determination procedure involves monitoring whether the duration of continuous selection of inactive behavior elements exceeds a set duration threshold, such as 24 hours. Once triggered, the system does not wait indefinitely but instead forcibly selects and executes a special wake-up detection element with a preset low probability, such as 5% per hour. This wake-up detection element is set... Consider a tiny perturbation to the system state, such as a very short charging or discharging operation with a preset extremely small current. This operation itself produces almost no economic benefits. Its core purpose is to actively change the system's state of charge at a very low cost, even if it is only a tiny change. This change in state will allow the system to face a completely new set of opportunity costs when it conducts opportunity cost assessment in the next decision cycle. This creates the possibility of escaping the local optimum trap caused by continuous stagnation and rediscovering potential market opportunities. It gives the system the ability to resist entropy increase and actively seek a better global solution in a mediocre market environment.

[0035] Example 1: In a real-world operating scenario of an industrial-grade energy storage power station providing high-frequency auxiliary services to the power grid, the station has been operating for several years, and its battery system exhibits signs of aging. Specifically, when faced with drastic fluctuations in grid commands or market prices, although the system strives to capture arbitrage opportunities through rapid charging and discharging, its internal battery management unit increasingly triggers protection against instantaneous overvoltage or excessive temperature rise, forcing the charging and discharging process to be interrupted. This not only results in the loss of expected economic benefits but also poses a potential threat to the long-term health of the battery system. This increasingly acute contradiction between economic goals and physical safety boundaries constitutes the core challenge faced by this aging energy storage system in its continuous operation. Under these conditions, the technical solution of this invention guides the charging and discharging decisions of the energy storage system. During a period of relatively stable market prices and low system load, the system automatically triggers a periodic diagnostic behavior primitive. By executing a standardized micro-current charging and discharging process and recording the current decay curve during its constant voltage phase, the system retrieves parameters characterizing the current internal health status of the battery. This parameter Compared to the baseline value during the initial operation of the system, an observable increase has been shown. At the same time, the system uses voltage and current abrupt change data before and after the most recent charge and discharge behavior to update the estimated value of the battery's macroscopic dynamic internal resistance through the voltage drop method. This estimated value also confirms the trend of battery internal resistance increasing with aging. These two parameters, which are sensed and characterized in real time through different paths, are directly used to calibrate the opportunity cost quantification rule base. Specifically, the physical characteristic deviation preprocessing gateway converts the percentage deviation between the current internal resistance estimate and the baseline internal resistance value into a dynamic penalty factor for the opportunity cost of all high-power charge and discharge behavior primitives.

[0036] Correspondingly, when the electricity market price experiences another brief but sharp rise, entering the peak price range of historical percentile levels, the system's decision-making process is adjusted based on updated parameters. The opportunity cost quantification rule base, based on high electricity prices and the current state of charge, determines that high-power discharge is the action element with the lowest opportunity cost. However, before making the final probabilistic selection, a penalty factor generated by the physical characteristic deviation preprocessing gateway and positively correlated with the internal resistance increment is applied to the opportunity cost of all high-power discharge elements. As a result, the 1C rate discharge element, which originally had the lowest opportunity cost, has its corrected opportunity cost significantly increased due to the greatest physical impact on the aging battery. Meanwhile, the cost difference between the corrected opportunity cost and the original optimal element, such as the 0.7C rate discharge element (a suboptimal power), is narrowed. The system then... The probabilistic selection method ultimately chose a discharge element with slightly lower power but better suited to the battery's physical state as the execution action with a high probability. In the final stage before the command was issued, the physical limit boundary arbitration module performed a virtual simulation of this 0.7C rate discharge command based on the updated macroscopic dynamic internal resistance, confirming that its estimated terminal voltage and temperature rise were still within the absolute safety boundary, thus allowing the operation to proceed. This process resolved the inherent contradiction between simply pursuing maximum economic efficiency and ensuring the physical safety of aging batteries. It does not prohibit high-power operation, but rather quantifies the real-time physical health state as a dynamic cost item in economic decision-making, making the system... It can autonomously find the balance between economic benefits and physical capacity under the current state; through the continuous operation of the above mechanism, the operation of the energy storage power station enters a new stable stage. The system can still respond sensitively to every fluctuation in market price and execute arbitrage operations, but the protective shutdown events caused by instantaneous overvoltage or excessive temperature rise are effectively suppressed, the overall availability of the system is maintained, and its decision-making behavior mode changes from reckless extreme utilization to dynamic adjustment adapted to its own health state. Thus, without relying on complex battery degradation prediction models, the continuous optimization of the operating efficiency of energy storage assets throughout their entire life cycle is achieved.

[0037] Example 2: To objectively verify the effectiveness of the adaptive adjustment of the technical solution of the present invention when the physical state of the battery changes, this comparative experiment was established. The purpose of the experiment is to quantitatively evaluate the specific performance of the method of the present invention in terms of operational safety and efficiency maintenance when dealing with the performance degradation caused by battery aging, compared with the benchmark method using a fixed strategy. This experiment was built on a standardized battery testing platform, which consists of a 100 Ah lithium iron phosphate single cell, a programmable high-precision DC power supply and electronic load, and a control system as a host computer. The control system is responsible for running the decision-making algorithm and issuing commands to the power supply and load. The electricity price data used in the experiment came from a 72-hour record of real electricity spot market prices. This price data was refreshed every 5 minutes and loaded in a loop during the experiment to ensure that the two experimental groups faced the exact same external market environment. The key parameter of the experiment, the decision time window, needed to balance the response speed of the market signal with the computational load of the system itself. Since the decision-making process of this scheme does not rely on complex forecasts, the calculation time for a single decision was measured to be less than 100 milliseconds, which is far lower than the 5-minute (300-second) market price refresh cycle. Therefore, the decision time window was set to 300 seconds to achieve synchronization with the market rhythm and avoid unnecessary consumption of computational resources.

[0038] The test procedure first involves initial characteristic calibration of the test battery with 100% health status to obtain its baseline macroscopic dynamic internal resistance and characteristic time constant. Based on this, an initial opportunity cost quantification rule base was constructed. Subsequently, the battery was accelerated to age by continuously executing deep charge-discharge cycles until its capacity decayed to 85% of its initial capacity. All subsequent experiments were conducted on this aged battery. The experiment was divided into a control group and an experimental group. The decision algorithm loaded in the control group adopted the initial opportunity cost quantification rule base and did not have the function of adaptive adjustment according to the battery state. The experimental group loaded the complete technical solution of the present invention, including all mechanisms such as periodically triggered diagnostic behavior primitives and preprocessing gateways based on physical characteristic deviations. After the experiment started, the experimental group was run first. Before formally loading the cycle electricity price data, the system first executed a diagnostic behavior primitive to obtain the current macroscopic dynamic internal resistance and time constant of the aged battery. Based on this, the penalty factor in the opportunity cost quantification rule base was adaptively calibrated. Subsequently, during a continuous 24-virtual-hour test period, it was observed that when the market price showed a peak that met the conditions for high-power discharge arbitrage, both the control group and the experimental group identified this opportunity. However, the control group tended to directly select the 1C high-power discharge element with the lowest opportunity cost defined in the original rule base, while the experimental group, due to the deviation of its physical characteristics, had already penalized the cost of the high-power element based on the aged internal resistance data of the preprocessing gateway. As a result, the experimental group ultimately selected a relatively mild 0.75C discharge element with a high probability. This difference occurred repeatedly throughout the test period. The specific data comparison is shown in Table 1.

[0039] Table 1: Comparison of operating data of the two strategies on aged batteries.

[0040]

[0041] Data shows that when faced with the same arbitrage opportunities, the control group's decision-making model failed to detect the deterioration of battery physical characteristics. As a result, its high-power discharge command was forcibly interrupted by the underlying battery management system 16 times due to triggering the preset voltage or temperature rise safety boundary, resulting in substantial missed opportunities. In contrast, the experimental group's decision-making system, through its built-in adaptive calibration mechanism, transformed the battery's aging state into an economical suppression of high-risk behavior, actively reducing the output power and almost avoiding all situations that touched the safety boundary. This enabled it to complete a stable and complete discharge process in every arbitrage window, and its cumulative effective arbitrage amount actually exceeded that of the more aggressive control group.

[0042] Example 3: This example combines Figures 1 to 3 This paper describes an optimization and evaluation method for energy storage charging and discharging strategies based on big data analysis. Figure 1As shown, firstly, the system acquires real-time electricity prices from the power grid and the current state of charge (SOC) of the energy storage system's batteries through a real-time data acquisition module. Next, the system inputs this data into a price quantile analysis and SOC interval coding module to perform statistical analysis on the historical quantile levels of the real-time electricity prices and the SOC of the batteries. Then, based on these analysis results, the opportunity cost quantification rule base maps the current SOC of the energy storage system to the real-time opportunity costs of corresponding behavioral primitives, including charging, discharging, and idle behaviors. Based on this opportunity cost data, the system calculates the current cost of each behavioral primitive through an opportunity cost calculation module and selects the optimal behavioral primitive as the action to be executed in the next time period using a probabilistic selection mechanism. During the selection process, actions with lower opportunity costs are prioritized. The behavioral primitives will have a high probability of being selected. In order to optimize the long-term operating efficiency of the battery, the system acquires battery health status parameters by periodically executing the standardized micro-current charge and discharge process in the diagnostic module. By collecting these parameters, the system can adaptively calibrate the opportunity cost quantification rule base to improve the accuracy of decision-making. Under specific conditions, such as the static dormant state, the system will execute the wake-up probe primitive to promote the change of the state of charge of the energy storage system with a small power change, ensuring that the system remains active in a long-term stable market environment and avoids falling into decision deadlock. Finally, through physical limit boundary arbitration, the system ensures that the selected behavior will not exceed the safe operating range of the battery, thereby ensuring a balance between operational safety and economic benefits.

[0043] like Figure 2 As shown in the figure, the vertical axis represents the probability of strategy selection, displaying the probability distribution of the energy storage system choosing charging, discharging, and resting under different electricity price conditions. It can be seen from the figure that as the electricity price level increases, the probability of the energy storage system choosing the discharging strategy gradually increases, while the probability of choosing charging gradually decreases. This is especially true under high and extremely high electricity prices, where the probability of discharging increases significantly, while the probability of charging decreases sharply. Furthermore, the probability of the resting strategy is higher when the electricity price is low, and gradually decreases as the electricity price increases. This change in strategy selection reflects the process by which the energy storage system makes decisions based on opportunity cost under different electricity price levels, ultimately maximizing economic benefits while ensuring the safety and stability of the system.

[0044] like Figure 3As shown, the system first acquires data through real-time grid electricity prices and battery SOC (State of Charge). This data is analyzed and quantified by a quantile analyzer and a state encoder to help the system assess the current electricity market situation in real time. Data acquisition is performed every 5 minutes, and the system's health status is checked through a diagnostic module and an internal monitoring module. Based on this real-time acquired status information, the system calculates the opportunity costs of current behavioral primitives, including charging, discharging, and resting, using an opportunity cost quantification rule base. The system then selects the most suitable behavioral primitive using a probability selector. In this decision-making process, the system uses a short chain of machine learning algorithms for calculation and adjusts the T parameter to ensure the accuracy and real-time performance of the decision. Finally, the system transmits execution instructions to the power converter and execution control unit to implement the energy storage system's charging and discharging strategy through charging and discharging commands. This decision-making process ensures that the energy storage system can respond promptly and optimize economic benefits in the electricity spot market, while also protecting the battery's safety and health.

[0045] Example 4: This example aims to provide a specific engineering implementation procedure for the offline construction and parameter calibration of the core components of the aforementioned technical solution, namely the opportunity cost quantification rule base and its embedded penalty factors, before system deployment. This addresses the problem of how to generate an effective and reproducible decision model from raw data in practical applications. In the preparation stage before a storage system is put into commercial operation, its primary task is to assign a logically complete and data-driven initial quantification rule base to the opportunity cost quantification rule base. To achieve this, the system executes an offline modeling process based on historical data. The input to this process is a sequence of electricity price data for the region where the energy storage power station is located over the past full year, at 5-minute intervals, and the rated charge / discharge efficiency parameters of the energy storage system itself. The first step of the process is to perform quantile statistics on the historical electricity price data, dividing the historical electricity price data into six price level intervals based on the 10%, 25%, 50%, 75%, and 90% quantiles, and statistically analyzing any one of the historical price levels. The transition probability matrix for a flat interval to move to any other interval after the next decision time window; the process involves calculating the specific opportunity cost value for each cell in the opportunity cost quantification rule base, i.e., the combination of each system charge state interval and the historical percentile level of the real-time electricity price of the grid, for the three basic behavioral primitives of charging, discharging, and resting. Here, opportunity cost is deterministically defined as the expected future revenue forgone after performing a certain behavior. Under medium electricity price level and medium charge state, the opportunity cost of charging behavior includes not only the direct cost of current electricity purchase, but also the expected revenue that may be missed in a future high electricity price period due to the consumption of charging capacity and battery capacity. This expected revenue is calculated by weighted averaging the aforementioned transition probability matrix, combined with the average electricity price and charging / discharging efficiency of each future electricity price interval. By repeating this calculation process for all behavioral primitives under all state combinations, a complete opportunity cost quantification rule base that transforms abstract market statistical laws into specific numerical decision-making basis is constructed.

[0046] Furthermore, to enable the rule base to adapt to changes in the physical characteristics of the battery throughout its entire life cycle, the key parameters of the penalty factor used to dynamically adjust opportunity costs also need to be deterministically calibrated. When the system detects a deviation of the physical parameter characterizing battery aging, namely the macroscopic dynamic internal resistance, from its initial baseline value, a penalty factor is applied. The penalty factor, which will be applied to the opportunity cost calculation of high-power behavioral primitives, is calculated as a function linearly related to the relative change in internal resistance. ,in, For real-time estimation of internal resistance, This is the reference internal resistance of the battery at the time of manufacture, and It is a punishment sensitivity coefficient; this coefficient The value is determined through a one-time calibration experiment. The calibration principle is to anchor a specific physical boundary condition to a desired economic decision effect. Specifically, based on the setting of the physical characteristic deviation preprocessing gateway, when the internal resistance deviation percentage, i.e. When the set threshold of 20% is reached, the system should significantly suppress high-risk behaviors. Based on this, the calibration experiment will... The value is set so that the deviation reaches 20%. The value is exactly 2. At this physical risk threshold, the economic cost of high-power behavior is twice the initial cost. Through this procedure, the response strength of the penalty factor is directly and verifiable mathematically related to a clear engineering boundary condition. After the above offline modeling and parameter calibration process, the decision-making core of the energy storage system is given a complete initial state with deterministic data sources and calculation rules before it is put into operation. Each value in the opportunity cost quantification rule base is supported by its historical data statistics and expected return calculation logic. The strength of the penalty factor is also precisely quantified by anchoring it to the physical safety threshold, so that the entire decision-making system has a traceable quantitative basis, providing an initial decision basis for subsequent online operation and adaptive adjustment.

[0047] Example 5: When deploying the technical solution of this invention in a newly built energy storage power station with specific physical characteristics, to ensure that the behavioral primitive library in the decision-making system can accurately match the hardware performance boundaries it controls, the system executes a pre-emptive hardware characteristic calibration procedure during the integration phase. This procedure controls the power conversion module of the energy storage system to perform a series of standardized charge-discharge test sequences covering its rated power range on the battery system. At each power step of the test sequence, the actual output power, energy conversion efficiency, and core temperature rise rate of the system are recorded with high precision. These measured data are compared with the design specifications provided by the battery manufacturer. If, at a certain power point, the measured energy conversion efficiency is lower than the preset economic threshold or the temperature rise rate exceeds the long-term operational reliability limit, then the system is terminated. The corresponding behavioral primitives are not included in the final deployed behavioral primitive library. Through this procedure, a behavioral primitive library optimized for this specific hardware and excluding uneconomical and high-risk operational options is ultimately formed. Next, after the energy storage system completes on-site installation and initial commissioning, entering the final commissioning phase before formal commercial operation, the system executes a one-time baseline state parameter acquisition process. This process aims to establish a unique digital profile representing the initial health state of the physical asset. After the process starts, the system will fully charge the battery system with a preset, extremely low constant current, and then execute a standardized micro-current discharge process as a diagnostic behavioral primitive. During this process, the system not only accurately records the initial macroscopic dynamic internal resistance calculated using the step-down method, but also... The required voltage and current data were obtained, and a complete current decay curve was captured during the constant voltage phase. The time constant characterizing the initial electrochemical properties of the battery was then derived from this curve. The baseline values, this set of values ​​was measured in actual installation environments. and The baseline value will serve as a fixed reference for all adaptive algorithms throughout their entire lifecycle.

[0048] Example 6: To ensure the adaptability and robustness of the decision-making system of the present invention in different market environments and during long-term operation, the system performs a set of periodic operating parameter calibration and online data health self-check procedures after deployment. In the parameter calibration stage, the system first retrieves and analyzes the real-time electricity price data of the power grid for the past statistical period, such as 30 days, and calculates the standard deviation of the sequence to quantitatively characterize the recent market activity. Accordingly, two key control parameters related to the dormant state in the system, namely the duration threshold for triggering dormancy and the probability of the wake-up detection primitive used to break out of dormancy, will be adjusted in conjunction. The set value of the duration threshold is inversely proportional to the market volatility. In a market environment with high volatility, this threshold will be automatically lowered to encourage the system to examine market opportunities more frequently. The probability of the wake-up detection primitive being selected is directly proportional to the market volatility. In a similarly active market, this probability will be increased to increase the frequency of the system actively detecting potential arbitrage windows.

[0049] Furthermore, to avoid abnormal jumps in physical parameter estimation caused by instantaneous sensor failures or external electromagnetic interference during online operation, which could mislead the adaptive decision-making logic, the system integrates a sliding window filtering and outlier removal module into its real-time estimation of macroscopic dynamic internal resistance. This module does not directly use the latest calculated instantaneous internal resistance value, but maintains a queue containing the past ten valid internal resistance measurements. The median value of this queue is used as input to the physical characteristic deviation preprocessing gateway for subsequent penalty factor calculation. Simultaneously, a deterministic data health arbitration logic is executed in parallel. If the difference between a newly calculated instantaneous internal resistance value and the average value within the current sliding window exceeds three times the standard deviation of the data within the window, the measurement result will be judged as a statistically significant outlier. The system will then mark it as outlier data and discard it, excluding it from the sliding window queue update. Through this filtering and arbitration mechanism, the system can effectively distinguish between the trend-like increase in internal resistance caused by slow battery aging and the sudden data jumps caused by external disturbances or failures, thereby ensuring the stability and reliability of its core adaptive decision-making.

[0050] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for evaluating energy storage charge and discharge strategy optimization based on big data analysis, characterized in that, Includes the following steps: Step a, construct a behavior primitive library, which contains standard action units for charging, discharging and stationary scenarios of energy storage systems; Step b: Construct an opportunity cost quantification rule base. The rule base establishes quantitative rules that map the current operating state of the energy storage system to the real-time opportunity cost of the corresponding behavioral primitives based on historical electricity price data and historical system state of charge data. The current operating state includes the historical percentile level of the current real-time electricity price of the power grid and the corresponding interval of the current system state of charge. Step c: Obtain the current operating status data of the energy storage system in real time; Step d: For each executable behavioral primitive in the behavioral primitive library, calculate the current opportunity cost of each executable behavioral primitive based on the current running state data and using the opportunity cost quantification rule library; Step e: Based on the current opportunity cost of each executable action primitive, select one action primitive as the execution action for the next period in a probabilistic manner, wherein the probability of any action primitive being selected is inversely proportional to its current opportunity cost. Step f involves periodically executing diagnostic behavioral primitives to obtain parameters characterizing the internal health status of the battery, and adaptively calibrating the opportunity cost quantification rule base based on the health status parameters. The probabilistic selection method is as follows: calculate the first element in the primitive library of computational behavior. The probability of a behavior primitive being selected probability The calculation formula is: ,in, Indicates the first The current opportunity cost of an individual behavioral element. This represents the total number of executable behavioral primitives in the behavioral primitive library; In step f, the diagnostic behavioral primitive includes a standardized microcurrent charge-discharge process, which is executed by the power electronics of the energy storage system itself, and the current decay curve of the battery is recorded with high precision during the process; the internal health state parameters of the battery are obtained by fitting the current decay curve to inversely derive the time constant characterizing the electrochemical impedance of the battery.

2. The method for optimizing and evaluating energy storage charging and discharging strategies based on big data analysis as described in claim 1, characterized in that, In the opportunity cost quantification rule base, when the current system state of charge is in an extremely high or extremely low range determined based on battery safety characteristics, a penalty factor is applied to the opportunity cost of charging and discharging behavior primitives. The penalty factor is determined according to the degree to which the system state of charge deviates from the safe range, so as to guide the behavior of the energy storage system towards the battery safe operating range.

3. The method for optimizing and evaluating energy storage charging and discharging strategies based on big data analysis as described in claim 1, characterized in that, The historical percentile of the current real-time electricity price is determined by comparing the current real-time electricity price with multiple percentiles obtained from historical electricity price data. These multiple percentiles are used to characterize the position of the electricity price in the historical distribution. 4.The method of claim 1, wherein, After step d is executed and before step e is executed, the method further includes: determining whether the energy storage system has entered a dormant state, where the dormant state refers to the time during which the dormant behavior primitive is continuously selected exceeds a set duration threshold; when the energy storage system has entered a dormant state, forcibly selecting and executing a wake-up probe primitive in the behavior primitive library with a preset low probability, whereby the wake-up probe primitive is used to change the system state of charge of the energy storage system with a small power, so as to prompt the energy storage system to evaluate a new set of opportunity costs for behavior primitives in the next decision cycle.

5. The energy storage charge and discharge strategy optimization evaluation method based on big data analysis according to claim 4, characterized in that, The wake-up detection element includes charging or discharging operations with a preset, extremely small current.

6. The method of claim 1, wherein the method is based on big data analysis. The method further includes: after selecting the behavior primitive as the execution action for the next time period in step e, before issuing the execution command to the power control unit of the energy storage system, performing physical limit boundary arbitration; the physical limit boundary arbitration performs virtual simulation of the behavior primitive to be executed based on the fixed absolute safety boundary rule library, and determines whether the estimated voltage and estimated temperature rise of the battery after executing the behavior primitive will touch the absolute safety boundary; when it is determined that the absolute safety boundary has been touched, the original execution command is rejected and replaced with a static primitive in the behavior primitive library.

7. The energy storage charge and discharge strategy optimization evaluation method based on big data analysis according to claim 6, characterized in that, The physical limit boundary arbitration also includes: during the execution of the behavioral primitive or at the moment of completion, using the battery voltage and current change data measured before and after the execution of the behavioral primitive, estimating the macroscopic dynamic internal resistance of the battery by the voltage drop method; and the virtual simulation calculates the predicted voltage and predicted temperature rise of the battery based on the macroscopic dynamic internal resistance. 8.The method of claim 1, wherein, Before step d is executed, the method further includes: evaluating the real-time effectiveness of the opportunity cost quantification rule base through a physical characteristic deviation preprocessing gateway. The physical characteristic deviation preprocessing gateway compares the real-time aging drift characteristics of the battery with the reference internal resistance value when the opportunity cost quantification rule base was established. When the percentage deviation between the aging drift characteristics and the reference internal resistance value exceeds a set threshold of 20%, a penalty factor is applied to the opportunity cost of all high-power charging and discharging behavior primitives in the opportunity cost quantification rule base. The magnitude of the penalty factor is positively correlated with the percentage deviation.

Images