Energy storage planning and identification method based on fatigue sensitivity and marginal utility

Abstract

The application discloses a kind of energy storage planning and identification method based on fatigue sensitivity and marginal utility, it is related to power system planning and energy management technical field.The method includes: obtaining user historical load data, extracting the power, duration and frequency characteristics of load event, and screening out short-time peak load event according to this, form effective load sequence;Effective load sequence is input into virtual battery model to calculate unit throughput loss cost;Combined with electricity price data and loss cost, based on marginal net income growth rate inflection point determines optimal energy storage configuration capacity;Through simulating load fluctuation scene, the configuration capacity is verified, and finally the energy storage potential identification report is output.The application is used to solve the problems of energy storage power configuration redundancy, system full life cycle performance evaluation deviation and weak anti-fluctuation interference ability caused by ignoring load impact characteristics and battery dynamic loss in the prior art.

Description

Technical Field

[0001] This invention relates to the field of power system planning and energy management technology, and more specifically, to a method for energy storage planning and identification based on fatigue sensitivity and marginal utility. Background Technology

[0002] User-side distributed energy storage systems are crucial physical infrastructures for regulating industrial power load fluctuations, improving power quality, and enhancing power supply reliability. In complex industrial power supply networks, accurately identifying load characteristics and accordingly planning the capacity and power of energy storage systems is fundamental to ensuring the safe and stable operation of energy storage devices throughout their entire lifecycle and maximizing the effectiveness of peak shaving and valley filling technologies. Currently, analyzing industrial load characteristics and configuring systems based on historical operating data is a common practice in the engineering field.

[0003] Existing technologies for planning user-side energy storage configurations typically employ static estimation based on low-frequency metering data (such as monthly electricity consumption) or linear programming models based on typical daily load curves. The static estimation method primarily uses transformer capacity and monthly maximum demand statistics to estimate the rated power and capacity of the energy storage system at a fixed ratio. The typical daily method selects load curves for specific time periods, assumes the load has deterministic periodic repetitive characteristics, and uses mathematical modeling to solve for system configuration parameters that satisfy power balance constraints.

[0004] However, actual industrial loads are highly nonlinear, transient, and time-varying. Existing technologies have significant technical shortcomings in physical model construction and system configuration: First, existing methods lack fine-grained differentiation of the frequency domain characteristics and time domain persistence of the load, failing to distinguish between steady-state peak loads and transient impact loads. If millisecond- or second-level large motor starting impacts are misjudged as demand regulation targets, it will lead to severe redundancy in the power configuration of energy storage converters, and the system will fail in actual operation due to its inability to respond quickly to high-frequency impacts, resulting in control strategy failure. Second, traditional models typically use simplified linear life decay models, estimating battery status only through total throughput, ignoring the battery micro-circulation effect caused by high-frequency fluctuations in industrial loads. This lack of a physical model leads to biased predictions of the battery's state of electrochemical health (SOH), causing the system to experience performance drops or safety failures before reaching its expected lifespan. Third, models based on static typical days lack robustness verification against operating condition disturbances. When faced with load shifts or random fluctuations caused by changes in production schedules, the established capacity configuration often cannot maintain the dynamic balance of the state of charge (SoC), and is very prone to power depletion (SoC lower limit exceeding the limit), which leads to system protection tripping or regulation function interruption, and fails to meet the requirements of industrial-grade power supply reliability. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an energy storage planning and identification method based on fatigue sensitivity and marginal utility. This method addresses the problems of redundant energy storage power configuration, biased system lifecycle performance evaluation, and weak anti-fluctuation interference capability caused by prior art neglecting load impact characteristics and battery dynamic losses. This is achieved by eliminating short-term peak-type ineffective loads, determining the optimal capacity based on the system's marginal utility inflection point, and simulating load fluctuation scenarios for verification.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A fatigue-sensitivity and marginal utility-based energy storage planning and identification method includes the following steps: identifying load events based on historical load data and generating a first load sequence containing the power and duration characteristics of each event; filtering the first load sequence according to the power and duration characteristics, removing short-term peak events to obtain a target load sequence; inputting a preset virtual battery model based on the target load sequence, performing time-series tracking simulation under preset state-of-charge safety boundary constraints, and calculating the unit throughput loss cost based on the generated actual charge and discharge response trajectory; combining electricity price data and the unit throughput loss cost to calculate the marginal net benefit of energy storage; denoising the mapping relationship between the marginal net benefit of energy storage and capacity, and determining the optimal energy storage configuration capacity based on the inflection point of the marginal net benefit growth rate; verifying the optimal energy storage configuration capacity by simulating the impact of preset load fluctuation scenarios on daily net benefit, and outputting an energy storage potential identification report.

[0007] In a preferred embodiment, the step of filtering the first load sequence and removing short-term peak load events to obtain the target load sequence includes: extracting waveform segments from historical load data that continuously exceed a preset power as load events to form the first load sequence; traversing the first load sequence to obtain the power and duration of each load event; if the duration of a load event is less than a preset time threshold and the power is greater than a preset power threshold, then it is removed from the sequence, and the remaining load events form the target load sequence.

[0008] In a preferred embodiment, the calculation of unit throughput loss cost includes: using the rainflow counting method to count the cycle depth and corresponding number of cycles of the virtual battery model when tracking the target load sequence; calculating the total battery life loss value based on the statistical results and the battery life decay curve; converting the total life loss value into monetary value and dividing it by the total throughput of the target load sequence to obtain the unit throughput loss cost.

[0009] In a preferred embodiment, determining the optimal energy storage capacity based on the inflection point of the marginal net revenue growth rate includes: sorting by comprehensive power characteristics, covering load events in the target load sequence, gradually increasing the energy storage capacity and calculating the marginal net revenue, and plotting the relationship curve between capacity and net revenue; the comprehensive power characteristics are the weighted composite value of power, duration, and frequency characteristics of the load time; the marginal net revenue is the electricity cost savings brought about by increasing the unit energy storage capacity minus the corresponding unit throughput loss cost.

[0010] In a preferred embodiment, before determining the optimal energy storage configuration capacity, the method further includes determining the optimal demand control threshold, specifically including: Set an initial demand control threshold, and gradually decrease the threshold with a preset step size to generate multiple sets of simulated control strategies; For each set of simulated control strategies, the corresponding fatigue sensitivity index is calculated using the rainflow counting method. The fatigue sensitivity index is the ratio of the increase in battery life loss to the increase in demand electricity cost savings. The threshold corresponding to the inflection point where the fatigue sensitivity index undergoes a sudden increase relative to the demand control threshold is determined as the optimal demand control threshold.

[0011] In a preferred embodiment, the step of verifying the optimal energy storage configuration capacity based on the simulation results includes: superimposing random noise on the target load sequence or adjusting the event occurrence time to generate several simulated load scenarios; running the control strategy of the optimal energy storage configuration capacity under the simulated load scenarios, and calculating the risk probability of energy storage power being exhausted or unable to meet demand limits; if the risk probability exceeds a preset safety value, reducing the optimal energy storage configuration capacity until the risk probability meets the requirements.

[0012] In a preferred embodiment, before inputting the target load sequence into the preset virtual battery model, the method further includes: determining the operating parameters of the virtual battery model, specifically: performing cluster analysis on users based on the power, duration, and frequency characteristics of all load events in the first load sequence; classifying users into stable arbitrage type, shock smoothing type, and demand sensitive type, and matching differentiated virtual battery model operating parameters for different types of users.

[0013] In a preferred embodiment, the output energy storage potential identification report includes the output of the optimal energy storage configuration capacity, the expected investment return period, and the risk assessment results under the load fluctuation scenario.

[0014] In a preferred embodiment, the step of performing time-series tracking simulation under a preset state-of-charge safety boundary constraint and calculating the unit throughput loss cost based on the generated actual charge-discharge response trajectory includes: calculating the actual response discharge power based on the theoretical peak-shaving power demand of the load event, combined with the remaining power and maximum inverter power at the current moment; updating the actual state of charge based on the actual response discharge power to generate an actual state of charge fluctuation sequence; and calculating the cycle depth and corresponding cycle number of the fluctuation sequence using the rainflow counting method, and combining it with the battery life decay curve to calculate the unit throughput loss cost.

[0015] In a preferred embodiment, determining the optimal energy storage configuration capacity based on the changing trend of the marginal net income includes: smoothing the relationship curve between capacity and net income using a Savitzky-Golay filter with local polynomial least squares fitting to eliminate local data jumps caused by discrete load events; performing difference calculation on the smoothed curve to obtain a second derivative sequence; and when the second derivative sequence is lower than a preset negative threshold in a continuous interval, taking the capacity corresponding to the starting point of the continuous interval as the optimal energy storage configuration capacity.

[0016] This invention provides an energy storage planning and identification system based on fatigue sensitivity and marginal utility, comprising: a feature extraction module, used to extract load events from historical load data based on load power changes, generate a first load sequence, and calculate the power, duration, and frequency characteristics of each load event in the sequence; a sequence processing module, used to filter the first load sequence according to the power and duration characteristics, remove short-term peak load events, and obtain a target load sequence; a capacity optimization module, used to input a preset virtual battery model based on the target load sequence, perform time-series tracking simulation under preset state-of-charge safety boundary constraints, calculate the unit throughput loss cost based on the generated actual charge and discharge response trajectory; combine electricity price data and the unit throughput loss cost to calculate the marginal net benefit of energy storage, perform noise reduction processing on the mapping relationship between the marginal net benefit of energy storage and capacity, and determine the optimal energy storage configuration capacity based on the inflection point of the marginal net benefit growth rate; and a verification output module, used to simulate the impact of preset load fluctuation scenarios on daily net benefit, verify the optimal energy storage configuration capacity based on the simulation results, and output an energy storage potential identification report.

[0017] An energy storage planning and identification device based on fatigue sensitivity and marginal utility includes a memory and a processor: the memory is used to store a program; the processor is used to execute the program to implement the various steps of the energy storage planning and identification method based on fatigue sensitivity and marginal utility.

[0018] The technical effects and advantages of this invention's energy storage planning and identification method based on fatigue sensitivity and marginal utility are as follows: This invention effectively eliminates the interference of transient impacts on system configuration calculations by eliminating short-term peak-type invalid load events based on power and duration characteristics, thus avoiding redundancy in energy storage converter power configuration and erroneous responses of control strategies to false targets. By introducing a virtual battery model to quantify the degree of battery life loss under dynamic operating conditions and combining it with the inflection point of system efficiency change rate to determine the optimal capacity, it overcomes the shortcomings of traditional linear models that cannot characterize the nonlinear electrochemical decay of batteries, and achieves accurate matching between energy storage capacity and load physical characteristics. Finally, the configuration capacity is verified through load fluctuation scenario simulation, which significantly improves the system configuration scheme's ability to balance state of charge (SoC) and its operational robustness in the face of random disturbances in production conditions. Attached Figure Description

[0019] Figure 1 A schematic diagram of the energy storage planning and identification method based on fatigue sensitivity and marginal utility provided in an embodiment of the present invention; Figure 2 This is a scatter plot of the load event filtering effect based on PDF features in an embodiment of the present invention; Figure 3 This is a schematic diagram of the optimal capacity search process determined based on the marginal net revenue inflection point in an embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the verification of the risk probability distribution based on Monte Carlo simulation in an embodiment of the present invention; Figure 5 This is a block diagram of an energy storage planning and identification system based on fatigue sensitivity and marginal utility, provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1, Figure 1 This invention presents an energy storage planning and identification method based on fatigue sensitivity and marginal utility, comprising the following steps: S1, extract load events from historical load data based on load power changes, generate a first load sequence, and calculate the power, duration, and frequency characteristics of each load event in the sequence.

[0022] It should be noted that in this embodiment, high-frequency load data of the user within a preset historical period is obtained through smart metering devices such as AMI smart meters or SCADA systems installed on the industrial user side. To ensure accurate identification of instantaneous impacts and short-term spikes, the sampling frequency is set to the minute level, and the acquired data is recorded as a time series set. After acquiring the raw data, instead of directly performing statistical averaging, an event-driven preprocessing logic is used to transform the continuous time series into a discrete set of physical events using waveform slicing technology.

[0023] Specifically, the process of generating the first load sequence is as follows: 1) Set a baseline power level based on the user's transformer capacity, historical average load, or basic electricity demand declaration. The baseline is a preset power level used to distinguish between basic electrical load and fluctuating characteristic load.

[0024] 2) Traverse historical load data Extracting data that continuously exceeds the baseline power level. The waveform segments, each of which is truncated, are defined as an independent load event. All extracted load events are arranged in chronological order to form the first load sequence.

[0025] The above processing method can separate the fluctuations caused by equipment start-up and shutdown and high-power operation in industrial production from the background load, thereby focusing on the key fluctuation segments that truly affect energy storage configuration and battery life.

[0026] Furthermore, after generating the first load sequence, for each load event in the sequence... Perform multidimensional feature extraction and calculate its power features respectively. Duration characteristics and frequency characteristics Among them, power characteristics Characterizing the intensity and duration of the event. Characterizing the span and frequency features of the event It represents the degree to which events with similar characteristics repeat within historical cycles.

[0027] The power characteristics and duration characteristics The calculation formula is as follows: , , In the formula, Indicates the first The duration of each load event This indicates that the waveform of this load event has crossed the base power baseline for the first time. At that moment, This indicates that the waveform of this load event has crossed below and returned to the baseline power level for the first time. The moment; Indicates the first The peak power (i.e., maximum power value) of a load event over its duration. For the original high-frequency load data at time The power value.

[0028] For frequency characteristics To avoid statistical failure due to minute numerical differences, a statistical method based on feature similarity is adopted. Specifically, a power tolerance range is set. and time tolerance range Statistically, all load sequences in the first load sequence that satisfy the power requirement are... And the duration is The total number of events within the range, after normalization, is used as the frequency characteristic of the load event. .

[0029] The power, duration, and frequency characteristics calculated through the above steps together constitute a "three-dimensional fingerprint" describing the load event. The technical advantage of calculating these characteristics lies in the fact that traditional methods only focus on the total electrical charge generated by the change in power over time, ignoring the physical shape of the load waveform; while this invention extracts... and It can keenly detect peak loads that are "extremely high in power but extremely short in duration." These types of loads often lead to frequent high-rate charging and discharging of energy storage systems without generating significant peak-shaving benefits. Simultaneously, frequency characteristics are introduced. It can determine whether a certain fluctuation is occasional (such as annual maintenance) or normal (such as daily shift start-up), providing data support for subsequent steps to screen effective loads and assess battery cycle life.

[0030] S2, the first load sequence is filtered according to the power and duration characteristics, and short-term peak load events are removed to obtain the target load sequence.

[0031] It should be noted that in this embodiment, the first load sequence based on the "three-dimensional fingerprint" of load events generated in step S1 needs to undergo further data cleaning and validity verification. The purpose is to eliminate "pseudo-demand" fluctuations that have no economic value or technical feasibility for the energy storage system. Since industrial loads often include peak loads caused by motor starting, arcing, or measurement noise, these loads, although having extremely high instantaneous power, have extremely short durations. If energy storage is configured according to such loads, it will lead to inflated system power and idle capacity, and is highly susceptible to damage to battery life due to high-rate instantaneous discharge. Therefore, this step uses preset physical constraints to traverse and filter the first load sequence.

[0032] Specifically: The process of filtering the first load sequence to obtain the target load sequence is as follows: 1) Set two judgment thresholds, namely the preset time threshold. and preset power threshold Among them, the time threshold Power thresholds are typically determined based on the minimum response time of the energy storage battery (e.g., PCS ramp-up time) or the shortest settlement period in the electricity market; It is determined based on the transformer's rated capacity or the user's base load level, and is used to define high power fluctuations.

[0033] 2) For each load event in the first load sequence Perform a traversal and read its duration feature. and power characteristics It also performs dual condition judgments to exclude short-term peak load events.

[0034] The judgment logic for removing short-term peak load events is as follows: if a load event simultaneously meets the characteristics of "extremely short duration" and "extremely high power," it is determined to be a short-term peak load event (or an invalid load event). Once determined to be a short-term peak load event, the event is physically removed from the sequence or marked as invalid and does not participate in subsequent capacity calculations; conversely, if a load event does not meet the above removal conditions, it is retained. All load events retained after screening are reorganized in their original time order to form the target load sequence.

[0035] The judgment condition for the short-term peak load event is expressed as follows: , In the formula, Indicates the first The duration of each load event Indicates the first Peak power of a load event over its duration For the preset time threshold, This is a preset power threshold. To more intuitively demonstrate the effect of the above filtering logic, this embodiment visualizes the characteristics of all load events in the first load sequence. For example... Figure 2 As shown in the figure, the horizontal axis represents the duration of the load event. The vertical axis represents the peak power of the load event. The intensity of the color of each scatter point represents the frequency of that event. The area in the upper left corner, framed by the red dashed line in the diagram, is the "short-term peak load" area (i.e., meeting the criteria). and It can be seen that although the power values ​​of the scattered points falling into this area are very high, their duration is extremely short, and they are typical ineffective loads; while the scattered points outside the area form an effective target load sequence.

[0036] Through the above screening steps, the target load sequence retains only those effective load events (i.e., "platform-type" or "long-term" fluctuations) with the potential for "peak shaving and valley filling" and "demand management." The technical advantage of this process is that it filters out a large amount of interfering data through low-computational-cost logical judgments before entering the complex economic calculation model. This improves the convergence speed of subsequent algorithms and avoids the problem of inflated energy storage configuration costs caused by blindly tracking instantaneous peaks, ensuring that the ultimately identified energy storage potential is based on the premise that "the battery can respond and there are economic benefits after the response."

[0037] S3. Input the target load sequence into a preset virtual battery model to calculate the unit throughput loss cost; combine the electricity price data with the unit throughput loss cost to calculate the marginal net benefit of energy storage, and determine the optimal energy storage configuration capacity based on the inflection point of the marginal net benefit growth rate.

[0038] It should be noted that, in this embodiment, to ensure the accuracy of the evaluation results, the operating parameters of the virtual battery model need to be determined before inputting the target load sequence into the virtual battery model. This involves classifying and identifying the user's electricity consumption characteristics. Specifically, using the power characteristics, duration characteristics, and frequency characteristics of all load events in the first load sequence obtained in step S1 as input vectors, a density clustering algorithm (such as DBSCAN or K-Means) is used to perform cluster analysis on the target industrial users. Based on the clustering results, users are divided into stable arbitrage users (characterized by high frequency, small and regular power fluctuations), shock smoothing users (characterized by extremely high power and short duration), and demand-sensitive users (characterized by obvious monthly extreme values ​​and low frequency). For different types of users, the system will automatically match differentiated virtual battery model operating parameters. For example, for shock smoothing users, the model will be set to have a higher charge-discharge rate (C-rate) and a faster response time; while for stable arbitrage users, the model focuses on deep cycle (DOD) lifetime simulation. To more intuitively illustrate the differentiated parameters matched for different user types, this embodiment provides Table 1 as a preferred example of parameter configuration. It should be noted that the values ​​in the table are only reference values ​​for typical scenarios. In practical applications, the system can dynamically adjust the parameters according to the specific specifications of the battery cells.

[0039] Table 1

[0040] After determining the model parameters, the selected target load sequence is input into a preset virtual battery model. The core task is to quantify the battery's "hidden cost," i.e., calculate the cost per unit throughput loss. Traditional calculations often ignore the nonlinear degradation of battery life, while this embodiment uses the Rainflow-counting Algorithm to accurately calculate the battery's fatigue level. In an optional implementation, to address the distortion of the state of charge (SoC) caused by ignoring physical boundaries in long-duration, high-power discharge scenarios, a time-series tracking simulation is performed under preset SoC safety boundary constraints. The cost per unit throughput loss is calculated based on the generated actual charge-discharge response trajectory. The specific process is as follows: 1) Set the physical boundary parameters of the virtual battery model, including the rated capacity. Charge and discharge efficiency and the minimum threshold for the state of charge ; 2) Within the set time step Within, the theoretical peak-shaving power demand in the target load sequence. Physical constraints are imposed. Specifically, this involves combining the remaining battery power at the current moment with the maximum inverter power. Calculate the actual response discharge power subject to physical constraints. The calculation formula is as follows: , Where Soc(t) is the state of charge at time t; 3) Based on the actual response discharge power Update the actual state of charge at each time step This generates a continuous sequence of actual state of charge (SoC) fluctuations; 4) The actual state of charge fluctuation sequence is statistically analyzed using the rainflow counting method to determine the cycle depth and the corresponding number of cycles; 5) Based on the battery manufacturer's lifespan degradation curve (i.e., SOH-Cycle curve), calculate the total battery lifespan loss required to complete the load sequence tracking task. To convert the technical indicators into economic indicators, multiply the total lifespan loss by the total hardware purchase cost of the battery system to obtain the total depreciation cost, and then divide it by the total amount of electricity handled by the battery during the process to obtain the unit throughput loss cost.

[0041] The formula for calculating the unit throughput loss cost is as follows: , In the formula, Cost per unit throughput loss The total hardware investment cost of the energy storage system, The number of levels for the cycle depth identified by the rainflow counting method. In the first The actual number of iterations under the specified loop depth. For the battery in the first File cycle depth The corresponding theoretical maximum cycle life is as follows. This represents the total amount of electricity accumulated and pumped by the battery during the simulation.

[0042] After obtaining the unit throughput loss cost, fatigue sensitivity analysis is required to address the problem of nonlinear increases in battery life loss caused by blindly pursuing low demand targets in traditional methods. Specifically, the optimal demand control threshold needs to be determined before determining the configured capacity, as follows: 1) Set the initial demand control threshold (take the user's historical maximum load value), and use a preset step size ( The threshold is gradually reduced to generate multiple sets of simulated control strategies.

[0043] 2) For each strategy group, simulate the operation of the energy storage system and calculate the corresponding battery life loss increment using the aforementioned rainflow counting method, while also calculating the corresponding demand cost savings increment. Based on this, define and calculate the fatigue sensitivity index function value.

[0044] The formula for calculating the fatigue sensitivity index is as follows: , In the formula, This is the current demand control threshold; When the threshold is from Reduce to At that time, the resulting increased cost of battery life reduction (based on) (conversion) This will save incremental costs on the corresponding demand-based electricity bills.

[0045] 3) Analyze the fatigue sensitivity index function value. Typically, as the threshold decreases, the load curve flattens out, forcing the battery into a deep charge / discharge state, leading to a sharp increase in lifespan costs. This is identified... Relative to threshold The inflection point where a sudden increase occurs (i.e., a sudden change in slope or a numerical step) is used to lock the threshold corresponding to this inflection point as the optimal demand control threshold, denoted as . This step ensures that the energy storage strategy achieves the best balance between economic benefits and battery health.

[0046] Based on the determined optimal demand control threshold, the process enters the optimal capacity search phase. This embodiment abandons the traditional one-step linear programming solution method and instead adopts an iterative method that is more in line with the principles of marginal economics. Specifically, local electricity price data is introduced, including peak and off-peak electricity price periods and prices, the maximum demand electricity price standard, and marginal net revenue is defined. To improve the search efficiency, all load events in the target load sequence are first prioritized based on comprehensive power characteristics.

[0047] The comprehensive power characteristic is a weighted sum of the power, duration, and frequency characteristics of the load event. The purpose of calculating this characteristic is to comprehensively evaluate the cost-effectiveness of eliminating the load event: high power indicates significant potential for demand savings, moderate duration indicates reasonable battery capacity requirements, and high frequency indicates recurring events with significant long-term benefits after elimination. Since power, duration, and frequency have different dimensions, each dimension needs to be normalized before calculating the weighted value. The system sorts the load events from largest to smallest based on the calculated comprehensive power characteristic value, thereby determining the priority order for energy storage capacity coverage.

[0048] The formula for calculating the comprehensive power characteristic is as follows: , In the formula, For the first The comprehensive power characteristic value of each load event; These represent the power, duration, and frequency characteristics of the event, respectively; subscripts. and These represent the maximum and minimum values ​​of the corresponding features in the target load sequence, respectively; These are the weighting coefficients for power, duration, and frequency, respectively. In this embodiment, considering the high-return characteristics of demand management, the following settings are configured: ( ).

[0049] After sorting, the energy storage capacity is gradually increased in preset steps (50kW or 100kWh) to sequentially cover the load events ranked higher. It should be noted that when calculating the electricity cost savings for each increase, the optimal demand control threshold should be used. This is a target constraint. That is, the discharge strategy of the energy storage system is limited to: discharging only when the load power exceeds... Discharge peak clipping is performed during the process, and the peak clipping target does not exceed [the specified value]. .

[0050] Under these constraints, calculate the effective electricity cost savings resulting from this incremental capacity (including those based on...). (Demand cost savings under constraints and peak-valley price arbitrage profits), minus the corresponding throughput loss cost incurred during the operation of this incremental capacity (i.e., throughput power multiplied by) Thus, the marginal net revenue of the capacity node is obtained.

[0051] Plot a curve showing the relationship between energy storage capacity (x-axis) and marginal net revenue (y-axis). Identify the slope of this curve and find the inflection point of the marginal net revenue growth rate. Typically, as capacity increases, the high-value (high-value) assets that can be eliminated decrease. As value events become less frequent, marginal returns will show a decreasing trend.

[0052] In an optional implementation, to address the robustness deficiency caused by sawtooth fluctuations in the revenue curve due to discrete load events, which could lead to premature misjudgment of capacity extremes, the optimal energy storage configuration capacity is determined based on the continuous decreasing inflection point of the marginal net revenue growth rate after denoising the mapping relationship between the energy storage marginal net revenue and capacity using a polynomial smoothing filtering algorithm. This specifically includes the following steps: 1) Calculate the original net marginal revenue under each iteration capacity: , In the formula, Indicates the configured capacity is Total net marginal revenue at that time; For the corresponding capacity The total amount of demand-based electricity cost savings that can be obtained from the next simulation run, this value varies with The increase exhibits non-linear growth; For electricity arbitrage profits; This corresponds to the total throughput power; 2) A polynomial smoothing filter algorithm is used to denoise the discrete revenue curves: Constructing a system with different capacity steps Corresponding The discrete relationship curves are then smoothed using a Savitzky-Golay (SG) filter with local polynomial least squares fitting. Specifically, the sliding window size is set to 5 steps, and the polynomial order is 2, resulting in a smoothed continuous curve of marginal net revenue. 3) Determine the inflection point based on the smoothed curve: Take the first derivative of the smoothed curve and second derivative When the first derivative (growth rate) of the total net income curve begins to decline significantly, i.e., the second derivative... The absolute value exceeds the preset negative threshold. When the optimal energy storage configuration capacity is determined, the capacity corresponding to that point is then defined as the optimal energy storage configuration capacity. The determination formula is: and , To further prevent misjudgments caused by single-point data jumps, this embodiment upgrades the above judgment condition to a continuous interval judgment: when the smoothed second derivative satisfies the above condition within m consecutive set iteration steps (i.e., satisfies it m times consecutively). When the starting point of the continuous interval is determined, the capacity corresponding to that point is determined as the optimal energy storage configuration capacity.

[0053] The beneficial effects of this determination method are twofold. On the one hand, it prioritizes the most cost-effective load based on comprehensive value ranking, avoiding the disproportionately high energy storage costs in order to eliminate low-value long-tail loads, and finding the capacity with the highest return on investment. On the other hand, through the synergistic effect of polynomial filtering and continuous multi-step judgment conditions, it effectively eliminates local data jumps caused by discrete load events and filters out false inflection point misjudgments caused by single-point data anomalies. This makes the optimization algorithm still highly robust under extreme discrete noise conditions, and the optimal capacity obtained is closer to the true global optimum.

[0054] Based on the above iterative calculation process, the nonlinear relationship between capacity and economic benefits is plotted. For example... Figure 3As shown in the figure, the dual Y-axis chart displays the cumulative net income curve and the marginal net income growth rate curve, respectively. With the increase in energy storage capacity, the cumulative net income shows a trend of first rising rapidly and then leveling off. The marked dots in the figure represent sudden changes in the marginal net income growth rate that fall below a preset threshold. The location of this inflection point. The left side of this inflection point represents the "high return on investment zone," and the right side represents the "inefficient and redundant zone." The capacity selected at this inflection point in this embodiment (e.g., at 660kWh as shown in the figure) effectively avoids over-investment while ensuring coverage of the vast majority of high-value demand returns.

[0055] S4, simulate the impact of a preset load fluctuation scenario on daily net income, verify the optimal energy storage configuration capacity based on the simulation results, and output an energy storage potential identification report.

[0056] It should be noted that, in this embodiment, although step S3 calculates the theoretically optimal energy storage configuration capacity based on historical data, considering the high degree of uncertainty in industrial production, such as scheduling changes caused by sudden orders and efficiency fluctuations due to equipment aging, directly using historical data to draw conclusions may pose a risk of overfitting. Therefore, this step introduces a robustness verification mechanism, which constructs multi-dimensional virtual scenarios to test the stability of the configuration scheme.

[0057] Specifically, the process of verifying the optimal energy storage configuration capacity based on the simulation results is as follows: 1) Based on the target load sequence generated in step S2, 1000 sets of simulated load scenarios are generated using the Monte Carlo simulation method. To simulate the randomness of real operating conditions, the system adopts two strategies: superimposing random noise and adjusting the occurrence time of events. On the one hand, a random disturbance factor conforming to a normal distribution is superimposed on the original load power data to simulate small fluctuations in equipment operating power and generate multiple simulated load sequences; on the other hand, the start time of load events is randomly shifted to simulate the advance or delay of production shifts.

[0058] The formula for generating the simulated load scenario is as follows: , , In the formula, Indicates the generated first A simulated scenario at time... The load value, The original values ​​of the target load sequence. For random time offsets (e.g.) (evenly distributed over minutes) To conform to a mean of 0 and a variance of The random noise coefficients of a normally distributed system. The volatility is set according to the typical volatility of the user's industry (e.g., 5% to 10%).

[0059] 2) After generating a large number of simulated scenarios, the theoretically optimal energy storage configuration capacity (including power and capacity parameters) determined in step S3 is substituted into each of the above simulated load scenarios, and the preset control strategy is run. During operation, the system focuses on monitoring whether the energy storage system experiences situations where it is "power depleted (SoC=0)" or "power overflowed (SoC=100%)", resulting in an inability to respond to demand management commands. The system counts the number of scenarios in all simulated scenarios where the energy storage system's insufficient capacity or limited charging and discharging capabilities cause the user's maximum monthly demand to exceed the set target value, and calculates the probability of operational risk accordingly.

[0060] The formula for calculating the probability of operational risk is as follows: , In the formula, For risk probability, To simulate the total number of scenarios, This represents the number of scenarios in which demand management fails during simulation (i.e., the actual demand value exceeds the target demand value).

[0061] 3) After obtaining the risk probability, execute the iterative correction logic: adjust the calculated risk probability... Compared with the preset safety value Perform a comparison. If... This indicates that the current configuration capacity has sufficient robustness and can be directly adopted; if If the current configuration is too aggressive and cannot cope with production fluctuations, the system will automatically reduce the optimal energy storage capacity (in increments of 5%) and repeat the above simulation verification steps until the calculated risk probability meets the requirements. The capacity determined at this point is the final recommended capacity after noise reduction and verification. The robustness performance of the optimal configuration capacity under 1000 random scenarios generated by Monte Carlo simulation is as follows: Figure 4 As shown in the figure, the probability distribution density of the lowest possible State of Charge (SoC) value of the energy storage system on the worst-case operating day is illustrated. The red shaded area in the figure represents the risk zone where the SoC drops to 0 (i.e., the risk of demand management failure). As the statistical results show, under the current configuration, only 1.2% of the scenarios fall into the red risk zone, which is lower than the preset 5% safety threshold, thus verifying the reliability of the configured capacity under complex and fluctuating environments. If the red area accounts for too large a proportion, the configured capacity needs to be increased according to the aforementioned logic.

[0062] Finally, by integrating all the above calculation and verification results, a user-side energy storage potential identification report is automatically generated. Specific outputs include: the final revised optimal energy storage configuration capacity, the predicted investment return period and internal rate of return range based on this configuration, and the risk assessment results under load fluctuation scenarios. This report is sent to the user terminal or the energy service provider's management platform via a standardized communication interface (such as JSON format or PDF file), providing quantitative basis for investment decisions on energy storage projects.

[0063] Example 2, Figure 5 An energy storage planning and identification system based on fatigue sensitivity and marginal utility is presented, including: The feature extraction module is used to extract load events from historical load data based on load power changes, generate a first load sequence, and calculate the power, duration, and frequency characteristics of each load event in the sequence. The sequence processing module is used to filter the first load sequence according to the power and duration characteristics, and after removing short-term peak load events, obtain the target load sequence. The capacity optimization module is used to input a preset virtual battery model based on the target load sequence, perform time-series tracking simulation under preset state-of-charge safety boundary constraints, calculate the unit throughput loss cost based on the generated actual charge and discharge response trajectory, calculate the marginal net benefit of energy storage by combining the electricity price data and the unit throughput loss cost, denoise the mapping relationship between the marginal net benefit of energy storage and the capacity, and determine the optimal energy storage configuration capacity based on the inflection point of the marginal net benefit growth rate. The verification output module is used to simulate the impact of preset load fluctuation scenarios on daily net income. After verifying the optimal energy storage configuration capacity based on the simulation results, it outputs an energy storage potential identification report.

[0064] Example 3: An energy storage planning and identification device based on fatigue sensitivity and marginal utility, comprising a memory and a processor: the memory is used to store a program; the processor is used to execute the program to implement any of the embodiments in Example 1.

[0065] Since the energy storage planning and identification device based on fatigue sensitivity and marginal utility described in this embodiment is the device used to implement the method in Embodiment 1 of this invention, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in Embodiment 1 of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application falls within the scope of protection of this application.

[0066] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0067] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0068] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0069] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0070] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0071] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for energy storage planning and identification based on fatigue sensitivity and marginal utility, characterized in that, Includes the following steps: Load events are identified based on historical load data, and short-term spikes are removed to generate a target load sequence containing the power and duration characteristics of each event. The target load sequence is input into a preset virtual battery model, and a time-series tracking simulation is performed under preset state-of-charge safety boundary constraints. Based on the generated actual charge and discharge response trajectory, the unit throughput loss cost is calculated as follows: The rainflow counting method was used to determine the cycle depth and corresponding number of cycles of the virtual battery model when tracking the target load sequence. Calculate the total battery life loss based on cycle depth, number of cycles, and battery life degradation curve; The total lifetime loss is converted into total depreciation cost, and combined with the total throughput of the target load sequence, the unit throughput loss cost is calculated. By combining electricity price data with the unit throughput loss cost, the marginal net benefit of energy storage is calculated. After denoising the mapping relationship between the marginal net benefit of energy storage and capacity, the optimal energy storage configuration capacity is determined based on the inflection point of the marginal net benefit growth rate. By simulating the impact of preset load fluctuation scenarios on daily net income, the optimal energy storage configuration capacity is verified, and an energy storage potential identification report is output.

2. The fatigue sensitivity and marginal utility based energy storage planning and identification method of claim 1, wherein, The process of generating the target load sequence by removing short-term spike interference includes: Iterate through each identified load event to obtain its power and duration; The target load sequence is generated by filtering based on preset power and duration conditions, removing load events that meet the characteristics of short-term spikes.

3. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 1, characterized in that, The determination of the optimal energy storage capacity based on the inflection point of the marginal net revenue growth rate includes: The load events in the target load sequence are sorted according to a preset sorting rule, which is determined based on a weighted composite value of the power, duration and frequency characteristics of the load events. Based on the ranking results, the energy storage capacity is increased iteratively and the corresponding marginal net benefit is evaluated. The relationship curve between capacity and net benefit is plotted. The marginal net benefit is the difference between the electricity cost savings and the loss cost calculated based on the unit throughput loss cost. The optimal energy storage capacity is determined based on the changing trend of the marginal net revenue.

4. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 1, characterized in that, Before determining the optimal energy storage configuration capacity, the method further includes determining the optimal demand control threshold, the specific steps of which include: Set an initial demand control threshold, and gradually decrease the threshold with a preset step size to generate multiple sets of simulated control strategies; For each set of simulated control strategies, the corresponding fatigue sensitivity index is calculated using the rainflow counting method. The fatigue sensitivity index is the ratio of the increase in battery life loss to the increase in demand electricity cost savings. The threshold corresponding to the inflection point where the fatigue sensitivity index undergoes a sudden increase relative to the demand control threshold is determined as the optimal demand control threshold.

5. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 1, characterized in that, The verification of the optimal energy storage configuration capacity includes: A random perturbation is applied to the target load sequence to generate multiple simulated load sequences; Under the simulated load sequence, assess the probability of operational risk for the pre-selected energy storage configuration capacity; The pre-selected energy storage configuration capacity is adjusted based on the comparison results between the operational risk probability and the preset threshold.

6. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 1, characterized in that, Before inputting the target load sequence into the preset virtual battery model, the method further includes: determining the operating parameters of the virtual battery model, specifically including: Based on the power, duration, and frequency characteristics of all identified load events, users are clustered to form different user categories; Match virtual battery model operating parameters to each user category to their energy consumption characteristics.

7. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 1, characterized in that, The output energy storage potential identification report includes the output of the optimal energy storage configuration capacity, the expected investment return period, and the risk assessment results under the load fluctuation scenario.

8. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 1, characterized in that, The step of performing time-tracking simulation under preset state-of-charge safety boundary constraints, and calculating the unit throughput loss cost based on the generated actual charge-discharge response trajectory, includes: Based on the theoretical peak-shaving power demand of the load event, combined with the remaining power and maximum inverter power at the current moment, the actual response discharge power is calculated. The actual state of charge is updated based on the actual response discharge power to generate an actual state of charge fluctuation sequence. The cycle depth and corresponding cycle number of the fluctuation sequence are statistically analyzed using the rainflow counting method, and the unit throughput loss cost is calculated in combination with the battery life decay curve.

9. The energy storage planning and identification method based on fatigue sensitivity and marginal utility according to claim 3, characterized in that, The step of determining the optimal energy storage capacity based on the changing trend of the marginal net revenue includes: The relationship curve between capacity and net income is smoothed by local polynomial least squares fitting using a Savitzky-Golay filter to eliminate local data jumps caused by discrete load events. The smoothed curve is differentially calculated to obtain the second derivative sequence. When the second derivative sequence is lower than the preset negative threshold in a continuous interval, the capacity corresponding to the starting point of the continuous interval is taken as the optimal energy storage configuration capacity.

Images