Power peak load-oriented self-adaptive optimization method and system for charging and discharging duration of energy storage power station

Abstract

The invention discloses a power peak load-oriented self-adaptive optimization method and system for charging and discharging duration of an energy storage power station, and the method comprises the steps: obtaining real-time load data and energy storage unit state information from a power grid through a sensor network, and obtaining a current load curve and a battery residual capacity value; a time sequence prediction model is adopted to analyze the peak starting time and the initial intensity index obtained through judgment, and the predicted duration and the intensity change trend of the peak load are obtained; for the obtained peak load prediction duration and the intensity change trend, calculating the maximum dischargeable duration of each unit by combining the battery residual capacity value of the energy storage unit, and determining a preliminary discharge duration distribution scheme; and generating a control instruction sequence according to the adjusted discharge duration sequence obtained by optimization and the battery residual capacity value monitored in real time, and determining a final charge and discharge scheduling plan.

Description

Technical Field

[0001] This invention belongs to the field of information processing technology, specifically relating to an adaptive optimization method and system for the charging and discharging duration of an energy storage power station oriented towards peak power loads. Background Technology

[0002] Energy storage technology plays a crucial role in power systems, balancing power supply and demand, alleviating grid pressure, and ensuring the stability and reliability of energy supply. With the rapid development of renewable energy and the continuous growth of electricity demand, the role of energy storage systems in handling peak loads is particularly prominent. Peak loads typically occur during specific periods, such as summer afternoons or winter evenings, due to the concentrated use of air conditioning or heating equipment, leading to a surge in grid load. Energy storage systems effectively smooth load fluctuations by charging during off-peak hours and discharging during peak hours. However, current energy storage technologies still have significant limitations when facing complex and ever-changing power load scenarios, necessitating more efficient operating strategies to improve system performance.

[0003] Existing energy storage systems often struggle to accurately adapt to dynamic load changes during peak load periods. Many methods rely on fixed charge / discharge schedules, lacking the ability to respond quickly to real-time fluctuations in grid load. For example, when the duration or intensity of peak loads changes, the energy storage system may fail to adjust its charge / discharge duration in time, leading to low energy utilization efficiency and even the risk of overcharging or over-discharging. In practice, this manifests as some energy storage units aging faster and having shorter lifespans due to frequent deep charge / discharge cycles, while other units may be underutilized, resulting in resource waste.

[0004] The core technical challenge of energy storage systems under peak load scenarios lies in how to dynamically adjust the charging and discharging duration of energy storage units based on real-time changes in grid load. This challenge stems primarily from the unpredictability of grid load. The occurrence and duration of peak loads are often influenced by various factors such as weather and user behavior, making them difficult to predict accurately using static models. This unpredictability further leads to safety concerns regarding the operation of energy storage systems, as excessively long or short charging and discharging times can cause battery overheating or capacity degradation. For example, during a sudden peak load event, if the energy storage system fails to complete efficient discharge within a short time, the grid may face the risk of short-term power shortages; conversely, excessively long discharge times can lead to excessive battery wear, affecting the long-term economic viability.

[0005] Therefore, how to dynamically optimize the charging and discharging duration of each energy storage unit based on real-time monitoring of grid load and the technical characteristics of the energy storage system, such as maximum charging and discharging power and battery cycle life, has become a key issue for improving the operational efficiency and safety of energy storage systems under peak load scenarios. This issue not only concerns the stable operation of the power grid but also directly affects the service life and economic benefits of energy storage equipment. Summary of the Invention

[0006] To address the problems existing in the prior art, the present invention provides a method, apparatus, system, and storage medium.

[0007] To achieve the above objectives, the present invention provides the following solution: An adaptive optimization method for charging and discharging duration of energy storage power stations oriented towards peak power loads includes: Real-time load data and energy storage unit status information are obtained from the power grid through a sensor network to obtain the current load curve and the remaining battery capacity value; Based on the current load curve, the peak start time and initial intensity index are obtained; Based on the start time of the peak load and the initial intensity index, the predicted duration and intensity trend of the peak load are obtained. Based on the predicted duration and intensity of peak load and the remaining battery capacity, a preliminary discharge duration allocation scheme is determined. If the maximum discharge duration of a certain unit in the calculated preliminary discharge duration allocation scheme is less than the predicted duration, the preliminary discharge duration allocation scheme is iteratively optimized using a genetic algorithm to obtain the adjusted discharge duration sequence. Based on the optimized and adjusted discharge duration sequence and the real-time monitored remaining battery capacity, a control command sequence is generated to determine the final charge and discharge scheduling plan.

[0008] Preferably, the current load curve is compared with a preset peak threshold. If the current load curve exceeds the preset peak threshold, it is determined that the peak load state has been entered, and the peak start time and initial intensity index are obtained.

[0009] As a preferred approach, a time series forecasting model is used to analyze the peak start time and initial intensity index obtained from the judgment, so as to obtain the predicted duration and intensity change trend of the peak load.

[0010] As a preferred approach, based on the predicted duration and intensity trends of peak loads, and combined with the remaining battery capacity of the energy storage units, the maximum discharge duration of each unit is calculated, and a preliminary discharge duration allocation scheme is determined.

[0011] This invention also provides an adaptive optimization system for the charging and discharging duration of an energy storage power station oriented towards peak power loads, comprising: The first processing module is used to obtain real-time load data and energy storage unit status information from the power grid through a sensor network, and to obtain the current load curve and the remaining battery capacity value. The second processing module is used to obtain the peak start time and initial intensity index based on the acquired current load curve; The third processing module is used to obtain the predicted duration and intensity change trend of the peak load based on the peak start time and initial intensity index. The fourth processing module is used to determine the initial discharge duration allocation scheme based on the predicted duration and intensity trend of peak load and the remaining battery capacity. The fifth processing module is used to iteratively optimize the preliminary discharge duration allocation scheme by using a genetic algorithm if the maximum discharge duration of a certain unit in the calculated preliminary discharge duration allocation scheme is less than the predicted duration, so as to obtain the adjusted discharge duration sequence. The sixth processing module is used to generate a control command sequence and determine the final charge and discharge scheduling plan based on the optimized and adjusted discharge duration sequence and the real-time monitored remaining battery capacity.

[0012] Preferably, the second processing module compares the acquired current load curve with a preset peak threshold. If the current load curve exceeds the preset peak threshold, it determines that it has entered a peak load state and obtains the peak start time and initial intensity index.

[0013] As a preferred option, the third processing module uses a time series prediction model to analyze the peak start time and initial intensity index obtained from the judgment, and obtains the predicted duration and intensity change trend of the peak load.

[0014] As a preferred option, the fourth processing module calculates the maximum discharge time of each unit based on the predicted duration and intensity trend of the peak load, combined with the remaining battery capacity of the energy storage unit, and determines the preliminary discharge time allocation scheme.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention determines the peak load state when the load curve exceeds a preset peak threshold and captures the start time and initial intensity. It uses a time series prediction model to analyze the duration and intensity trends, and integrates the remaining battery capacity to calculate the maximum discharge time of each cell to form a preliminary allocation scheme. If any cell has insufficient discharge time, a genetic algorithm is introduced to iteratively optimize and adjust the discharge sequence. Finally, a charging and discharging scheduling plan is formed based on the optimized sequence and real-time capacity to generate control commands. This invention achieves accurate peak load prediction and dynamic optimization allocation of energy storage resources, ensuring stable peak shaving of the power grid and improving energy storage utilization efficiency. Attached Figure Description To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of an adaptive optimization method for charging and discharging duration of an energy storage power station for peak power loads, according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Example 1 like Figure 1 As shown, this invention provides an adaptive optimization method for the charging and discharging duration of an energy storage power station oriented towards peak power loads, comprising: S101. Obtain real-time load data and energy storage unit status information from the power grid through a sensor network to obtain the current load curve and battery remaining capacity value.

[0020] Real-time load data and energy storage unit status information are acquired from the power grid via a sensor network to obtain a preliminary load sequence and battery capacity indicators. For the preliminary load sequence, a moving average data smoothing method is used to process noise, resulting in a smoothed load sequence. Peak and trough time series features are extracted from the smoothed load sequence to determine the current load curve. Based on the battery capacity indicators and the current load curve, if load fluctuations exceed a preset threshold, the response is adjusted to determine the energy storage response demand and obtain the remaining battery capacity value.

[0021] For example, in smart grid systems, sensor networks deployed in substations and distribution lines can collect real-time load data and energy storage unit status information. These sensors, including current transformers and voltage sensors, monitor the grid's power consumption and battery charging status every second, thus forming preliminary load sequences, such as the load value sequence per minute within an hour, and battery capacity indicators, such as the current remaining percentage of charge. The data acquired in this way provides the foundation for subsequent analysis, ensuring the system's real-time responsiveness.

[0022] In one possible implementation, a moving average data smoothing method is used to handle noise in the preliminary load sequence. This method involves calculating the average value for each point in the sequence, for example, taking the average of the first 5 and last 5 data points as the smoothed value for the current point, thereby eliminating fluctuations caused by sensor errors or sudden disturbances. For instance, if the preliminary sequence shows that the load suddenly jumps to an abnormally high value at a certain moment, a moving average can smooth it to a value closer to the actual trend, ultimately resulting in a smoothed load sequence. This helps to accurately reflect the true operating state of the power grid and reduces the risk of misjudgment.

[0023] For example, the process of extracting peak and trough time series features from a smoothed load sequence can be achieved by scanning the sequence to identify local maxima and minima. For instance, in a daily load curve, peaks might occur during the peak electricity consumption period at 8 PM, while troughs occur during the low-demand period at 3 AM. These features are used to determine the current load curve, such as whether it exhibits a typical "duck curve" shape, where daytime solar power generation leads to deeper load troughs. This extraction helps identify load patterns and provides a basis for energy storage management decisions.

[0024] In one possible implementation, the response strategy is adjusted based on the battery capacity index and the current load curve, if the load fluctuation exceeds a preset threshold such as 10%.

[0025] Specifically, if the curve indicates that a peak load is imminent and the remaining battery capacity is below 50%, the system will determine that immediate discharge is needed to balance the grid, and will calculate an adjusted remaining battery capacity value, for example, reducing it from the initial 80% to an adjusted 60%. This judgment mechanism ensures the optimal use of the energy storage unit and avoids overload situations.

[0026] For example, in practical applications, this technology can be used in urban microgrids. In a residential power grid in Beijing, a sensor network collects load sequences during peak summer air conditioning seasons. After smoothing, the peak value at 5 PM is extracted. If fluctuations exceed a threshold, the system releases stored energy based on battery capacity indicators to support peak reduction, ultimately yielding the remaining capacity. This not only stabilizes power supply but also reduces energy costs.

[0027] In one possible implementation, the continuity of the entire process lies in the data acquisition and processing. For example, the smoothing of the initial sequence directly affects the accuracy of peak extraction, while curve determination guides response judgment. This logical chain ensures a complete analytical path from real-time data to capacity values, achieving efficient energy management goals in business operations.

[0028] S102. Compare the current load curve with the preset peak threshold. If the current load curve exceeds the preset peak threshold, it is determined that the peak load state has been entered, and the peak start time and initial intensity index are obtained.

[0029] The current load curve and a preset peak threshold are obtained. If the current load curve exceeds the preset peak threshold, it is determined that a peak load state has been entered, and the peak start time and initial intensity index are obtained. Based on the peak start time and initial intensity index, the duration of the load peak is determined, and the peak decay rate is obtained.

[0030] For example, in a power grid load monitoring system, the current load curve needs to be obtained from a real-time data source, which is usually achieved through a sensor network deployed in substations.

[0031] Specifically, these sensors collect electricity demand data every minute, forming a continuous load curve. This curve is then compared to a preset peak threshold, which may be based on historical peak data, such as 1.5 times the normal load. If the current curve exceeds this threshold at a certain point in time, the system automatically determines that it has entered a peak load state. At this time, it records the start time of the peak, such as 10:00 AM, and the initial intensity index, which can be the difference between the peak and average load, used to quantify the severity of the peak. In this way, operators can promptly identify load anomalies, thus providing a basis for subsequent energy dispatching.

[0032] In one embodiment, once the peak start time and initial intensity index are determined, the system continues to monitor changes in the load curve to calculate the duration of the load peak.

[0033] Specifically, this involves tracking the curve from the beginning until it falls back below the threshold.

[0034] For example, if a peak starts at 10:00 AM and ends at 10:30 AM, then the duration is 30 minutes. During this process, the system may use time series analysis to ensure accuracy and avoid misjudgments due to short-term fluctuations. The resulting duration data not only reflects the persistence of the peak but also helps assess the stress distribution on the power grid.

[0035] For example, based on the aforementioned duration, when further calculating the peak decay rate, one can observe the speed at which the peak value decreases from its initial intensity.

[0036] Specifically, the decay rate is determined by the rate of change of peak intensity over time. For example, if the initial intensity is 200 MW and drops to 100 MW after 30 minutes, the decay rate is approximately 3.33 MW per minute. This calculation helps predict when the peak will end, thereby optimizing the discharge strategy of the energy storage unit. In practical applications, this type of analysis is often used for peak-hour management of urban power grids. For instance, during peak summer periods when air conditioning loads surge, calculating the decay rate early can guide the timing of standby generator startup and avoid blackout risks.

[0037] In one embodiment, the continuity of the entire process lies in the continuity of the data flow, from acquiring the curve to judging the status, and then to calculating the duration and rate, forming a closed-loop monitoring mechanism.

[0038] Specifically, if the start time of a spike is not recorded accurately, it will directly affect subsequent duration estimations. Therefore, the system often integrates calibration algorithms to verify the accuracy of the data.

[0039] For example, in a power grid application in an industrial area, once the initial intensity index exceeds the threshold, the system will immediately notify the control center to adjust the load distribution of the transformers based on the duration of the delay. Ultimately, the supply and demand are balanced by predicting the attenuation rate. This not only improves response efficiency but also achieves a more stable power supply target in business operations.

[0040] For example, when extended to more complex scenarios, the system can cumulatively calculate the decay rate of each peak when multiple spikes occur consecutively.

[0041] Specifically, this involves comparing the initial intensity with historical data. For example, if the previous peak decay rate was 2 MW per minute and this one is 3 MW, it indicates faster load recovery, thus guiding the dynamic allocation of energy reserves. In this approach, the operational process expands from single peak analysis to sequence analysis, ensuring the overall resilience of the power grid.

[0042] S103. The time series prediction model is used to analyze the peak start time and initial intensity index obtained by the judgment to obtain the predicted duration and intensity change trend of the peak load.

[0043] By analyzing historical load curves, the peak start time and initial intensity index are obtained. The peak occurrence probability is calculated by dividing the ratio of the peak start time to the initial intensity index by the average value of historical load curves. Based on this peak occurrence probability, a time series prediction model is employed. This model takes the peak occurrence probability and the period fluctuation amplitude as inputs and outputs the predicted duration. The period fluctuation amplitude is analyzed using an autoregressive integral moving average process to determine the predicted duration. Using this predicted duration, a rate of change estimate is calculated against the load peak threshold. This calculation determines the intensity change trend by dividing the predicted duration by the load peak threshold. If the intensity change trend exceeds the load peak threshold, trend fitting parameters are obtained from real-time monitoring signals. An extended predicted value for the load data sequence is obtained by linearly regressing the intensity change trend using these trend fitting parameters.

[0044] For example, in power load management systems, analyzing historical load curves to obtain the peak start time and initial intensity index is a crucial process.

[0045] Specifically, suppose a power grid company has historical data records where load curves show daily electricity consumption patterns. The peak start time can be defined as the point in time when the load value first exceeds a preset threshold, such as during the hot summer months when the load begins to rise sharply at 2 PM. The initial intensity index might be the difference between the load value at that moment and the baseline value. By calculating the ratio of the peak start time to the initial intensity index—for example, converting the start time to hours, such as 14, and dividing this by the difference in intensity index, such as 500 kW—and then dividing by the average of the historical load curve, such as a daily average of 300 kW, the probability of peak occurrence is obtained. This probability value might be 0.15, representing a 15% chance of peak load. This method helps grid operators assess risks in advance and avoid blind dispatching.

[0046] In one embodiment, when using a time series forecasting model based on the probability of peak occurrence, an ARIMA model, or Autoregressive Integral Moving Average, can be introduced to handle the amplitude of fluctuations over time. The core of the ARIMA model lies in capturing the dependencies between data through the autoregressive part, handling non-stationarity through the integral part, and smoothing random fluctuations through the moving average part.

[0047] For example, in a city power grid scenario, given a peak occurrence probability of 0.15 and time-period fluctuations such as hourly load variations of ±50 kW, the model analyzes the fluctuation data from the past week, uses differencing to stabilize the sequence, and then fits parameters such as autoregression order p=2, differencing order d=1, and moving average order q=1, ultimately outputting a predicted duration of 3 hours. This allows operators to simulate the persistence of peak loads, thereby optimizing the timing of generator start-up.

[0048] For example, the process of calculating the rate of change for the predicted duration can be understood as a trend judgment mechanism.

[0049] Specifically, if the predicted duration is 3 hours and the peak load threshold is 1000 kW, the calculation involves dividing the duration by the difference between the current load and the threshold. For example, if the difference is 200 kW, the rate of change is 3 / 200 = 0.015 kW per kilowatt-hour. This is used to determine whether the intensity change trend is increasing or decreasing. In practical applications, such as the power supply for an industrial park, if the trend shows an increase exceeding the threshold, this calculation can enable timely adjustments to the power distribution strategy, reducing the risk of overload.

[0050] In one embodiment, if the intensity change trend exceeds the peak load threshold, the process of obtaining trend fitting parameters from the real-time monitoring signal involves linear regression analysis.

[0051] For example, in a smart grid system, real-time signals come from minute-by-minute load data collected by sensors. Fitting parameters such as slope and intercept are extracted, and the trend is fitted using a linear regression formula. Assuming the data sequence is time t and load y, the regression yields an extended forecast value, such as the load expected to reach 1200 kilowatts in the next hour. This extended forecast helps in developing emergency response plans and ensuring power continuity.

[0052] For example, when expanding to more complex scenarios, the impact of seasonal factors on these steps can be considered.

[0053] Specifically, during the peak heating season in winter, the average value of the historical load curve may be higher, such as 400 kilowatts. This will affect the calculation of the probability of peak occurrence, causing the probability to be adjusted to 0.20. Then, the time series model will increase its sensitivity to the fluctuation range accordingly, and the predicted duration may be extended to 4 hours. This will emphasize a more conservative trend judgment in the estimation of the rate of change. The final linear regression prediction value can support the implementation of dynamic pricing strategies and optimize energy allocation efficiency.

[0054] In one embodiment, the logical progression of these steps ensures a complete chain from probability assessment to forecast extension, for example, starting with historical data acquisition and progressing to model prediction and trend analysis, with each link supporting each other to form a consistent load management framework.

[0055] S104. Based on the predicted duration and intensity trend of peak load, and combined with the remaining battery capacity of the energy storage unit, calculate the maximum discharge duration of each unit and determine the preliminary discharge duration allocation scheme.

[0056] Peak load forecasting is used to obtain intensity change trends, and the remaining battery value is calculated based on the energy storage unit capacity. The calculation process involves dividing the remaining capacity by the discharge rate to obtain the maximum discharge duration. The duration is then assessed based on the maximum discharge duration, and load forecasting is used to integrate and process the intensity change trends to determine optimal duration allocation. If the optimized duration allocation exceeds a preset threshold, trend analysis is applied to adjust the remaining battery value, resulting in a preliminary scheme determination. This preliminary scheme is then combined with capacity calculations, and real-time monitoring feedback is obtained to determine the final allocation scheme.

[0057] In one possible implementation, the process of obtaining the intensity change trend through peak load forecasting can begin by collecting historical power load data. For example, in a smart grid system, the intensity change trend of load peaks caused by air conditioning use during the summer peak season can be analyzed.

[0058] Specifically, this trend is predicted using time series models such as ARIMA, which take historical load curves as input and output intensity rise or fall curves for the next hour. For example, if the prediction shows that the intensity will rise from an initial 80% load rate to 95%, the remaining battery capacity is calculated based on the energy storage unit capacity. Here, the remaining capacity refers to the current available battery power, such as 500 kWh, and the discharge rate is a preset 200 kW per hour. The maximum discharge duration is obtained by dividing the remaining capacity by the discharge rate, i.e., 500 divided by 200 equals 2.5 hours, thereby assessing the duration for which the energy storage system can support peak loads.

[0059] For example, in the assessment of duration, the load forecast processing intensity change trend is integrated based on the maximum discharge duration, such as 2.5 hours. The specific process involves matching the forecast trend with the duration. For example, if the intensity change trend shows that the load will remain high for 1 hour, the duration allocation optimization is determined by an optimization algorithm such as linear programming to ensure that the energy storage discharge is evenly distributed and avoids overload.

[0060] In one possible implementation, if the time allocation optimization exceeds a preset threshold, for example, if the optimization result shows that the allocated time exceeds 3 hours while the threshold is 2 hours, then the remaining battery value is adjusted through trend analysis. This trend analysis can use the least squares method to fit the intensity curve and adjust the remaining value, such as from 500kWh to 450kWh, to match the actual demand, thereby obtaining a preliminary solution. For example, the preliminary solution is to prioritize discharging for 2 hours during peak periods to smooth the load.

[0061] For example, the process of determining the integrated capacity for the preliminary scheme involves calculations. Integrated capacity refers to combining the capacity of multiple energy storage units. For example, the remaining capacity of two battery packs is added together to obtain a total value of 700kWh. Then, the calculations are combined with real-time monitoring and feedback. Here, the feedback comes from sensor data, such as the current load rate of 95%, to determine whether the preliminary allocation scheme is feasible. For example, if the feedback shows that the load has dropped to 85%, the scheme is approved; otherwise, adjustments are needed to ensure grid stability.

[0062] In one possible implementation, the logical connection of the entire process lies in the progression from predicting trends to determining solutions. For example, the trend of intensity change serves as a basic input, affecting the calculation of the maximum discharge duration. The duration allocation optimization depends on this assessment. If the threshold is exceeded and adjustments are made, the fusion capacity calculation is further verified by real-time feedback to form a closed-loop optimization. In actual business applications such as urban power grid management, this can effectively reduce blackout risks and improve power supply reliability.

[0063] For example, in the extended scheme, more variables such as weather factors can be introduced to affect load forecasting, enriching the diversity of the initial scheme. For instance, if the forecast trend takes into account the increase in intensity due to high temperatures, the discharge rate can be adjusted from 200kW to 150kW, and the maximum discharge duration can be extended to 3.3 hours, supporting longer duration allocation optimization.

[0064] It should be noted that the details of trend analysis applications include linear regression of intensity change trends. For example, data points from the past 24 hours are collected, and the slope is fitted to predict future changes. If the slope is greater than 0.1, it indicates a rapid increase. The remaining battery value is adjusted accordingly to discharge earlier, resulting in a more accurate preliminary solution. This helps businesses cope with sudden spikes.

[0065] For example, real-time monitoring feedback can be obtained through IoT devices. When judging the initial allocation plan, if the deviation between the feedback data and the prediction is less than 5%, the plan is effective. Otherwise, the fusion capacity calculation needs to be iterated, such as recalculating the total capacity and allocating the duration, to ensure the consistency of the overall plan.

[0066] S105. If the maximum discharge duration of a certain unit in the calculated preliminary discharge duration allocation scheme is less than the predicted duration, the preliminary discharge duration allocation scheme is iteratively optimized using a genetic algorithm to obtain the adjusted discharge duration sequence.

[0067] The initial allocation scheme is used to obtain the maximum discharge duration and predicted duration of each unit. It is then determined whether the maximum discharge duration is less than the predicted duration, yielding a judgment result. Based on this result, if the maximum discharge duration is less than the predicted duration, a genetic algorithm is used to iterate the initial allocation scheme to determine an optimal fitness function. This optimal fitness function calculates individual fitness values ​​based on unit capacity limitations and balanced discharge allocation. Using the optimized fitness function, a set of individuals after crossover and mutation operations is obtained. These operations involve probabilistic crossover and mutation of individuals in the population during the genetic algorithm iteration. The set of individuals is then judged to meet convergence criteria, which are based on the fitness value change across generations being less than a preset threshold, resulting in elite individuals. Based on these elite individuals, a balanced scheme is selected under the unit capacity limitations. This balanced scheme is chosen by comparing the duration distribution uniformity among the elite individuals to determine the balanced discharge allocation. Finally, the optimized scheme is obtained through the balanced discharge allocation, resulting in an adjusted duration sequence.

[0068] For example, in an energy storage system, when obtaining the maximum discharge duration and predicted duration of each unit through an initial allocation scheme, a specific grid peak management scenario can be considered. Assume there are three energy storage units A, B, and C, with remaining battery capacities of 500kWh, 400kWh, and 300kWh respectively, and a discharge rate of 100kW. The maximum discharge duration is calculated by dividing the remaining capacity by the discharge rate, resulting in 5 hours for unit A, 4 hours for unit B, and 3 hours for unit C. Simultaneously, based on peak load forecasting, the predicted duration is 6 hours. At this point, it is determined that the maximum discharge duration of A, B, and C is all less than 6 hours, resulting in a "yes" judgment, thus triggering the subsequent optimization process. This judgment ensures that the system avoids the risk of insufficient capacity during actual operation.

[0069] In one possible implementation, if the maximum discharge duration is less than the predicted duration, a genetic algorithm is used to iterate the initial allocation scheme. First, an optimal fitness function is determined, which calculates the individual fitness value based on cell capacity constraints and discharge balance allocation.

[0070] Specifically, the fitness function can be designed as a weighted sum formula, where the capacity constraint part evaluates whether the duration of each unit exceeds its maximum value, and the balanced allocation part calculates the variance of the duration allocation of all units. The smaller the variance, the higher the fitness value.

[0071] For example, in the scenario described above, the initial population includes multiple allocation schemes. Scheme 1 involves discharging A for 4 hours, B for 4 hours, and C for 3 hours, which together cover the peak demand. The fitness function calculates that the balance has a low variance value, thus serving as the basis for iteration. Through multiple iterations, the algorithm gradually improves the overall efficiency of the schemes.

[0072] For example, when obtaining the set of individuals after crossover and mutation operations, the genetic algorithm performs probabilistic crossover and mutation on the individuals in the population. Assuming a crossover probability of 0.8 and a mutation probability of 0.1, in the iteration, a single-point crossover is performed on two parent individuals, such as scheme 1 and scheme 2, swapping the duration allocations of A and B to generate child individuals; then, the mutation operation randomly adjusts the duration of C by ±0.5 hours. This results in a more diverse set of new individuals, which helps in exploring the globally optimal solution. In practical applications, this operation simulates natural evolution, ensuring that the allocation scheme is progressively optimized under capacity constraints.

[0073] In one possible implementation, the convergence criterion is determined by the change in fitness value across generations being less than a preset threshold, such as 0.01. If the fitness value changes in the first three generations are 0.05, 0.02, and 0.009, the condition is met, the iteration stops, and elite individuals are selected.

[0074] For example, an elite individual might have the highest fitness value if schemes A discharge for 5 hours, B discharge for 3.5 hours, and C discharge for 3.5 hours. This convergence mechanism avoids infinite iteration, improves computational efficiency, and can respond to peak changes in real time during power grid dispatching.

[0075] For example, when selecting a balancing scheme under cell capacity constraints, the selection is made by comparing the uniformity of time distribution among elite cells. Uniformity can be quantified as a standard deviation; the cell with the smallest standard deviation is selected as the balancing scheme, thus determining the discharge balance distribution. In the example, if there are two elite cells, the first with a standard deviation of 0.5 hours and the second with 1 hour, the first one is selected to ensure balanced load across cells and extend battery life.

[0076] In one possible implementation, an optimized scheme is obtained through discharge balance allocation, resulting in an adjusted duration sequence.

[0077] For example, the final sequence is A: 4.8 hours, B: 4.2 hours, and C: 3.0 hours, which sums to match the predicted duration. This optimized solution achieves efficient energy utilization in operations and reduces grid pressure during peak periods.

[0078] S106. Based on the optimized and adjusted discharge duration sequence and the real-time monitored remaining battery capacity, generate a control command sequence and determine the final charge and discharge scheduling plan.

[0079] By adjusting the discharge sequence and monitoring the remaining battery capacity in real time, differential calculations are performed on the capacity value sequence to obtain capacity fluctuation analysis results, resulting in a preliminary energy balance state. Based on this energy balance state, a load demand forecasting method is used to fit a trend curve using historical load data to determine the criteria for safety threshold judgment and obtain a resource allocation optimization scheme. For the resource allocation optimization scheme, instruction execution logic parameters are obtained, and if the capacity fluctuation exceeds a preset threshold, the adjustment plan is iteratively updated to obtain a control instruction sequence. From the control instruction sequence, feedback loop mechanism data is obtained, and by comparing the actual execution feedback with the expected value, the charging and discharging scheduling details are determined. Based on the charging and discharging scheduling details, a final charging and discharging scheduling plan is generated.

[0080] For example, in an electric vehicle battery management system, the capacity value sequence is differentially calculated by combining the adjusted discharge sequence with the real-time monitored remaining battery capacity value to obtain the capacity fluctuation analysis results, thereby obtaining the preliminary energy balance state.

[0081] Specifically, this differential calculation involves subtracting the capacity values ​​at consecutive time points to identify the fluctuation range. For example, assuming a battery pack has a capacity sequence of 100Ah initially, followed by 98Ah and 95Ah, the fluctuation trend can be analyzed by calculating the differences -2Ah and -3Ah. If the fluctuation rate exceeds 5%, it indicates that the energy balance is unstable, which helps to initially assess the stability of the system.

[0082] In one embodiment, the application of this method can ensure that the battery discharges evenly under high load, avoiding premature depletion of individual cells, thereby improving the overall system reliability and lifespan.

[0083] For example, based on the above preliminary energy balance state, the load demand forecasting method is used to fit the trend curve through historical load data to determine the basis for judging the safety threshold and obtain the resource allocation optimization scheme.

[0084] Specifically, the load demand forecasting method can use the least squares method to fit the load data of the past week to a curve. For example, if historical data shows that the peak load reaches 50kW at 8 pm, a quadratic curve can be obtained by fitting the curve to predict the future load trend. If the predicted curve shows that the peak may exceed 60kW, a safety threshold of 55kW is set as the basis for judgment, thereby optimizing the resource allocation scheme, ensuring that energy supply matches demand, and avoiding the risk of overload.

[0085] In one embodiment, for the resource allocation optimization scheme, instruction execution logic parameters are obtained, and if the capacity fluctuation exceeds a preset threshold, the adjustment plan is iteratively updated to obtain a control instruction sequence.

[0086] For example, the instruction execution logic parameters include discharge rate and duration settings. If fluctuations exceeding 10% are detected, the plan is iteratively updated. By repeatedly adjusting the parameters, such as reducing the discharge rate from 2C to 1.5C, a series of instruction sequences are ultimately formed to guide actual operations.

[0087] For example, feedback loop mechanism data can be obtained from the control command sequence, and the charging and discharging scheduling details can be determined by comparing the actual execution feedback with the expected value.

[0088] Specifically, the feedback loop mechanism involves real-time acquisition of capacity changes after execution. For example, if the expected capacity after discharge is 80Ah, but the actual capacity is 78Ah, the details are adjusted by comparing the differences to ensure more accurate scheduling.

[0089] In one embodiment, the final charge and discharge scheduling plan is generated through these charge and discharge scheduling details, for example, by integrating all the details into a 24-hour schedule to achieve energy-efficient utilization.

[0090] Example 2 This invention also provides an adaptive optimization system for the charging and discharging duration of an energy storage power station oriented towards peak power loads, comprising: The first processing module is used to obtain real-time load data and energy storage unit status information from the power grid through a sensor network, and to obtain the current load curve and the remaining battery capacity value. The second processing module is used to obtain the peak start time and initial intensity index based on the acquired current load curve; The third processing module is used to obtain the predicted duration and intensity change trend of the peak load based on the peak start time and initial intensity index. The fourth processing module is used to determine the initial discharge duration allocation scheme based on the predicted duration and intensity trend of peak load and the remaining battery capacity. The fifth processing module is used to iteratively optimize the preliminary discharge duration allocation scheme by using a genetic algorithm if the maximum discharge duration of a certain unit in the calculated preliminary discharge duration allocation scheme is less than the predicted duration, so as to obtain the adjusted discharge duration sequence. The sixth processing module is used to generate a control command sequence and determine the final charge and discharge scheduling plan based on the optimized and adjusted discharge duration sequence and the real-time monitored remaining battery capacity.

[0091] As one embodiment of the present invention, the second processing module compares the current load curve obtained with the preset peak threshold. If the current load curve exceeds the preset peak threshold, it is determined that the peak load state has been entered, and the peak start time and initial intensity index are obtained.

[0092] As one embodiment of the present invention, the third processing module uses a time series prediction model to analyze the peak start time and initial intensity index obtained by the judgment, and obtains the predicted duration and intensity change trend of the peak load.

[0093] As one embodiment of the present invention, the fourth processing module calculates the maximum discharge time of each unit based on the obtained peak load prediction duration and intensity change trend, combined with the remaining battery capacity value of the energy storage unit, and determines the preliminary discharge time allocation scheme.

[0094] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. An adaptive optimization method for charging and discharging duration of an energy storage power station oriented towards peak power loads, characterized in that, include: Real-time load data and energy storage unit status information are obtained from the power grid through a sensor network to obtain the current load curve and the remaining battery capacity value; Based on the current load curve, the peak start time and initial intensity index are obtained; Based on the start time of the peak load and the initial intensity index, the predicted duration and intensity trend of the peak load are obtained. Based on the predicted duration and intensity of peak load and the remaining battery capacity, a preliminary discharge duration allocation scheme is determined. If the maximum discharge duration of a certain unit in the calculated preliminary discharge duration allocation scheme is less than the predicted duration, the preliminary discharge duration allocation scheme is iteratively optimized using a genetic algorithm to obtain the adjusted discharge duration sequence. Based on the optimized and adjusted discharge duration sequence and the real-time monitored remaining battery capacity, a control command sequence is generated to determine the final charge and discharge scheduling plan.

2. The adaptive optimization method for charging and discharging duration of energy storage power stations oriented towards peak power loads as described in claim 1, characterized in that, The current load curve is compared with the preset peak threshold. If the current load curve exceeds the preset peak threshold, it is determined that the peak load state has been entered, and the peak start time and initial intensity index are obtained.

3. The adaptive optimization method for charging and discharging duration of energy storage power stations oriented towards peak power loads as described in claim 1, characterized in that, A time series forecasting model was used to analyze the start time and initial intensity index of the peak load to obtain the predicted duration and intensity trend of the peak load.

4. The adaptive optimization method for charging and discharging duration of energy storage power stations oriented towards peak power loads as described in claim 1, characterized in that, Based on the predicted duration and intensity trends of peak load, and combined with the remaining battery capacity of the energy storage unit, the maximum discharge duration of each unit is calculated, and a preliminary discharge duration allocation scheme is determined.

5. An adaptive optimization system for charging and discharging duration of an energy storage power station oriented towards peak power loads, characterized in that, include: The first processing module is used to obtain real-time load data and energy storage unit status information from the power grid through a sensor network, and to obtain the current load curve and the remaining battery capacity value. The second processing module is used to obtain the peak start time and initial intensity index based on the acquired current load curve; The third processing module is used to obtain the predicted duration and intensity change trend of the peak load based on the peak start time and initial intensity index. The fourth processing module is used to determine the initial discharge duration allocation scheme based on the predicted duration and intensity trend of peak load and the remaining battery capacity. The fifth processing module is used to iteratively optimize the preliminary discharge duration allocation scheme by using a genetic algorithm if the maximum discharge duration of a certain unit in the calculated preliminary discharge duration allocation scheme is less than the predicted duration, so as to obtain the adjusted discharge duration sequence. The sixth processing module is used to generate a control command sequence and determine the final charge and discharge scheduling plan based on the optimized and adjusted discharge duration sequence and the real-time monitored remaining battery capacity.

6. The adaptive optimization system for charging and discharging duration of energy storage power stations oriented towards peak power loads as described in claim 5, characterized in that, The second processing module compares the current load curve with the preset peak threshold. If the current load curve exceeds the preset peak threshold, it determines that it has entered the peak load state and obtains the peak start time and initial intensity index.

7. The adaptive optimization system for charging and discharging duration of energy storage power stations oriented towards peak power loads as described in claim 5, characterized in that, The third processing module uses a time series prediction model to analyze the peak start time and initial intensity index obtained from the judgment, and obtains the predicted duration and intensity change trend of the peak load.

8. The adaptive optimization system for charging and discharging duration of an energy storage power station oriented towards peak power load as described in claim 5, characterized in that, The fourth processing module, based on the predicted duration and intensity trend of peak load, and combined with the remaining battery capacity of the energy storage unit, calculates the maximum discharge time of each unit and determines the preliminary discharge time allocation scheme.

Images