Multi-dimensional load prediction household energy storage charging and discharging optimization method, system and device

Abstract

The invention discloses a multi-dimensional load prediction household energy storage charging and discharging optimization method, system and device, which are used for a higher-precision load prediction energy storage charging and discharging control strategy. The method comprises the following steps: preprocessing multi-source data related to household electricity consumption to obtain an input structure, and constructing corresponding energy consumption scene parameters; on the basis of the historical power distribution characteristics and the behavior priority parameters of the energy consumption scene parameters, energy consumption scene parameter potential energy values of the discrete time periods in the prediction time window are calculated, and a corresponding potential energy curve is obtained; performing time diffusion processing on the potential energy curve by starting a time fluctuation model to obtain a potential energy evolution curve; superposing the potential energy evolution curves of all the energy consumption scene parameters in the prediction time window according to each discrete time period to obtain a family total load prediction curve; inputting data such as a household total load prediction curve and the like into the energy consumption scene parameter game scheduling model to obtain target charging and discharging power; and controlling the household energy storage battery according to the target charging and discharging power.

Description

Technical Field

[0001] This embodiment relates to the field of home energy storage management, and in particular to a multi-dimensional load prediction method, system and device for home energy storage charging and discharging optimization. Background Technology

[0002] With the large-scale application of residential distributed photovoltaic (PV) and energy storage devices, household electricity consumption patterns are gradually evolving from single loads to a comprehensive energy consumption model composed of multiple devices and scenarios. The combined effects of family members' lifestyles, the operating status of household appliances, and changes in PV power generation cause household loads to exhibit superimposed, sudden, and unstable characteristics at different times, forming electricity consumption curves with significant fluctuations and randomness. To increase the self-consumption rate of PV power generation, reduce electricity costs, and avoid additional load pressure on the grid during peak hours, residential energy storage systems typically need to make reasonable charging and discharging arrangements based on future changes in electricity consumption, electricity prices, and the state of the storage batteries themselves. Therefore, higher demands are placed on the real-time performance and accuracy of scheduling.

[0003] Existing home energy storage dispatch technologies mainly rely on two types of methods: The first type is based on preset rules for control, such as following a fixed logic of "charging at low electricity prices and discharging at high electricity prices," or controlling the operation of the energy storage system according to manually set time-sharing strategies; the second type of method usually first performs time-series forecasting of the total household load, and then combines optimization techniques such as linear programming to generate an energy storage dispatch plan. Although both types of methods are simple to implement and have low deployment costs, they both have obvious limitations: existing methods are difficult to describe the structured characteristics of household energy consumption behavior, cannot effectively distinguish or identify the start-up patterns, power distribution characteristics, or time fluctuation characteristics of different household energy consumption scenarios, and are also difficult to handle load changes caused by changes in user behavior, environmental fluctuations, or differences in equipment usage habits, thus resulting in load forecasting biases and making it difficult for energy storage dispatch to respond to actual energy consumption changes in a timely manner. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a multi-dimensional load forecasting method, system, and apparatus for optimizing household energy storage charging and discharging, which is used to achieve higher-precision load forecasting based on household energy consumption behavior characteristics and generate a dynamically adaptive energy storage charging and discharging control strategy.

[0005] The technical solution provided in this application is described below: The first aspect of this application provides a method for optimizing the charging and discharging of residential energy storage based on multidimensional load forecasting, including: Preprocessing of multi-source data related to household electricity consumption yields the input structure; Based on the input structure, multiple household energy consumption scenarios are identified, and corresponding energy consumption scenario parameters are constructed. The parameters carried by the energy consumption scenario parameters include the historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters, and external environmental influencing factors of the household energy consumption scenario. Based on the historical power distribution characteristics and behavior priority parameters of each energy consumption scenario parameter, the potential energy value of the energy consumption scenario parameter in each discrete time period within the prediction time window is calculated, and the corresponding potential energy curve is obtained. Based on the start-up time fluctuation model of each of the energy consumption scenario parameters, the potential energy curve is subjected to time diffusion processing to obtain the potential energy evolution curve. Within the prediction time window, the potential energy evolution curves of all the energy consumption scenario parameters are superimposed according to each discrete time period to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. The household total load forecast curve, electricity price information and household energy storage battery status data are input into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window. Based on the target charging and discharging power, a charging and discharging control command is generated to control the home energy storage battery to perform charging or discharging actions in corresponding discrete time periods.

[0006] Optionally, the step of inputting the household total load forecast curve, electricity price information, and household energy storage battery status data into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window includes: The net load pressure sequence for each discrete period within the forecast time window is calculated based on the total household load forecast curve and the reference load value. Calculate the electricity price incentive sequence for each discrete time period relative to the preset electricity price benchmark based on electricity price information; Based on the state data of home energy storage batteries, the feasible range of charge and discharge power for each discrete time period is determined; Calculate the state trend items for each discrete time period based on the state data of the home energy storage battery; Based on the net load pressure sequence, the electricity price incentive sequence, the state trend term, and external environmental influencing factors, construct the energy storage action tendency function for each discrete time period; The target charge / discharge power is determined within the feasible range of the charge / discharge power based on the energy storage action tendency function.

[0007] Optionally, determining the target charge / discharge power within the feasible range of the charge / discharge power based on the energy storage action tendency function includes: The original energy storage action tendency value for each discrete time period is calculated based on the energy storage action tendency function, which integrates the net load pressure sequence, electricity price incentive sequence, state trend term and external environmental influencing factors. The original energy storage action tendency value is subjected to amplitude constraint processing based on a nonlinear activation function with a limited range to obtain a normalized energy storage action tendency value limited to a preset range. The type of energy storage action is determined based on the comparison between the normalized energy storage action tendency value and the preset action threshold. Based on the normalized energy storage action tendency value, the target charge and discharge power is determined through amplitude mapping within the corresponding feasible range of charge and discharge power.

[0008] Optionally, the potential energy curve is subjected to time diffusion processing based on the start-up time fluctuation model of each of the energy consumption scenario parameters to obtain the potential energy evolution curve, including: Based on the historical start-up time distribution type of the start-up time fluctuation model for each of the energy consumption scenario parameters, the corresponding time diffusion kernel function is determined, and the time diffusion range parameter is calculated based on the time fluctuation amplitude of the start-up time fluctuation model. Based on the time diffusion kernel function and the time diffusion range parameter, a convolution operation is performed on each of the potential energy curves to obtain the potential energy value distribution after diffusion along the time axis within the prediction time window; Arrange the potential energy values ​​in chronological order to obtain the potential energy evolution curves of the energy consumption scenario parameters.

[0009] Optionally, based on the historical power distribution characteristics and behavioral priority parameters of each energy consumption scenario parameter, the potential energy value of the energy consumption scenario parameter for each discrete time period within the prediction time window is calculated to obtain the corresponding potential energy curve, including: Based on the historical power distribution characteristics, the corresponding expected power value is calculated for each discrete time period; The power expectation values ​​for each discrete time period are weighted based on the aforementioned behavior priority parameters to obtain a sequence of basic potential energy values. Based on the external environmental influencing factors, the basic potential energy value sequence is amplified or attenuated to obtain the potential energy values ​​of each energy consumption scenario parameter in each discrete time period, and arranged in chronological order to form the potential energy curve.

[0010] Optionally, the step of identifying multiple household energy consumption scenarios based on the input structure and constructing corresponding energy consumption scenario parameters includes: The historical power sequence is divided into power intervals and statistically processed to obtain the historical power distribution characteristics of each household energy consumption scenario; A start-up time fluctuation model is fitted based on the start-up time distribution of household energy consumption scenarios in the historical power sequence; The behavioral priority parameters are determined based on the energy consumption ratio of each household's energy use scenario within a preset period, as well as the user's set comfort and economic needs. External environmental impact factors are determined based on environmental data, electricity price information, and photovoltaic power generation data. The historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters, and external environmental impact factors are then written into the corresponding energy consumption scenario parameters.

[0011] Optionally, the preprocessing of multi-source data related to household electricity consumption to obtain the input structure includes: Time alignment processing is performed on historical household electricity consumption data, real-time electricity consumption data, electricity price information, environmental data, and photovoltaic power generation data to obtain a multi-source time series arranged with a unified time step. Interpolation imputation and outlier removal are performed on the missing data in the multi-source time series to obtain a data series with missing data filled and outliers removed. The data sequence is subjected to statistical feature extraction at a preset time granularity to obtain a set of feature vectors for household energy consumption scenario identification; Based on the feature vector set, multiple household energy consumption scenarios are identified using clustering or rule matching methods, and the identification results are associated with the corresponding time periods to form the input structure.

[0012] A second aspect of this application provides a home energy storage charging and discharging optimization system based on multidimensional load forecasting, the system comprising: The preprocessing unit is used to preprocess multi-source data related to household electricity consumption to obtain the input structure; The construction unit is used to identify multiple household energy consumption scenarios based on the input structure and construct corresponding energy consumption scenario parameters. The parameters carried by the energy consumption scenario parameters include the historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters and external environmental influence factors of the household energy consumption scenario. The calculation unit is used to calculate the potential energy value of the energy consumption scenario parameter in each discrete time period within the prediction time window based on the historical power distribution characteristics and behavior priority parameters of each energy consumption scenario parameter, and to obtain the corresponding potential energy curve. A diffusion unit is used to perform time diffusion processing on the potential energy curve based on the start-up time fluctuation model of each of the energy consumption scenario parameters to obtain the potential energy evolution curve. The superposition unit is used to superimpose the potential energy evolution curves of all the energy consumption scenario parameters in each discrete time period within the prediction time window to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. The input unit is used to input the household total load forecast curve, electricity price information and household energy storage battery status data into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window; The control unit is used to generate charge and discharge control commands based on the target charge and discharge power, and control the home energy storage battery to perform charging or discharging actions in corresponding discrete time periods.

[0013] Optionally, the input unit is specifically used for: The net load pressure sequence for each discrete period within the forecast time window is calculated based on the total household load forecast curve and the reference load value. Calculate the electricity price incentive sequence for each discrete time period relative to the preset electricity price benchmark based on electricity price information; Based on the state data of home energy storage batteries, the feasible range of charge and discharge power for each discrete time period is determined; Calculate the state trend items for each discrete time period based on the state data of the home energy storage battery; Based on the net load pressure sequence, the electricity price incentive sequence, the state trend term, and external environmental influencing factors, construct the energy storage action tendency function for each discrete time period; The target charge / discharge power is determined within the feasible range of the charge / discharge power based on the energy storage action tendency function.

[0014] Optionally, the input unit is further configured to: The original energy storage action tendency value for each discrete time period is calculated based on the energy storage action tendency function, which integrates the net load pressure sequence, electricity price incentive sequence, state trend term and external environmental influencing factors. The original energy storage action tendency value is subjected to amplitude constraint processing based on a nonlinear activation function with a limited range to obtain a normalized energy storage action tendency value limited to a preset range. The type of energy storage action is determined based on the comparison between the normalized energy storage action tendency value and the preset action threshold. Based on the normalized energy storage action tendency value, the target charge and discharge power is determined through amplitude mapping within the corresponding feasible range of charge and discharge power.

[0015] Optionally, the diffusion unit is specifically used for: Based on the historical start-up time distribution type of the start-up time fluctuation model for each of the energy consumption scenario parameters, the corresponding time diffusion kernel function is determined, and the time diffusion range parameter is calculated based on the time fluctuation amplitude of the start-up time fluctuation model. Based on the time diffusion kernel function and the time diffusion range parameter, a convolution operation is performed on each of the potential energy curves to obtain the potential energy value distribution after diffusion along the time axis within the prediction time window; Arrange the potential energy values ​​in chronological order to obtain the potential energy evolution curves of the energy consumption scenario parameters.

[0016] Optionally, the computing unit is specifically used for: Based on the historical power distribution characteristics, the corresponding expected power value is calculated for each discrete time period; The power expectation values ​​for each discrete time period are weighted based on the aforementioned behavior priority parameters to obtain a sequence of basic potential energy values. Based on the external environmental influencing factors, the basic potential energy value sequence is amplified or attenuated to obtain the potential energy values ​​of each energy consumption scenario parameter in each discrete time period, and arranged in chronological order to form the potential energy curve.

[0017] Optionally, the building unit is specifically used for: The historical power sequence is divided into power intervals and statistically processed to obtain the historical power distribution characteristics of each household energy consumption scenario; A start-up time fluctuation model is fitted based on the start-up time distribution of household energy consumption scenarios in the historical power sequence; The behavioral priority parameters are determined based on the energy consumption ratio of each household's energy use scenario within a preset period, as well as the user's set comfort and economic needs. External environmental impact factors are determined based on environmental data, electricity price information, and photovoltaic power generation data. The historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters, and external environmental impact factors are then written into the corresponding energy consumption scenario parameters.

[0018] Optionally, the preprocessing unit is specifically used for: Time alignment processing is performed on historical household electricity consumption data, real-time electricity consumption data, electricity price information, environmental data, and photovoltaic power generation data to obtain a multi-source time series arranged with a unified time step. Interpolation imputation and outlier removal are performed on the missing data in the multi-source time series to obtain a data series with missing data filled and outliers removed. The data sequence is subjected to statistical feature extraction at a preset time granularity to obtain a set of feature vectors for household energy consumption scenario identification; Based on the feature vector set, multiple household energy consumption scenarios are identified using clustering or rule matching methods, and the identification results are associated with the corresponding time periods to form the input structure.

[0019] A third aspect of this application provides a home energy storage charging and discharging optimization device based on multidimensional load forecasting, the device comprising: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to execute the first aspect and any one of the optional methods in the first aspect.

[0020] A fourth aspect of this application provides a computer-readable storage medium storing a program that, when executed on a computer, performs the methods of the first aspect and any one of the first aspects.

[0021] As can be seen from the above technical solutions, this application has the following advantages: By preprocessing household multi-source electricity consumption data and identifying household energy consumption scenarios, complex household energy consumption behaviors can be structurally expressed in the form of energy consumption scenario parameters. Potential energy curves are generated based on power distribution characteristics, priority parameters, and time fluctuation models according to energy consumption scenario parameters. These curves, after time diffusion, yield potential energy evolution curves, which can more accurately reflect future trends in household load. By inputting load forecasts, electricity price information, and battery status data into the scheduling model, a forward-looking target charging and discharging power can be generated. This allows the household energy storage system to adaptively adjust its charging and discharging strategies at different time periods, thereby improving the accuracy of load forecasting. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic flowchart of an embodiment of the household energy storage charging and discharging optimization method based on multidimensional load prediction in this application; Figure 2a This is a schematic flowchart of an embodiment of the home energy storage charging and discharging optimization method based on multidimensional load prediction in this application. Figure 2b This is a schematic flowchart of an embodiment of the second stage of the household energy storage charging and discharging optimization method based on multidimensional load prediction in this application; Figure 3 This is a schematic diagram of an embodiment of the home energy storage charging and discharging optimization system based on multidimensional load prediction in this application; Figure 4 This is a schematic diagram of an embodiment of the home energy storage charging and discharging optimization device based on multidimensional load prediction in this application. Detailed Implementation

[0024] It should be noted that the household energy storage charging and discharging optimization method based on multidimensional load forecasting provided in this application can be applied to terminals, systems, and servers. For example, terminals can be fixed terminals such as smartphones, computers, tablets, smart TVs, smartwatches, portable computer terminals, or desktop computers. For ease of explanation, this application uses terminals as the implementing entity for illustrative purposes.

[0025] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0026] Please see Figure 1 This application first provides an embodiment of a method for optimizing the charging and discharging of residential energy storage based on multidimensional load forecasting, which includes: S101. Preprocess the multi-source data related to household electricity consumption to obtain the input structure; Multi-source data refers to a collection of various data that reflects household electricity consumption behavior and external conditions, including historical household power consumption sequences, real-time electricity consumption data, electricity price information, environmental data, and photovoltaic power generation data. The input structure represents the data organization format formed by the terminal after completing time alignment, missing data completion, and feature extraction of the multi-source data, which is used for subsequent household energy consumption scenario identification.

[0027] Specifically, the terminal first performs time alignment on multi-source data from different acquisition paths, arranging all data based on a unified time step to ensure consistent temporal correlation in subsequent processing. For missing data due to acquisition delays or sensor errors, the terminal uses interpolation to complete the missing data and removes outliers that significantly deviate from the normal range to improve data quality. Subsequently, the terminal extracts statistical features describing household electricity consumption behavior from the aligned data according to the specified statistical time granularity, such as average power, peak power, power change rate, and photovoltaic power generation stability indicators.

[0028] After completing the above processing, the terminal assembles all the extracted features in chronological order to form an input structure, providing a unified and standardized data foundation for subsequent identification of household energy consumption scenarios.

[0029] S102. Identify multiple household energy consumption scenarios based on the input structure and construct corresponding energy consumption scenario parameters. The parameters carried by the energy consumption scenario parameters include the historical power distribution characteristics of the household energy consumption scenario, the start-up time fluctuation model, the behavior priority parameters, and the external environmental influencing factors. Household energy consumption scenarios represent relatively fixed energy consumption patterns in daily life, such as evening cooking, air conditioning use, washing machine operation, and home entertainment. These scenarios describe the impact of different lifestyle behaviors on electricity demand. Energy consumption scenario parameters represent a set of parameters describing the behavioral trends of household energy consumption scenarios within the forecast time window; these are the core input units for subsequent load forecasting.

[0030] Based on the feature vectors in the input structure, the terminal identifies time slices with similar power change characteristics through clustering algorithms or rule-based matching methods, and categorizes them into corresponding household energy consumption scenarios. Subsequently, the terminal constructs energy consumption scenario parameters for each scenario, and writes historical power distribution characteristics, start-up time fluctuation models, behavioral priority parameters, and external environmental impact factors derived from environmental data, electricity price information, and photovoltaic power generation data obtained in the previous steps.

[0031] After completing the identification of all scenarios and the construction of energy scenario parameters, the terminal obtains multiple energy scenario parameters that describe the characteristics of different energy use activities in the household, providing structured input for subsequent potential energy calculation steps.

[0032] S103. Based on the historical power distribution characteristics and behavior priority parameters of each energy consumption scenario, calculate the potential energy value of the energy consumption scenario parameters for each discrete time period within the prediction time window, and obtain the corresponding potential energy curve. The potential energy value of an energy consumption scenario parameter represents the tendency of the energy consumption scenario parameter to start up or the trend of power demand during a certain discrete period in the future, and is a key intermediate result in the load forecasting process. The potential energy curve represents the sequence of potential energy values ​​of the energy consumption scenario parameter over all discrete periods within the forecast time window, and is used to express the future behavioral trend of the energy consumption scenario parameter.

[0033] The terminal first generates multiple consecutive discrete time periods based on the time granularity of the prediction time window. For each time slice, it references the historical power distribution characteristics corresponding to the energy consumption scenario parameters to calculate the expected power value reflecting the typical power demand of that scenario during that time period. Subsequently, the terminal weights the expected power value according to behavioral priority parameters, giving greater weight to more important household energy consumption scenarios in the potential energy representation. The weighted result comprehensively reflects both the scenario's own characteristics and the importance of the behavior, making the future trend representation more consistent with actual household electricity consumption preferences.

[0034] The terminal arranges the potential energy values ​​of all discrete time periods in chronological order to form a potential energy curve describing the future trend of the parameters of the energy consumption scenario, providing basic data for the time diffusion calculation in the next step.

[0035] S104. Based on the start-up time fluctuation model of each energy consumption scenario, perform time diffusion processing on the potential energy curve to obtain the potential energy evolution curve. Time diffusion processing is used to extend the potential energy value of energy consumption scenario parameters in a certain discrete time period to adjacent time slices where start-up offset may occur, so that the behavioral trend can reflect the actual start-up advance or delay characteristics of the energy consumption scenario parameters.

[0036] After obtaining the potential energy curves of various energy consumption scenarios, the terminal performs a time diffusion operation on the potential energy curves based on the start-up time distribution characteristics described in its start-up time fluctuation model. Specifically, when a household energy consumption scenario has a relatively stable start-up range in its historical operation, its start-up time fluctuation range is small, and the terminal diffuses the potential energy value only within a small number of time slices; while when a scenario has greater start-up time uncertainty, the terminal expands the potential energy value over a wider time range, so that the potential energy value distribution can cover all possible start-up time points that the scenario may have in the actual environment.

[0037] The time diffusion process transforms the potential energy values, which were originally concentrated in a single time period, into a smooth and continuous trend expression, thereby more accurately representing the overall behavioral trend of energy consumption scenario parameters within the prediction time window. After completing the diffusion, the terminal arranges the results in chronological order, obtaining potential energy evolution curves corresponding to multiple energy consumption scenario parameters, providing structured time-series data for the next step of calculating the household total load prediction curve.

[0038] S105. Within the prediction time window, the potential energy evolution curves of all energy consumption scenario parameters are superimposed according to each discrete time period to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. The total household load forecast curve represents the combined load trend of multiple household energy consumption scenarios within a future forecast time window, serving as an important basis for energy storage systems to make charging and discharging decisions.

[0039] After obtaining the potential energy evolution curves for various energy consumption scenarios, the terminal first ensures that all potential energy evolution curves use the same discrete time period as the basis for time alignment, allowing the trend values ​​of different household energy consumption scenarios to be superimposed within the same time slice. Subsequently, within the prediction time window, the terminal numerically superimposes the potential energy evolution values ​​of all energy consumption scenario parameters for each discrete time slice to obtain a comprehensive potential energy value reflecting the overall energy demand of the household. The superposition result can reflect the combined effect of multiple scenarios in a certain time slice. For example, when there are simultaneous trends in air conditioning operation and cooking load, the comprehensive potential energy during that time slice will increase significantly.

[0040] The terminal arranges the superimposed results of all time periods in chronological order to form a total household load forecast curve covering the entire forecast time window, providing input data for subsequent energy storage charging and discharging scheduling steps.

[0041] S106. Input the household total load forecast curve, electricity price information and household energy storage battery status data into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window. The game-theoretic scheduling model for energy usage scenario parameters represents a decision-making model used to comprehensively evaluate the charging or discharging strategies that a home energy storage system should execute in future discrete time periods under the combined influence of multiple factors. This model incorporates factors such as household load trends, electricity price changes, and battery operating status into the same decision space, achieving overall optimization of the energy storage system through multi-factor trade-offs.

[0042] After obtaining the total household load forecast curve, the terminal inputs it along with future electricity price information and energy storage battery status data. The energy consumption scenario parameter game scheduling model first determines the electricity consumption pressure for each time period based on the load forecast curve, uses electricity price information to express cost-driven trends, and then combines the battery's remaining capacity, allowable power, and operating limitations to construct constraints for the energy storage strategy. After considering the above factors, the model determines the appropriate action direction and power level for each discrete time period through its internal strategy generation mechanism, prioritizing discharging during high-load, high-electricity-price periods and prioritizing charging during low-load, low-electricity-price periods, while ensuring that the entire decision-making process does not violate the battery's safe operating boundaries.

[0043] After completing the calculation, the terminal obtains the target charging and discharging power corresponding to all discrete time periods within the prediction time window, providing a power basis for the subsequent generation of actual control commands.

[0044] S107. Generate charging and discharging control commands based on the target charging and discharging power, and control the household energy storage battery to perform charging or discharging actions in the corresponding discrete time periods.

[0045] Charge and discharge control commands represent the underlying control commands used to drive the energy storage system to perform charging, discharging, or holding actions during a specific period of time. They are execution signals that translate the optimized calculation results into physical energy storage devices.

[0046] After obtaining the target charge / discharge power for each discrete time period, the terminal converts the target power into a control command format that the device can recognize, according to the communication protocol and control interface format of the home energy storage battery. For example, when the target power for a certain time period is the discharge power, the terminal generates a corresponding discharge control command; when the target power is the charging power, the terminal generates a charging command; if the target power is close to zero, the terminal generates a hold command to maintain the current state of the battery.

[0047] Within the predicted time window, the terminal sends the aforementioned control commands to the energy storage system in chronological order, enabling the energy storage battery to accurately perform the corresponding charging and discharging actions in each discrete time period, thereby achieving optimized scheduling of household energy.

[0048] This embodiment sequentially executes steps such as multi-source data preprocessing, household energy consumption scenario identification, potential energy trend construction, time diffusion processing, and load superposition prediction through the terminal, forming a complete load prediction process for household energy storage systems. In this embodiment, the terminal generates potential energy evolution curves of future trends based on historical power characteristics, priority parameters, and start-up time fluctuation models for different household energy consumption scenarios. It then superimposes the trends of all scenarios within the prediction time window to obtain a total household load prediction curve reflecting overall household energy consumption changes, providing an accurate basis for subsequent energy storage system scheduling. By dividing household energy consumption into multiple scenarios and constructing trend potential energy expressions, and then combining this with time diffusion to form a load trend closer to actual usage behavior, the terminal ensures that the total household load prediction results fully reflect usage habits and time uncertainties, thereby significantly improving the accuracy of household load prediction and laying a reliable foundation for the subsequent optimized charging and discharging scheduling of the energy storage system.

[0049] Please see Figures 2a to 2b This application provides another embodiment of a residential energy storage charging and discharging optimization method based on multidimensional load forecasting, which includes: S201. Perform time alignment processing on household historical electricity consumption data, real-time electricity consumption data, electricity price information, environmental data, and photovoltaic power generation data to obtain a multi-source time series arranged with a unified time step. Historical household electricity consumption data represents the power consumption sequence of a household over past operating cycles, reflecting the typical energy consumption structure of a household; real-time electricity consumption data represents the power data of a household within the current sampling period; electricity price information represents the electricity price level for the corresponding time period; environmental data represents external factors such as temperature and humidity that affect household electricity consumption behavior; and photovoltaic power generation data represents the power generation of photovoltaic modules during the corresponding time period.

[0050] The terminal first acquires the aforementioned data types and determines the arrangement of the target time series based on a uniformly set time step. The terminal resamples data from different sources according to their original timestamps, arranging all data at the same time granularity to ensure consistency across the timeline. After this processing, the terminal obtains a multi-source time series arranged with a uniform time step.

[0051] S202. Perform interpolation imputation and outlier removal on the missing data in the multi-source time series to obtain a data series with missing data filled and outliers removed. Missing data indicates gaps in time-slice data due to data acquisition interruptions or communication problems. Outliers indicate sudden erroneous data that does not conform to normal power consumption behavior, such as load spikes caused by sensor jitter.

[0052] The terminal scans the multi-source time series, identifies the locations of missing data, and fills the gaps using interpolation based on neighboring time slices to maintain the continuity of the sequence. Subsequently, the terminal identifies outliers in the sequence according to preset anomaly detection conditions and replaces or removes them based on their location and the preceding and following normal data. After processing, the terminal obtains a data sequence with missing data filled and anomalies removed.

[0053] S203. Statistical feature extraction is performed on the data sequence according to a preset time granularity to obtain a set of feature vectors for household energy consumption scenario identification. The preset time granularity represents the time interval used to divide the data sequence into multiple statistical time periods, such as 5 minutes, 10 minutes, or 15 minutes. This time granularity is used to generate statistical characteristics of household energy consumption behavior.

[0054] The terminal divides the completed data sequence into continuous statistical windows based on the time granularity, and extracts feature combinations such as mean power, power variance, maximum power, minimum power, and photovoltaic output within each window to form a feature vector characterizing the energy consumption characteristics of that time period. After integrating the feature vectors from all windows, the terminal obtains a feature vector set for identifying household energy consumption scenarios.

[0055] S204. Based on the feature vector set, clustering or rule matching is used to identify multiple household energy consumption scenarios, and the identification results are associated with the corresponding time period to form an input structure.

[0056] The terminal performs clustering analysis on the data based on the feature differences in the feature vector set, automatically aggregating feature vectors with similar power behaviors into the same category, thereby identifying different household energy consumption scenarios. For certain scenarios with clear regular characteristics (such as the periodic operation of refrigerators), the terminal can identify them based on rule matching. The terminal associates each identified household energy consumption scenario with its corresponding statistical window on the time axis and organizes the identification results into an input structure in chronological order, providing a basis for subsequently constructing energy consumption scenario parameters.

[0057] S205. Divide the historical power sequence into power intervals and perform statistical processing to obtain the historical power distribution characteristics of each household energy consumption scenario; Historical power series represent power records of corresponding household energy consumption scenarios in the past, used to statistically analyze the typical power levels and distribution patterns of the scenario.

[0058] Based on the identified household energy consumption scenarios in the input structure, the terminal extracts the historical power sequence for each scenario and divides the sequence according to a set power range. For example, the terminal can divide the sequence into several segments based on power magnitude and statistically analyze the frequency, average power, and main fluctuation range of each segment in the historical operation of that scenario. The terminal obtains the historical power distribution characteristics of the scenario through statistical processing, which describes the power pattern of the scenario in its historical operation and is an important basis for subsequent calculation of potential energy values.

[0059] S206. Fit a start-up time fluctuation model based on the start-up time distribution of household energy consumption scenarios in historical power sequences; Start-up time distribution represents the set of specific times when a household energy consumption scenario begins operation within a historical cycle, reflecting the temporal pattern of triggering behaviors in that scenario. Start-up time fluctuation model represents the range of deviations between the actual start-up time of a scenario and its typical start-up time, expressing the temporal variation characteristics caused by differences in family member behavior.

[0060] After obtaining the historical power sequence of each household energy consumption scenario, the terminal first identifies the start time of each past operation of the scenario and organizes these start times in chronological order by date and time to form start time distribution data. Based on this start time distribution, the terminal analyzes the trigger concentration of the scenario at different times of the day. For example, air conditioning energy consumption may be more concentrated in the evening, while washing machine operation may occur more frequently at night.

[0061] The terminal determines the typical startup time location of a scenario based on the clustering degree of startup time distribution and calculates the offset of each startup time in historical data relative to this typical startup time. The terminal then fits a startup time fluctuation model for the scenario based on the range of variation of this offset, which describes the potential future time offset characteristics of the scenario within the household. Through this processing, the terminal obtains a startup time fluctuation model that reflects the variability of scenario startup times, serving as the basis for subsequent time diffusion.

[0062] For example, after analyzing the launch time distribution of the "evening cooking and heating scenario" over the past month, it was found that the launch times for this scenario are mainly concentrated between 18:00 and 18:30 daily, with 18:10 being the most frequent time. The terminal uses 18:10 as the typical launch time for this scenario and calculates the offset of all historical launch times relative to 18:10. For example, if the launch time on a certain day is 18:05, then the offset is... If the startup time on another day is 18:25, then the offset is +15 minutes. After statistically analyzing all offsets, the terminal forms a startup time fluctuation model for this scenario based on the actual range of change of the offset values ​​in the positive and negative directions. This model is used to represent the startup behavior that may occur in this scenario within a certain time interval around 18:10 in the future.

[0063] S207. Determine behavioral priority parameters based on the energy consumption ratio of each household's energy consumption scenario within a preset cycle, as well as the user's set comfort and economic needs. Energy consumption percentage indicates the proportion of a household's energy consumption within a preset analysis period (e.g., one day, one week, or one month), reflecting the importance of that scenario. Comfort needs indicate the degree to which users rely on the continuous usage experience of that scenario; for example, scenarios such as air conditioning and water heaters are usually highly correlated with household comfort. Economic needs indicate the extent to which users want to reduce energy costs and are closely related to sensitivity to electricity price changes; for example, washing machines or electric vehicle charging are usually highly dispatchable.

[0064] After obtaining the historical power sequences of each household's energy consumption scenarios, the terminal first calculates the total energy consumption of each scenario within a preset period and its proportion relative to the overall household energy consumption, reflecting the actual importance of that scenario in the household's energy structure. Then, based on the user's preference parameters set in the system, the terminal determines the degree of impact of each scenario on household comfort. For example, air conditioning energy consumption directly affects the living experience, so its comfort requirements are usually high; while washing machine operation can usually be delayed, so its comfort requirements are lower. The terminal also analyzes the usage distribution of scenarios during different electricity price periods to assess the economic demand for that scenario. For example, if a scenario mostly operates during low electricity price periods, it indicates that its economic demand is high and can be optimized and adjusted in the scheduling process.

[0065] The terminal constructs a weighted relationship model based on energy consumption ratio, comfort requirements, and economic requirements. It combines these three factors according to set weights to generate behavioral priority parameters for each household energy consumption scenario, reflecting its overall importance. After completing the calculation, the terminal obtains the behavioral priority parameters and writes them into the corresponding household energy consumption scenario, providing a priority basis for subsequent potential energy value calculations.

[0066] S208. Based on environmental data, electricity price information and photovoltaic power generation data, determine the external environmental impact factors, and write the historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters and external environmental impact factors into the corresponding energy consumption scenario parameters. External environmental influencing factors represent a set of external conditions that can affect the probability of triggering household energy consumption scenarios and power demand. These factors express the moderating effect of changes in the environment, prices, and energy supply on the operational trends of these scenarios. This factor is typically composed of elements such as temperature, humidity, weather changes, electricity price levels, and photovoltaic output.

[0067] After acquiring environmental data, electricity price information, and photovoltaic power generation data, the terminal selects external influencing factors relevant to the characteristics of the household energy consumption scenario. For example, air conditioning energy consumption is highly sensitive to changes in temperature and humidity; therefore, when calculating the external environmental influencing factors for this scenario, the terminal focuses on temperature change curves and real-time weather conditions. Cooking and heating scenarios are generally less affected by weather but are more closely related to the user's lifestyle; therefore, the terminal primarily focuses on changes in electricity prices for this scenario. For scenarios relying on photovoltaic self-generation and self-consumption, such as household water heater heating, the terminal will focus on the changing trends of photovoltaic output.

[0068] The terminal performs a weighted combination of environmental data, electricity price information, and photovoltaic power generation data based on the sensitivity of the scenario to different external factors, resulting in an external environmental impact factor that reflects the adjustment trend of external conditions in the scenario. This factor is used to correct the trend of energy consumption scenario parameters when calculating potential energy values ​​in subsequent calculations, so that the energy consumption scenario parameters can more accurately express the actual operating trend.

[0069] After obtaining the external environmental influencing factors, the terminal writes these factors, along with the historical power distribution characteristics, startup time fluctuation model, and behavior priority parameters obtained in the previous steps, into the corresponding energy consumption scenario parameters, so that each energy consumption scenario parameter has a complete set of parameters that describe its future behavior trend.

[0070] S209. Calculate the expected power value for each discrete time period based on the historical power distribution characteristics; The expected power value represents the typical power demand that an energy consumption scenario parameter may exhibit at a certain discrete time period in the future, and is used as the basic input for calculating the subsequent potential energy value of that energy consumption scenario parameter. The expected power value is derived from a model based on typical power ranges, power averages, and frequency of occurrence extracted from historical power distribution characteristics.

[0071] After obtaining the historical power distribution characteristics of a household energy consumption scenario, the terminal first divides the future time axis into multiple consecutive time slices based on the discrete time period setting of the prediction time window, and then matches the historical power statistics of the scenario within the same time period for each time slice. For example, in the "evening cooking and heating scenario", if historical data shows that there is a high power concentration in the interval of 18:00–18:30, the terminal uses the typical power average of that time period as the basis for the power expectation in the corresponding time slice within the prediction window.

[0072] The terminal combines the frequency of occurrence of historical power intervals to perform weighted calculations on the power level of each time slice, so that each historical pattern is reflected in the expected value according to its probability of occurrence. After this calculation, the terminal obtains the expected power values ​​of each energy consumption scenario parameter in future discrete time periods, which are used as the data source for the next step of calculating the basic potential energy value sequence.

[0073] S210. The power expectation values ​​for each discrete time period are weighted based on the behavior priority parameter to obtain the basic potential energy value sequence. The behavior priority parameter represents the overall importance of each household energy consumption scenario in the overall household energy consumption structure, and has been determined in previous steps based on energy consumption proportion, comfort needs, and economic needs. The basic potential energy value sequence represents the basic values ​​of the behavior trend of energy consumption scenario parameters over future discrete time periods. It is used to reflect the impact of scenario importance on its power demand trend and is an intermediate result for calculating the potential energy value of energy consumption scenario parameters.

[0074] After obtaining the expected power values ​​for each energy consumption scenario parameter in future discrete time periods, the terminal combines the corresponding scenario's behavior priority parameter with its expected power value. Specifically, the terminal weights the expected power values ​​according to the magnitude of the behavior priority parameter, making the expression of scenarios with higher importance more significant in the basic potential energy value sequence. For example, in a household, the energy consumption of air conditioners is usually more important than that of washing machines; the behavior priority parameter for the air conditioner scenario is larger, therefore, under the same expected power value for the same time period, the basic potential energy value of the air conditioner scenario will be relatively higher.

[0075] After processing each discrete time slice, the terminal arranges the weighted values ​​according to the time order of the prediction time window to form a basic potential energy value sequence.

[0076] S211. Based on the external environmental influence factors, perform amplification or attenuation calculations on the basic potential energy value sequence to obtain the potential energy values ​​of each energy consumption scenario parameter in each discrete time period, and arrange them in time order to form a potential energy curve.

[0077] The potential energy value of the energy consumption scenario parameter is used to quantify the startup tendency or power trend of the energy consumption scenario parameter in a certain discrete period in the future, reflecting the future behavior intensity of the scenario after comprehensively considering historical behavior characteristics and external conditions.

[0078] After obtaining the basic potential energy value sequence, the terminal adjusts the sequence according to external environmental influencing factors. The direction and intensity of the effect of these external environmental influencing factors vary depending on the scenario. For example, air conditioning scenarios are highly sensitive to temperature increases. When the external temperature rises, the terminal uses the temperature-related influencing factors to increase the potential energy value of that scenario in the corresponding time slot; conversely, when the temperature is low, it correspondingly reduces the potential energy level for that period. For scenarios relying on photovoltaic power output, such as household water heater heating, when the photovoltaic power generation capacity is strong, the terminal uses corresponding external factors to attenuate the basic potential energy value to reflect the substitution effect of external renewable energy on the scenario's demand.

[0079] The terminal sequentially performs the amplification or attenuation processing on each discrete time period within the prediction time window, and arranges the processed potential energy values ​​in chronological order to form the potential energy curve of the energy consumption scenario parameters, providing continuous time-series input for subsequent time diffusion processing.

[0080] S212. Determine the corresponding time diffusion kernel function based on the historical start-up time distribution type of the start-up time fluctuation model based on the parameters of each energy consumption scenario, and calculate the time diffusion range parameter based on the time fluctuation amplitude of the start-up time fluctuation model. The time-diffusion kernel function describes the potential start-up time offset of energy consumption scenario parameters in the future. It is a structure that mathematically expresses the historical start-up time distribution characteristics of energy consumption scenario parameters. The time fluctuation amplitude represents the offset range of the start-up time of the energy consumption scenario parameters in historical operation, and is used to limit the range of influence of the potential energy value on the time axis during the diffusion process. The time-diffusion range parameter represents the length of the diffusion period generated based on the time fluctuation amplitude, and is used to control the coverage and attenuation degree of the diffusion process.

[0081] After constructing the start-up time fluctuation model for energy consumption scenario parameters, the terminal first analyzes the historical start-up time distribution types included in the model. For example, when the historical start-up times of energy consumption scenario parameters exhibit a single-peak concentrated distribution, it indicates that the start-up times of the scenario have mostly been concentrated in a specific time period in the past. Based on this, the terminal selects a diffusion kernel function with central peak characteristics to maintain the concentration of diffusion. When the start-up times exhibit a bimodal or multimodal structure, it indicates that the scenario has a high trigger probability in multiple time periods. The terminal selects a kernel function that can express a multimodal structure so that the diffusion results can reflect the potential triggering tendency in multiple time periods. When the start-up times exhibit a relatively smooth wide-range distribution, the terminal selects a kernel function with slow decay characteristics to reflect the relatively free offset characteristics of the start-up times.

[0082] Subsequently, the terminal calculates the time diffusion range parameter based on the time offset range recorded in the startup time fluctuation model. For example, when the historical startup behavior of a certain scenario fluctuates significantly within a range of ±10 minutes, the terminal sets the diffusion range according to this offset amplitude, ensuring that the potential energy value only expands within this range. When setting the range parameter, the terminal introduces an attenuation factor at the offset boundary, causing the diffusion intensity to gradually decrease with the time of deviation from the center, thereby ensuring that the diffusion covers the actual possible offset range without exceeding the actual triggering range of the scenario.

[0083] After determining the time-diffusion kernel function and diffusion range parameters, the terminal obtains a diffusion configuration with parameters suitable for the current energy consumption scenario, providing the necessary parameter basis for subsequent time-diffusion processing of the potential energy curve.

[0084] S213. Perform convolution operation on each potential energy curve based on the time diffusion kernel function and the time diffusion range parameter to obtain the potential energy value distribution after diffusion along the time axis within the prediction time window. Potential energy distribution is used to express the trend of the future start-up tendency of energy consumption scenario parameters under the influence of time diffusion. It is the unfolded form of the potential energy curve in the time dimension.

[0085] After determining the time-spreading kernel function and time-spreading range parameters in the previous steps, the terminal uses the kernel function as the basic structure of the convolution operator and performs hourly convolution processing on the potential energy curve on the time axis of the prediction time window. Specifically, the terminal limits the effective boundary of the convolution operation according to the spread range parameters, so that the potential energy value only spreads in the region where the start time may be offset, avoiding the convolution result from affecting time segments beyond the physical meaning.

[0086] During convolution, the terminal will use the shape of the selected time-spreading kernel function, such as a central peak, multi-peak, or slowly decaying type, without specifying a particular type here. This will spread the original potential energy value to adjacent time slices with kernel function weights, causing the potential energy value, originally concentrated in a certain time period, to gradually spread within an allowable offset range. The result of this operation can realistically reflect the behavioral characteristics of energy consumption scenario parameters, including potential early start-up, delayed start-up, and cross-time-slice diffusion.

[0087] After completing the convolution operation for all discrete time periods, the terminal obtains the potential energy value distribution covering the entire prediction time window, providing structured data for the next step of forming a continuous potential energy evolution curve.

[0088] S214. Arrange the potential energy values ​​in chronological order to obtain the potential energy evolution curves of each energy consumption scenario parameter.

[0089] The potential energy evolution curve represents the future behavior trend of energy consumption scenario parameters after time diffusion processing. It is the final time-series expression of the future power demand or start-up tendency of the energy consumption scenario parameters, and is used to support the prediction of total household load.

[0090] After obtaining the potential energy value distribution, the terminal arranges it according to the discrete time periods of the prediction time window, so that the potential energy values ​​of different time slices form a continuous sequence. Through this arrangement process, each potential energy value is organized into a structurally complete continuous trend on the time axis, and the time offset characteristics of the energy scene parameters and future action tendencies can be visualized.

[0091] After the terminal completes the sorting, it obtains the potential energy evolution curve of the application energy scenario parameters. This curve serves as the input for the subsequent household total load prediction step, enabling the trend characteristics of different household energy consumption scenarios to be superimposed and analyzed on the same time axis.

[0092] S215. Within the prediction time window, the potential energy evolution curves of all energy consumption scenario parameters are superimposed according to each discrete time period to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. Step S215 in this embodiment is similar to step S105 in the previous embodiment, and will not be described in detail here.

[0093] S216. Calculate the net load pressure sequence for each discrete period within the forecast time window based on the household total load forecast curve and reference load value. Net load pressure represents the deviation of the predicted electricity demand for a specific future period from the reference load value. It describes the degree of electricity shortage that households may face during that period and is an important basic input for subsequent energy storage decisions.

[0094] The reference load value represents a threshold used to measure household load pressure. It can be determined based on factors such as the household's historical electricity consumption characteristics, the inverter's rated output capacity, or the grid-side operating requirements, and does not constitute a limitation on this method.

[0095] After obtaining the total household load forecast curve, the terminal calculates the difference between each forecast load value and the reference load value based on the discrete time slices of the forecast time window. This reflects the degree of deviation of future electricity demand relative to the baseline level. For example, when the forecast load for a certain period is significantly higher than the reference load value, the net load pressure is greater, reflecting the potential for increased energy storage discharge demand during that period. Conversely, when the forecast load is lower, the net load pressure is smaller, making the corresponding period more suitable for charging activities.

[0096] After performing the above difference calculation on all discrete time slices within the prediction time window, the terminal obtains a net load pressure sequence arranged in chronological order, which serves as the data input for subsequently constructing the energy storage action tendency function.

[0097] S217. Calculate the electricity price incentive sequence for each discrete time period relative to the preset electricity price benchmark based on electricity price information; Electricity price incentives represent the deviation of electricity prices from a reference price level for a future discrete time period, used to express the driving force of electricity price changes on the charging or discharging behavior of energy storage systems. Preset electricity price benchmarks represent a reference value used to measure electricity price levels, which can be automatically generated by the terminal from various data sources, such as historical average electricity prices, segmented averages of different price tiers, or user-defined electricity price sensitivity preferences in the system.

[0098] The terminal first analyzes the electricity price level for each discrete time period within the prediction time window from the electricity price information. Then, based on historical electricity price records and user configuration parameters, the terminal generates a preset electricity price benchmark for calculating the offset. For example, when users prefer low-cost operation, the terminal selects the average of historically low electricity price periods as the benchmark; when users have higher requirements for comfort, the terminal uses a more lenient average electricity price, making the system more inclined to reduce fluctuations in electricity consumption behavior.

[0099] After generating the electricity price benchmark, the terminal calculates the difference between the electricity price for each time period and the benchmark value to obtain an electricity price incentive value reflecting the tendency to discharge or charge. The terminal arranges the calculated values ​​in chronological order to form an electricity price incentive sequence, providing input for the subsequent construction of the energy storage action tendency function.

[0100] S218. Determine the feasible range of charging and discharging power for each discrete time period based on the state data of household energy storage batteries; State of matter (SOC) data for home energy storage batteries represents a set of parameters that reflect the battery's current operational capabilities and safe operating boundaries. These parameters include the battery's remaining charge (SOC), allowable charge / discharge power, battery temperature, state of health (SOH), and current operating mode. The feasible charge / discharge power range represents the minimum and maximum range of charge / discharge power that the energy storage battery is allowed to perform in various discrete time periods in the future. This ensures that subsequent scheduling actions comply with battery safety operating conditions and equipment performance limitations.

[0101] After acquiring the energy storage battery status data, the terminal first calculates the maximum allowable charging power and maximum discharge power within the subsequent time slice based on the battery's remaining charge. For example, when the battery is near full charge, the terminal reduces the upper limit of the allowable charging power according to the energy storage protection mechanism; when the battery is at a low charge level, the upper limit of the allowable discharge power is reduced. Simultaneously, the terminal further limits the allowable charging and discharging power range based on the energy storage device's rated power parameters, current temperature conditions, and battery health status to ensure the battery remains within a safe operating range throughout the entire predicted time window.

[0102] The terminal performs the above judgments sequentially on each discrete time slice within the prediction time window, and finally obtains the feasible range of charging and discharging power arranged in time series, which provides a basis for the constraints on the subsequent energy storage operation tendency.

[0103] S219. Calculate the state trend item for each discrete time period based on the state data of household energy storage batteries; The state trend term represents the potential operating direction of a home energy storage battery at a certain discrete time in the future, including bias towards charging, bias towards discharging, or maintaining the current state. It is used to express the driving effect brought about by the battery's own state in the calculation of energy storage action tendency.

[0104] The terminal analyzes current battery status data, including remaining charge (SOC), state of health (SOH), battery temperature, and current operating mode, to predict future battery behavior trends. When the remaining charge is near the safety lower limit, the terminal sets the trend towards charging based on battery protection logic to ensure the battery does not enter a deep discharge region. When the remaining charge is high and the battery health is good, the trend leans towards discharging to improve overall energy utilization. When the remaining charge is in the middle range, the terminal sets the status trend to a relatively neutral "hold" state by combining historical operating curves, daily load patterns, and battery temperature change trends.

[0105] The terminal calculates the possible states of the battery for each time period according to the discrete time periods of the prediction time window, and finally obtains a sequence of state trend terms arranged in chronological order, which provides the basic input for the energy storage action tendency function.

[0106] S220. Construct energy storage action tendency functions for each discrete time period based on net load pressure sequence, electricity price incentive sequence, state trend term and external environmental influencing factors; The energy storage action tendency function represents the comprehensive driving force of the energy storage system on charging or discharging behavior in a certain discrete period of the future. It is a function formed by the combined effects of load pressure, electricity price changes, battery state trends and external environmental conditions, and is used to express the direction of energy storage action under multi-factor conditions.

[0107] After obtaining the net load pressure sequence, electricity price incentive sequence, state trend term, and external environmental influencing factors, the terminal constructs an energy storage action tendency function based on the correlation between each factor and energy storage behavior. For example, high load pressure usually increases the discharge tendency; low electricity price incentive usually increases the charging tendency; the state trend term is used to express the battery's own operating preferences; and the external environmental influencing factors are used to reflect the impact of external conditions such as changes in photovoltaic output and changes in electricity price structure on energy storage behavior.

[0108] Within each discrete time period of the prediction time window, the terminal combines the above four types of factors in a time-aligned manner to generate a function structure for subsequent calculation of the energy storage action tendency value. After this processing, the terminal obtains the complete energy storage action tendency function, providing function input for the next step of action tendency value calculation.

[0109] S221. Based on the energy storage action tendency function, the net load pressure sequence, electricity price incentive sequence, state trend term and external environmental influencing factors are fused to calculate the original energy storage action tendency value for each discrete time period. The raw energy storage action tendency value represents the intensity at which the energy storage system is more likely to perform charging or discharging during a certain discrete time period under the combined effect of multiple factors, and is used as the basic input for subsequent normalization processing.

[0110] Specifically, the formula for calculating the original energy storage tendency value is as follows: Z(t) = k1 * L(t) + k2 * C(t) + k3 * S(t) + k4 * E(t).

[0111] Wherein, L(t) represents the net load pressure in time period t, which is the deviation of the predicted load from the reference load. C(t) represents the electricity price incentive in time period t, which is the deviation of the electricity price from the preset electricity price benchmark. S(t) represents the state trend term in time period t, which is the trend value of the battery's future state biased towards charging, discharging, or maintaining. E(t) represents the external environmental influence factor in time period t, used to reflect the regulatory effect of external conditions on energy storage behavior. k1, k2, k3, and k4 are the weight coefficients of each factor, used to balance the influence intensity of different factors on the energy storage action tendency. The weight values ​​are preset by the system or obtained through training with historical operating data. Z(t) represents the original energy storage action tendency value in time period t.

[0112] Within each discrete time period of the prediction window, the terminal substitutes net load pressure, electricity price incentives, state trend terms, and external environmental influencing factors into the aforementioned calculation formula, and weights and fuses the effects of different factors using weighting coefficients to obtain the original energy storage action tendency value for the corresponding time period. The terminal performs this calculation sequentially for all discrete time periods and arranges the results in chronological order to form a sequence of original energy storage action tendency values, providing input for the activation function processing in subsequent steps.

[0113] S222. The original energy storage action tendency value is subjected to amplitude constraint processing based on the nonlinear activation function with a limited interval to obtain a normalized energy storage action tendency value limited to a preset range. Nonlinear activation functions are used to compress the original energy storage tendency value to a set range, avoiding excessively large or small values ​​due to the superposition of multiple factors, thereby ensuring the comparability and stability of the tendency value across different time slices. The defined interval represents the allowable output range of the tendency value, for example... The range is 1 to 1 or 0 to 1; this embodiment does not limit the specific range.

[0114] When selecting an activation function, the terminal determines the appropriate function type based on the numerical distribution characteristics of the original energy storage action tendency value.

[0115] Specifically, when the original tendency value shows a rapid jump trend in some time slices, the terminal selects an incremental activation function with smooth transition characteristics to keep the compression result continuous; when the original tendency value fluctuates greatly between different time slices, the terminal selects a saturated activation function with stronger amplitude limiting ability to ensure that the result does not exceed the set upper limit; when the original tendency value is concentrated in a narrow range, the terminal selects an activation function with a larger slope to enhance the distinguishability between time slices.

[0116] After completing the function type selection, the terminal processes all the original energy storage action tendency values ​​within the prediction time window one by one, so that each processed output value is limited to the set range, and arranges the processing results in chronological order to obtain a normalized energy storage action tendency value sequence, which is used to support the action type judgment in the next step.

[0117] S223. Determine the energy storage action type for each discrete time period based on the comparison results between the normalized energy storage action tendency value and the preset action threshold. The energy storage action type indicates the operational direction that the energy storage system should perform during a discrete time period, including three categories: charging, discharging, or remaining unchanged, which are used to guide the subsequent power mapping process. The preset action threshold is a numerical threshold used to distinguish different energy storage action types. It is usually set according to the energy storage system's operating strategy and safety range, and does not constitute a limitation of this method.

[0118] After obtaining the normalized energy storage action tendency value sequence, the terminal compares the tendency value for each time period with preset action thresholds. When the tendency value is higher than the discharge threshold, the terminal determines the action for that time period to be discharge; when the tendency value is lower than the charging threshold, the terminal determines the action for that time period to be charging; when the tendency value is between the charging and discharging thresholds, the terminal determines the energy storage action type for that time period to be hold. Through this judgment method, the terminal provides a clear basis for determining the action direction for subsequent target power.

[0119] The terminal arranges the judgment results of each discrete time period in chronological order to obtain the corresponding energy storage action type sequence.

[0120] S224. Based on the normalized energy storage action tendency value, the target charging and discharging power is determined by amplitude mapping within the corresponding feasible range of charging and discharging power.

[0121] Amplitude mapping refers to the process of mapping the normalized energy storage action tendency value to a specific power value within the feasible charging or discharging power range. This is used to generate an actually executable power command after the action type is determined. The feasible charging and discharging power range represents the power range derived from the energy storage battery state, equipment rated capacity, and safe operating limits. This range is used to ensure that the action result does not exceed the battery's tolerable operating boundaries.

[0122] After determining the energy storage action type for a given time period, the terminal first establishes an amplitude correlation between the normalized tendency value for that time period and the feasible power range. If the action type is discharging, the terminal performs a linear or non-linear interval mapping within the feasible discharge power range according to the magnitude of the tendency value, so that a higher tendency value corresponds to a higher discharge power. If the action type is charging, the mapping relationship is formed within the feasible charging power range, so that a higher tendency value corresponds to a higher charging rate. If the action type is holding, the terminal limits the target power to a range close to zero according to the energy storage system strategy to maintain a stable battery energy level.

[0123] After performing the above mapping process on all discrete time periods within the prediction time window, the terminal obtains the target charging and discharging power sequence arranged in chronological order, providing specific input basis for the final generation of energy storage control commands.

[0124] S225. Generate charging and discharging control commands based on the target charging and discharging power, and control the household energy storage battery to perform charging or discharging actions in the corresponding discrete time periods.

[0125] Step S225 in this embodiment is similar to step S107 in the previous embodiment, and will not be described in detail here.

[0126] This embodiment achieves structured organization of household electricity consumption data through time alignment, missing data completion, and feature extraction of multi-source data. Furthermore, through scene recognition, power distribution analysis, start-up time fitting, and priority calculation, it enables each type of household energy consumption scenario to possess predictable behavioral characteristics. Based on this, it further generates potential energy curves, potential energy evolution curves, and household total load prediction curves. Combining load pressure, electricity price incentives, battery status trends, and external influencing factors, it calculates energy storage action tendency values. Then, through activation, judgment, and power mapping, it generates target charging and discharging power, forming a complete implementation process from data processing to energy storage scheduling. This allows the terminal to incorporate the historical behavioral characteristics of household energy consumption scenarios, external environmental factors, and the operating status of energy storage batteries into the same scheduling system. This enables the generation of energy storage actions based on joint decision-making using comprehensive information, significantly improving the rationality and implementability of energy storage scheduling strategies. It ensures that the energy storage system can output charging and discharging power that meets actual operating requirements under different loads, electricity prices, and battery states.

[0127] The above provides a detailed description of the home energy storage charging and discharging optimization method based on multidimensional load prediction in the embodiments of this application. The following will provide a detailed description of the home energy storage charging and discharging optimization system and device based on multidimensional load prediction.

[0128] Please see Figure 3 This application provides an embodiment of a home energy storage charging and discharging optimization system based on multidimensional load forecasting, which includes: Preprocessing unit 301 is used to preprocess multi-source data related to household electricity consumption to obtain the input structure; The construction unit 302 is used to identify multiple household energy consumption scenarios based on the input structure and construct corresponding energy consumption scenario parameters. The parameters carried by the energy consumption scenario parameters include the historical power distribution characteristics of the household energy consumption scenario, the start-up time fluctuation model, the behavior priority parameters, and the external environmental influence factors. The calculation unit 303 is used to calculate the potential energy value of the energy scenario parameter in each discrete period within the prediction time window based on the historical power distribution characteristics and behavior priority parameters of each energy scenario parameter, and to obtain the corresponding potential energy curve. The diffusion unit 304 is used to perform time diffusion processing on the potential energy curve based on the start-up time fluctuation model of each energy consumption scenario parameter to obtain the potential energy evolution curve. The superposition unit 305 is used to superimpose the potential energy evolution curves of all energy consumption scenario parameters according to each discrete time period within the prediction time window to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. Input unit 306 is used to input the household total load forecast curve, electricity price information and household energy storage battery status data into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete period within the forecast time window; The control unit 307 is used to generate charge and discharge control commands based on the target charge and discharge power, and control the home energy storage battery to perform charging or discharging actions in corresponding discrete time periods.

[0129] Optionally, input unit 306 is specifically used for: The net load pressure sequence for each discrete period within the forecast time window is calculated based on the total household load forecast curve and reference load value. Calculate the electricity price incentive sequence for each discrete time period relative to the preset electricity price benchmark based on electricity price information; Based on the state data of home energy storage batteries, the feasible range of charge and discharge power for each discrete time period is determined; Calculate the state trend terms for each discrete time period based on the state data of home energy storage batteries; Based on the net load pressure sequence, electricity price incentive sequence, state trend term, and external environmental influencing factors, an energy storage action tendency function is constructed for each discrete time period. The target charge and discharge power is determined within the feasible range of charge and discharge power based on the energy storage action tendency function.

[0130] Optionally, the input unit 306 is further used for: The original energy storage action tendency value for each discrete time period is calculated by integrating the net load pressure sequence, electricity price incentive sequence, state trend term and external environmental influencing factors based on the energy storage action tendency function. The original energy storage action tendency value is subjected to amplitude constraint processing based on the nonlinear activation function with a limited interval, so as to obtain the normalized energy storage action tendency value within the preset range. The type of energy storage action is determined based on the comparison between the normalized energy storage action tendency value and the preset action threshold. Based on the normalized energy storage action tendency value, the target charge and discharge power is determined by amplitude mapping within the corresponding feasible range of charge and discharge power.

[0131] Optionally, the diffusion unit 304 is specifically used for: Based on the historical start-up time distribution type of the start-up time fluctuation model for each energy consumption scenario, the corresponding time diffusion kernel function is determined, and the time diffusion range parameter is calculated based on the time fluctuation amplitude of the start-up time fluctuation model. Convolution operations are performed on each potential energy curve based on the time-diffusion kernel function and the time-diffusion range parameter to obtain the potential energy value distribution after diffusion along the time axis within the prediction time window. Arrange the potential energy values ​​in chronological order to obtain the potential energy evolution curves of the parameters for each energy consumption scenario.

[0132] Optionally, the computing unit 303 is specifically used for: Calculate the expected power value for each discrete time period based on historical power distribution characteristics; The power expectation values ​​for each discrete time period are weighted based on the behavior priority parameter to obtain the basic potential energy value sequence. Based on the external environmental influencing factors, the basic potential energy value sequence is amplified or attenuated to obtain the potential energy values ​​of each energy consumption scenario parameter in each discrete time period, and then arranged in chronological order to form a potential energy curve.

[0133] Optionally, building unit 302 is specifically used for: By dividing the historical power sequence into power intervals and performing statistical processing, the historical power distribution characteristics of each household energy consumption scenario are obtained. A start-up time fluctuation model is fitted based on the start-up time distribution of household energy consumption scenarios in historical power sequences; The behavioral priority parameters are determined based on the energy consumption ratio of each household's energy use scenario within a preset period, as well as the user's set comfort and economic needs. External environmental impact factors are determined based on environmental data, electricity price information, and photovoltaic power generation data. Historical power distribution characteristics, start-up time fluctuation models, behavioral priority parameters, and external environmental impact factors are then incorporated into the corresponding energy consumption scenario parameters.

[0134] Optionally, the preprocessing unit 301 is specifically used for: Time alignment processing is performed on historical household electricity consumption data, real-time electricity consumption data, electricity price information, environmental data, and photovoltaic power generation data to obtain a multi-source time series arranged with a unified time step. Interpolation imputation and outlier removal are performed on missing data in multi-source time series to obtain a data series with missing data filled and outliers removed. Statistical features are extracted from the data sequence according to a preset time granularity to obtain a set of feature vectors for household energy consumption scenario identification. Multiple household energy consumption scenarios are identified using clustering or rule matching based on feature vector sets, and the identification results are associated with corresponding time periods to form an input structure.

[0135] In this embodiment, the functions of each unit are the same as those described above. Figures 1 to 2b The steps in the illustrated embodiments are the same and will not be repeated here.

[0136] Please see Figure 4 This application provides an embodiment of a home energy storage charging and discharging optimization device based on multidimensional load forecasting, comprising: Processor 401, memory 402, input / output unit 403, bus 404; The processor 401 is connected to the memory 402, the input / output unit 403, and the bus 404; Processor 401 performs specific operations Figures 1 to 2b The specific operations corresponding to the steps in the method will not be elaborated here.

[0137] This application also relates to a computer-readable storage medium on which a program is stored, characterized in that, when the program is run on a computer, it causes the computer to perform any of the methods described above.

[0138] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0139] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0140] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0141] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0142] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for optimizing the charging and discharging of household energy storage based on multidimensional load forecasting, characterized in that, The method includes: Preprocessing of multi-source data related to household electricity consumption yields the input structure; Based on the input structure, multiple household energy consumption scenarios are identified, and corresponding energy consumption scenario parameters are constructed. The parameters carried by the energy consumption scenario parameters include the historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters, and external environmental influencing factors of the household energy consumption scenario. Based on the historical power distribution characteristics and behavior priority parameters of each energy consumption scenario parameter, the potential energy value of the energy consumption scenario parameter in each discrete time period within the prediction time window is calculated, and the corresponding potential energy curve is obtained. Based on the start-up time fluctuation model of each of the energy consumption scenario parameters, the potential energy curve is subjected to time diffusion processing to obtain the potential energy evolution curve. Within the prediction time window, the potential energy evolution curves of all the energy consumption scenario parameters are superimposed according to each discrete time period to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. The household total load forecast curve, electricity price information and household energy storage battery status data are input into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window. Based on the target charging and discharging power, a charging and discharging control command is generated to control the home energy storage battery to perform charging or discharging actions in corresponding discrete time periods.

2. The method according to claim 1, characterized in that, The step of inputting the household total load forecast curve, electricity price information, and household energy storage battery status data into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window includes: The net load pressure sequence for each discrete period within the forecast time window is calculated based on the total household load forecast curve and the reference load value. Calculate the electricity price incentive sequence for each discrete time period relative to the preset electricity price benchmark based on electricity price information; Based on the state data of home energy storage batteries, the feasible range of charge and discharge power for each discrete time period is determined; Calculate the state trend items for each discrete time period based on the state data of the home energy storage battery; Based on the net load pressure sequence, the electricity price incentive sequence, the state trend term, and external environmental influencing factors, construct the energy storage action tendency function for each discrete time period; The target charge / discharge power is determined within the feasible range of the charge / discharge power based on the energy storage action tendency function.

3. The method according to claim 2, characterized in that, The step of determining the target charge / discharge power within the feasible range of the charge / discharge power based on the energy storage action tendency function includes: The original energy storage action tendency value for each discrete time period is calculated based on the energy storage action tendency function, which integrates the net load pressure sequence, electricity price incentive sequence, state trend term and external environmental influencing factors. The original energy storage action tendency value is subjected to amplitude constraint processing based on a nonlinear activation function with a limited range to obtain a normalized energy storage action tendency value limited to a preset range. The type of energy storage action is determined based on the comparison between the normalized energy storage action tendency value and the preset action threshold. Based on the normalized energy storage action tendency value, the target charge and discharge power is determined through amplitude mapping within the corresponding feasible range of charge and discharge power.

4. The method according to claim 1, characterized in that, The startup time fluctuation model based on the parameters of each energy consumption scenario performs time diffusion processing on the potential energy curve to obtain the potential energy evolution curve, including: Based on the historical start-up time distribution type of the start-up time fluctuation model for each of the energy consumption scenario parameters, the corresponding time diffusion kernel function is determined, and the time diffusion range parameter is calculated based on the time fluctuation amplitude of the start-up time fluctuation model. Based on the time diffusion kernel function and the time diffusion range parameter, a convolution operation is performed on each of the potential energy curves to obtain the potential energy value distribution after diffusion along the time axis within the prediction time window; Arrange the potential energy values ​​in chronological order to obtain the potential energy evolution curves of the energy consumption scenario parameters.

5. The method according to claim 1, characterized in that, Based on the historical power distribution characteristics and behavioral priority parameters of each energy consumption scenario parameter, the potential energy value of the energy consumption scenario parameter in each discrete time period within the prediction time window is calculated, and the corresponding potential energy curve is obtained, including: Based on the historical power distribution characteristics, the corresponding expected power value is calculated for each discrete time period; The power expectation values ​​for each discrete time period are weighted based on the aforementioned behavior priority parameters to obtain a sequence of basic potential energy values. Based on the external environmental influencing factors, the basic potential energy value sequence is amplified or attenuated to obtain the potential energy values ​​of each energy consumption scenario parameter in each discrete time period, and arranged in chronological order to form the potential energy curve.

6. The method according to any one of claims 1 to 5, characterized in that, The process of identifying multiple household energy consumption scenarios based on the input structure and constructing corresponding energy consumption scenario parameters includes: The historical power sequence is divided into power intervals and statistically processed to obtain the historical power distribution characteristics of each household energy consumption scenario; A start-up time fluctuation model is fitted based on the start-up time distribution of household energy consumption scenarios in the historical power sequence; The behavioral priority parameters are determined based on the energy consumption ratio of each household's energy use scenario within a preset period, as well as the user's set comfort and economic needs. External environmental impact factors are determined based on environmental data, electricity price information, and photovoltaic power generation data. The historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters, and external environmental impact factors are then written into the corresponding energy consumption scenario parameters.

7. The method according to any one of claims 1 to 5, characterized in that, The preprocessing of multi-source data related to household electricity consumption yields an input structure, including: Time alignment processing is performed on historical household electricity consumption data, real-time electricity consumption data, electricity price information, environmental data, and photovoltaic power generation data to obtain a multi-source time series arranged with a unified time step. Interpolation imputation and outlier removal are performed on the missing data in the multi-source time series to obtain a data series with missing data filled and outliers removed. The data sequence is subjected to statistical feature extraction at a preset time granularity to obtain a set of feature vectors for household energy consumption scenario identification; Based on the feature vector set, multiple household energy consumption scenarios are identified using clustering or rule matching methods, and the identification results are associated with the corresponding time periods to form the input structure.

8. A home energy storage charging and discharging optimization system based on multidimensional load forecasting, characterized in that, The system includes: The preprocessing unit is used to preprocess multi-source data related to household electricity consumption to obtain the input structure; The construction unit is used to identify multiple household energy consumption scenarios based on the input structure and construct corresponding energy consumption scenario parameters. The parameters carried by the energy consumption scenario parameters include the historical power distribution characteristics, start-up time fluctuation model, behavior priority parameters and external environmental influence factors of the household energy consumption scenario. The calculation unit is used to calculate the potential energy value of the energy consumption scenario parameter in each discrete time period within the prediction time window based on the historical power distribution characteristics and behavior priority parameters of each energy consumption scenario parameter, and obtain the corresponding potential energy curve. A diffusion unit is used to perform time diffusion processing on the potential energy curve based on the start-up time fluctuation model of each of the energy consumption scenario parameters to obtain the potential energy evolution curve. The superposition unit is used to superimpose the potential energy evolution curves of all the energy consumption scenario parameters in each discrete time period within the prediction time window to obtain the household total load prediction curve. The discrete time period is the time period obtained by discretizing the prediction time window at fixed time intervals. The input unit is used to input the household total load forecast curve, electricity price information and household energy storage battery status data into the energy consumption scenario parameter game scheduling model to obtain the target charging and discharging power for each discrete time period within the forecast time window; The control unit is used to generate charge and discharge control commands based on the target charge and discharge power, and control the home energy storage battery to perform charging or discharging actions in corresponding discrete time periods.

9. A household energy storage charging and discharging optimization device based on multidimensional load forecasting, characterized in that, The device includes: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains a program that, when executed on a computer, performs the method as described in any one of claims 1 to 7.

Images