A multi-scale power grid supply-demand balance regulation method, system and device

Abstract

A multi-scale power grid supply and demand balance regulation method, system and device, the method first acquires the historical data of clean energy, comprehensive load and environment of the target power grid, inputs the new energy and environment data into the fluctuation model to generate a multi-scale clean energy output prediction curve, inputs the load data into the load model to generate a comprehensive load prediction curve, then imports the two types of curves into the energy storage prediction model to output a multi-scale energy storage charging and discharging prediction curve, finally inputs all the prediction curves into the supply and demand balance model to calculate and generate regulation instructions, and completes power grid power regulation relying on the instructions; the present application faces short-time, intra-day, medium and long-term multi-scale operation scenarios of the power grid, optimizes new energy, load, energy storage prediction and power grid supply and demand regulation accuracy by constructing multi-scale models, effectively improves the energy storage charging and discharging power prediction accuracy and the regulation accuracy of power grid supply and demand matching.

Description

Technical Field

[0001] This invention relates to power grid supply and demand balance regulation methods, belonging to the field of power grid regulation technology, and particularly to a multi-scale power grid supply and demand balance regulation method, system and equipment. Background Technology

[0002] As the grid connection penetration rate of clean energy such as wind power and photovoltaics continues to increase, the composition of the power grid's comprehensive energy load is becoming more diversified. Coupled with the large-scale participation of energy storage devices in grid regulation, the complexity and dynamic fluctuation characteristics of grid operation continue to intensify, placing stringent requirements on the accuracy, response timeliness, and full-level adaptability of supply and demand balance regulation.

[0003] Existing power grid supply and demand balance control schemes fail to differentiate between the multi-timescale operating characteristics of the power grid, have incomplete coverage of power output forecast dimensions, and lack a differentiated load classification and processing mechanism, resulting in large overall load forecast errors. At the same time, energy storage control is not optimized and adapted to multi-scale operating conditions, and the accuracy of charging and discharging power forecasts is poor. Affected by the above factors, the power grid supply and demand balance lacks multi-scale and multi-level control methods, and the overall control accuracy is insufficient, which restricts the safe, efficient and economical operation of the power grid. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned defects and problems in the prior art and to provide a multi-scale power grid supply and demand balance regulation method, system and equipment with high regulation accuracy.

[0005] To achieve the above objectives, the technical solution of the present invention is: a multi-scale power grid supply and demand balance regulation method, comprising:

[0006] Acquire historical clean energy data, comprehensive load data, and environmental data of the target power grid;

[0007] Historical clean energy data and environmental data are input into a multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain multi-scale clean energy output prediction curves.

[0008] The historical data of the comprehensive load is input into the power grid load simulation model to simulate the dynamic changes of the load and obtain the comprehensive load prediction curve.

[0009] The multi-scale clean energy output prediction curve and the comprehensive load prediction curve are input into the energy storage regulation prediction model to predict the energy storage charge and discharge, and the multi-scale energy storage charge and discharge prediction curve is obtained.

[0010] The multi-scale clean energy output forecast curve, the comprehensive load forecast curve, and the multi-scale energy storage charging and discharging forecast curve are input into the supply and demand balance regulation model for supply and demand regulation analysis, generating regulation commands, and regulating the output of the target power grid based on the regulation commands.

[0011] Optionally, the step of inputting historical clean energy data and historical environmental data into a multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain a multi-scale clean energy output prediction curve specifically includes:

[0012] The multi-scale clean energy fluctuation simulation model includes micro-scale units, meso-scale units, and macro-scale units.

[0013] Acquire real-time output data, daily output data, and seasonal output data from historical clean energy data, as well as real-time sunlight, real-time wind speed, daily sunlight, daily wind speed, seasonal sunlight, and seasonal wind speed from historical environmental data.

[0014] Real-time power output data, real-time illumination, and real-time wind speed are input into microscale units, and short-term power output fluctuations are simulated using the random forest algorithm to obtain short-term power output prediction curves.

[0015] Intraday power output data, intraday sunshine and intraday wind speed are input into the mesoscale unit, and intraday power output fluctuations are simulated through time series prediction algorithms to obtain intraday power output prediction curves.

[0016] Seasonal power output data, seasonal sunlight, and seasonal wind speed are input into macro-scale units, and seasonal power output fluctuations are simulated through trend analysis algorithms to obtain long-term power output prediction curves.

[0017] The short-term output forecast curve, intraday output forecast curve, and long-term output forecast curve are combined to form a multi-scale clean energy output forecast curve.

[0018] Optionally, the step of inputting historical comprehensive load data into the power grid load simulation model to simulate dynamic load changes and obtain a comprehensive load forecast curve specifically includes:

[0019] The power grid load simulation model includes a load classification unit and a load prediction unit;

[0020] The comprehensive historical load data is input into the load classification unit for classification, and residential load historical data, industrial load historical data, commercial load historical data and emerging load historical data are obtained.

[0021] Historical data on residential load, industrial load, commercial load, and emerging load are input into the load forecasting unit. Load forecasting is performed through time-series forecasting and scenario adjustment algorithms to obtain residential load forecasting curves, industrial load forecasting curves, commercial load forecasting curves, and emerging load forecasting curves.

[0022] The residential load forecast curve, industrial load forecast curve, commercial load forecast curve, and emerging load forecast curve are combined to form a comprehensive load forecast curve.

[0023] Optionally, the multi-scale clean energy fluctuation simulation model and the power grid load simulation model are constrained by historical power grid operating data;

[0024] The historical power grid operation data includes historical power grid status data, power grid topology data, and power grid equipment data;

[0025] The historical power grid status data includes node voltage, line current, active power, reactive power, system frequency, power factor, line loss rate, and power fluctuation at clean energy grid connection points.

[0026] The power grid topology data includes power grid wiring methods, node and branch connection relationships, line impedance, line admittance, substation main wiring layout, and topology change records.

[0027] The power grid equipment data includes rated capacity, operating efficiency, and current carrying capacity.

[0028] Optionally, the step of inputting the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to predict energy storage charging and discharging, and obtaining the multi-scale energy storage charging and discharging prediction curve, specifically includes:

[0029] The energy storage regulation prediction model includes an energy storage type adaptation unit and a dynamic regulation strategy unit.

[0030] The multi-scale clean energy output prediction curve and the comprehensive load prediction curve are input into the energy storage type adaptation unit to adapt the operating conditions of different types of energy storage devices in the target power grid and obtain the adapted multi-scale output prediction curve and the adapted load prediction curve.

[0031] The adaptive multi-scale output prediction curve and the adaptive load prediction curve are input into the dynamic adjustment strategy unit. The charging and discharging quantities are calculated based on the fluctuation characteristics of the adaptive multi-scale output prediction curve and the change characteristics of the adaptive load prediction curve to obtain the multi-scale energy storage charging and discharging prediction curve.

[0032] Optionally, the step of inputting the multi-scale clean energy output prediction curve, the comprehensive load prediction curve, and the multi-scale energy storage charging and discharging prediction curve into the supply and demand balance regulation model for supply and demand regulation analysis and generating regulation commands specifically includes:

[0033] The supply and demand balance adjustment model includes a balance calculation unit, a supply and demand trend prediction unit, and an imbalance optimization adjustment unit.

[0034] The multi-scale clean energy output prediction curve, the comprehensive load prediction curve, and the multi-scale energy storage charging and discharging prediction curve are input into the balance calculation unit to calculate the supply-demand difference and obtain the multi-scale balance difference prediction curve.

[0035] The multi-scale balance difference prediction curve is input into the supply and demand trend prediction unit, and the imbalance level is predicted by the time series trend analysis algorithm to obtain the multi-scale imbalance level prediction curve.

[0036] The multi-scale imbalance level prediction curve is input into the imbalance optimization adjustment unit to solve the adjustment parameters, obtain the real-time adjustment parameters, intraday adjustment parameters and seasonal adjustment parameters, and generate adjustment instructions based on the real-time adjustment parameters, intraday adjustment parameters and seasonal adjustment parameters.

[0037] Optionally, the energy storage regulation prediction model and the supply-demand balance regulation model are constrained by energy storage device data;

[0038] The energy storage device data includes rated energy storage performance data, historical energy storage operating status data, and energy storage network control data.

[0039] The rated performance data of the energy storage includes rated capacity, charge and discharge power, conversion efficiency, upper and lower limits of SOC, response time, grid-connected rated voltage, and grid-connected rated frequency;

[0040] The historical operating status data of the energy storage includes historical SOC changes of the energy storage device, actual charging and discharging power and capacity, start-stop records, fault data, grid connection switching records and off-grid switching records;

[0041] The energy storage network control data includes virtual inertia, voltage and frequency support thresholds, charging and discharging triggering conditions, multi-energy storage collaborative control strategies, and scheduling communication parameters.

[0042] A multi-scale power grid supply and demand balance regulation system, which is applied to the above-mentioned method, the system comprising:

[0043] The data acquisition module is used to acquire historical clean energy data, historical comprehensive load data, and historical environmental data of the target power grid.

[0044] The multi-scale clean energy output prediction curve acquisition module is used to input historical clean energy data and environmental historical data into the multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain the multi-scale clean energy output prediction curve.

[0045] The integrated load forecasting acquisition module is used to input historical integrated load data into the power grid load simulation model to simulate dynamic load changes and obtain the integrated load forecast curve.

[0046] The multi-scale energy storage charge and discharge prediction curve acquisition module is used to input the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to perform energy storage charge and discharge prediction and obtain the multi-scale energy storage charge and discharge prediction curve.

[0047] The output regulation module is used to input the multi-scale clean energy output prediction curve, the comprehensive load prediction curve, and the multi-scale energy storage charging and discharging prediction curve into the supply and demand balance regulation model for supply and demand regulation analysis, generate regulation commands, and regulate the output of the target power grid based on the regulation commands.

[0048] A multi-scale power grid supply and demand balance regulation device, the device including a processor and a memory;

[0049] The memory is used to store computer program code and to transmit the computer program code to the processor;

[0050] The processor is used to execute the above-described multi-scale power grid supply and demand balance regulation method according to the instructions in the computer program code.

[0051] A computer-readable storage medium storing computer-executable instructions, which, when executed on a computer, implement the aforementioned multi-scale power grid supply and demand balance regulation method.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] 1. In the multi-scale power grid supply and demand balance regulation method, system, and equipment of the present invention, the method first acquires historical data on clean energy, comprehensive load, and environment of the target power grid, and inputs the historical data on clean energy and environment into a multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain a multi-scale clean energy output prediction curve. Then, the historical data on comprehensive load is input into a power grid load simulation model to perform load dynamic change simulation and obtain a comprehensive load prediction curve. Next, the multi-scale clean energy output prediction curve and the comprehensive load prediction curve are input into an energy storage regulation prediction model to perform energy storage charging and discharging prediction and obtain a multi-scale energy storage charging and discharging prediction curve. Finally, the above prediction curves are input into a supply and demand balance regulation model for supply and demand regulation analysis to generate a regulation curve. This invention addresses multi-scale operation scenarios of the power grid, including short-term, intraday, and medium-to-long-term scenarios. It establishes a multi-scale clean energy fluctuation simulation model, a power grid load simulation model, an energy storage regulation and prediction model, and a segmented supply-demand balance optimization model. The clean energy fluctuation simulation model simulates multi-time-series power generation output, improving the comprehensive prediction capability of new energy output. The power grid load simulation model enables refined load classification modeling, effectively reducing inherent system errors in comprehensive load prediction. The energy storage regulation and prediction model matches differentiated operating conditions, improving the accuracy of energy storage charging and discharging power prediction. Finally, the supply-demand balance model implements hierarchical and segmented coordinated regulation, thereby improving the overall precision of power grid supply-demand matching.

[0054] 2. In the multi-scale power grid supply and demand balance regulation method, system and equipment of the present invention, a multi-scale clean energy fluctuation simulation model is used to adapt random forest, time series forecasting and trend analysis algorithms for different scales such as short time, intraday and long time. Combined with environmental data, the power output fluctuation is simulated at different scales. The whole process is constrained by the power grid topology and equipment operating limits. The output multi-scale power output prediction curve can accurately match the characteristics of short-term strong fluctuation, intraday regularity and seasonal trend of clean energy, and improve the prediction accuracy of clean energy power output. The power grid load simulation model divides the comprehensive load of the power grid into four categories: residential, industrial, commercial and emerging. The classification mapping function achieves accurate division without omission and repetition. Then, time series forecasting is used for the operating characteristics of each type of load, which greatly reduces the error of load forecasting.

[0055] 3. In the multi-scale power grid supply and demand balance regulation method, system, and equipment of this invention, the rated performance, historical operating status, and grid configuration control data of energy storage are integrated into the energy storage regulation prediction model. A matching rule library between the multi-scale operating conditions of the power grid and different energy storage devices is established through the energy storage type adaptation unit, realizing accurate adaptation of energy storage and the power grid at multiple scales. At the same time, the charging and discharging quantity is calculated through the dynamic adjustment strategy unit, so that the charging and discharging regulation of the energy storage device is fully aligned with the actual needs of the power grid, improving the prediction accuracy of energy storage charging and discharging. Through the supply and demand balance regulation model, a hierarchical regulation command is generated through the process of balance calculation, imbalance level prediction, and optimization of regulation parameters, realizing full-dimensional coverage from short-term emergency regulation to intraday scheduling regulation and seasonal planning regulation, thus improving the regulation accuracy of the power grid. Attached Figure Description

[0056] Figure 1 This is a flowchart of the method of the present invention.

[0057] Figure 2 This is a system structure diagram of the present invention.

[0058] Figure 3 This is a structural diagram of the device of the present invention.

[0059] In the figure: Data acquisition module 1, multi-scale clean energy output prediction curve acquisition module 2, comprehensive load prediction acquisition module 3, multi-scale energy storage charge and discharge prediction curve acquisition module 4, output adjustment module 5, processor 6, memory 7, computer program code 71. Detailed Implementation

[0060] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] Example 1:

[0062] See Figure 1 A multi-scale power grid supply and demand balance regulation method, comprising:

[0063] S100: Obtain historical clean energy data, historical comprehensive load data, and historical environmental data of the target power grid;

[0064] In this embodiment, the historical clean energy data refers to the output data, as well as the grid connection parameters and equipment operating status of the power plant, collected from the grid dispatch automation system and the clean energy power plant SCADA system at the real-time, intraday, and seasonal scales, and archived at the minute, hour, and daily levels.

[0065] The comprehensive historical load data refers to data retrieved from the power grid electricity consumption information collection system and load management terminal, and the historical power and electricity consumption of residential, industrial, commercial and emerging loads, as well as peak and valley characteristic data of various loads, collected at the minute and hour level, and the load data of different periods in the past 3-5 years are archived simultaneously.

[0066] The aforementioned historical environmental data refers to: integrating data from the power grid's own meteorological monitoring stations and third-party meteorological platforms, collecting real-time sunlight, wind speed, and intraday and seasonal meteorological indicators that are strongly correlated with clean energy, and supplementing the collection of data such as temperature and humidity, with the collection frequency matching the spatiotemporal data of historical clean energy data.

[0067] The above three types of raw data are preprocessed, outliers are removed and redundant data are deleted using the 3σ principle, and missing data are filled in by methods such as linear interpolation and nearest neighbor mean. Then, data of different dimensions are normalized to unify units and formats. Finally, all data timestamps are calibrated to the power grid standard time to ensure spatiotemporal matching of data in the same time dimension.

[0068] Based on the preprocessed data, a target power grid time-series data fusion database is built. A three-level storage structure is established according to time scale, data type, and spatial dimension. At the same time, an association index is built for the data, binding clean energy output with corresponding environmental data and regional load with corresponding power grid topology nodes, so as to enable rapid data retrieval in the future.

[0069] S200. Input historical clean energy data and environmental historical data into the multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain the multi-scale clean energy output prediction curve.

[0070] Furthermore, historical clean energy data and historical environmental data are input into a multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain multi-scale clean energy output prediction curves, specifically including:

[0071] The multi-scale clean energy fluctuation simulation model includes micro-scale units, meso-scale units, and macro-scale units.

[0072] Acquire real-time output data, daily output data, and seasonal output data from historical clean energy data, as well as real-time sunlight, real-time wind speed, daily sunlight, daily wind speed, seasonal sunlight, and seasonal wind speed from historical environmental data.

[0073] Real-time power output data, real-time illumination, and real-time wind speed are input into microscale units, and short-term power output fluctuations are simulated using the random forest algorithm to obtain short-term power output prediction curves.

[0074] Intraday power output data, intraday sunshine and intraday wind speed are input into the mesoscale unit, and intraday power output fluctuations are simulated through time series prediction algorithms to obtain intraday power output prediction curves.

[0075] Seasonal power output data, seasonal sunlight, and seasonal wind speed are input into macro-scale units, and seasonal power output fluctuations are simulated through trend analysis algorithms to obtain long-term power output prediction curves.

[0076] The short-term output forecast curve, intraday output forecast curve, and long-term output forecast curve are combined to form a multi-scale clean energy output forecast curve.

[0077] In this embodiment, the multi-scale clean energy fluctuation simulation model relies on power grid topology data to clarify the node characteristics of clean energy grid connection points and transmission line boundaries. It combines the rated output and grid connection parameters of clean energy power generation equipment in power grid equipment data to define simulation limits. Based on the scaled operation rules of historical power grid data, micro, meso, and macro scale units are built. The micro unit is equipped with a random forest algorithm, the meso unit is equipped with a time series prediction algorithm, and the macro unit is embedded with a trend analysis algorithm. Each unit calibrates the algorithm parameters with three types of power grid data to realize the simulation of short-term, intraday, and long-term power output fluctuations of clean energy, and unifies the time granularity and format of the output data of each unit.

[0078] Historical clean energy data and environmental data at various scales are directly input into the corresponding scale units. A dedicated algorithm simulates power output fluctuations. All calculations are constrained by historical grid state data S, grid topology data T, and grid equipment data D. The output is a multi-scale power output prediction curve. The expressions for each scale are independent and match the analysis requirements of the corresponding time dimension, as detailed below:

[0079] In this scheme, all calculations are constrained by historical power grid state data S, power grid topology data T, and power grid equipment data D. This does not merely refer to the enumeration of data types, but rather that S, T, and D take effect in four stages: construction of input features for the prediction model, calibration of model parameters, correction of prediction results, and compliance verification. Specifically, historical power grid state data S reflects the actual operating state of the target power grid at different time scales; power grid topology data T determines the nodes, branches, and power flow transmission boundaries of the prediction object; and power grid equipment data D limits the prediction results to ensure they do not exceed the rated capacity, current carrying capacity, operating efficiency, and permissible operating range of the equipment.

[0080] Specifically, in the multi-scale clean energy output prediction process, firstly, based on the grid topology data T, the corresponding grid-connected nodes, upstream substations, related lines, and transmission channels of the clean energy power plants are determined, and topological parameters such as node number, branch connection relationship, line impedance, and line admittance are embedded as spatial constraints into the model input. Secondly, based on the grid equipment data D, the rated capacity, current carrying capacity, and operating efficiency of grid-connected equipment, lines, and transformers are read to form the output upper limit, ramp limit, and transmission capacity limit. Then, using the node voltage, line current, active power, reactive power, system frequency, line loss rate, and grid-connected point power fluctuations in the historical state data S, state label matching and parameter calibration are performed on the model training samples to ensure that the prediction results conform to the historical operating patterns of the node. After the model outputs, the predicted output is corrected through a compliance verification function: when the predicted value causes the node voltage to exceed the limit, the line current to exceed the limit, the line power to exceed the equipment capacity, or fails to meet the historical fluctuation boundaries, the prediction curve is peak-shaving, amplitude-limiting, or redistributed until the common constraints of S, T, and D are met.

[0081] In the comprehensive load forecasting process, grid topology data T is used to determine the grid nodes, power supply zones, and transformer substations to which various types of loads belong, avoiding mismatches between load classification and topology nodes; historical grid state data S is used to identify load fluctuation patterns, peak-valley characteristics, and power factor variation patterns at different nodes and time periods; and grid equipment data D is used to limit the carrying capacity of each node, line, and transformer. After residential, industrial, commercial, and emerging loads are forecasted separately, they are not simply added together, but rather aggregated and superimposed according to topology nodes, and the superimposed node loads are verified to ensure they do not exceed the rated capacity and current carrying capacity of the corresponding equipment. If they exceed equipment constraints, corrections are made based on historical load transfer patterns, demand response capabilities, and the carrying capacity margin of adjacent nodes, thereby forming a comprehensive load forecast curve that satisfies the physical structure of the grid and the operating limits of the equipment.

[0082] In the process of energy storage charging and discharging prediction and supply and demand balance regulation, S, T, and D further serve as regulation boundaries. The grid topology data T is used to determine the energy storage access node and its adjustable influence range. The grid equipment data D is used to determine the power variation range that the node and related branches can withstand. The grid historical state data S is used to determine the time pattern and fluctuation intensity of supply and demand imbalance in the region. The energy storage charging and discharging capacity is not only determined by the difference between clean energy output and load, but also needs to meet the requirements of access node voltage, line current, equipment capacity, power flow direction, and historical operating stability. When the predicted energy storage charging and discharging power may cause node voltage to exceed the limit, line overload, or equipment to operate beyond its limit, the system will limit the amplitude according to the equipment limits in the grid equipment data D, and redistribute the regulation amount according to the topology connection relationship in T, and finally generate an energy storage charging and discharging prediction curve that meets the grid safe operation boundary.

[0083] Furthermore, the constraint rules for power grid historical state data S, power grid topology data T, and power grid equipment data D differ in short-term, medium-term, and long-term multi-scale forecasting:

[0084] The short-term forecast scale is 0-1 hour into the future, outputting minute-level forecast curves. This scale is primarily constrained by real-time or near-real-time factors such as S, T, and D. At this scale, S mainly includes node voltage, line current, system frequency, real-time active / reactive power, and short-term fluctuations at clean energy grid connection points. T is mainly used to determine real-time power flow paths and the effective network structure after faults, maintenance, and changes in switch status. D is mainly used to constrain the instantaneous current carrying capacity, short-term overload capacity, energy storage response time, and upper limits of charging and discharging power. Therefore, S has the highest weight in short-term forecasting, while T and D are used for real-time safety verification and limiting correction, ensuring that the forecast curve reflects short-term fluctuations and meets immediate adjustment requirements.

[0085] The medium-term forecast scale is for the next 1-24 hours, outputting hourly forecast curves. This scale focuses on combining intraday load factors (S), time factors (T), and time factors (D). At this scale, S is used to extract intraday load peaks and valleys, intraday output changes of renewable energy sources, power factor, and line loss variation patterns; T is used to constrain power flow distribution under different operating modes, regional power supply relationships, and intraday maintenance plans; D is used to constrain the capacity, efficiency, and temperature rise tolerance of transformers, lines, switchgear, and energy storage devices under continuous operating conditions. Therefore, in the medium-term forecast, S, T, and D work together to optimize output plans, load curves, and energy storage charging and discharging sequences, with a focus on ensuring the executability of intraday scheduling plans.

[0086] The long-term forecast scale is from 1 day to 12 months, outputting daily or monthly forecast curves. This scale is primarily constrained by seasonality factors such as S, T, and D. At this scale, S is used to extract seasonal load changes, seasonal output trends of new energy sources, historical extreme operating conditions, and long-term supply and demand fluctuations; T reflects grid structure, planned access points, topology changes from renovations and expansions, and inter-regional transmission capacity; D constrains long-term equipment capacity margins, maintenance cycles, efficiency degradation, and lifespan constraints. Therefore, in long-term forecasting, T and D have a stronger role in defining the planning boundaries of the forecast results, while S is used for trend calibration, focusing on ensuring that long-term output forecasts, load forecasts, and energy storage strategies meet the requirements of grid planning and the long-term safe operation of equipment.

[0087] Input real-time historical clean energy output data, real-time solar irradiance data, and real-time wind speed data. Use the random forest algorithm to simulate short-term output fluctuations for the next 0-1 hours, and output a short-term output prediction curve with a time series of minutes. Its expression is as follows:

[0088] ;

[0089] in: This is the random forest function; For real-time lighting; Real-time wind speed; This indicates that the constraints satisfy the historical state data S of the power grid, the topology data T of the power grid, and the equipment data D of the power grid are met.

[0090] Input historical clean energy output data, daily sunshine and wind speed data, and use a time-series forecasting algorithm to simulate daily output fluctuations for the next 1-24 hours, outputting hourly time-series daily output forecast curves. Its expression is as follows:

[0091] ;

[0092] in: For time series prediction functions; Intraday sunlight; Wind speed for the day;

[0093] Input historical seasonal clean energy output data, seasonal sunshine and wind speed data, and simulate seasonal output fluctuations over the next 1 day to 12 months using trend analysis algorithms, outputting a long-term output prediction curve with a daily time series. Its expression is as follows:

[0094] ;

[0095] in: For trend analysis functions; For seasonal sunlight; Seasonal wind speed;

[0096] Centered on the time axis, a curve integration operator is used. The short-term, intraday, and long-term power output forecast curves are seamlessly stitched together in chronological order. Simultaneously, a final compliance calibration is performed using historical grid operation data (S, T, D) as constraints, resulting in a multi-scale clean energy power output forecast curve that incorporates short-term fluctuations, intraday variations, and seasonal trends. This provides data support for subsequent energy storage charging and discharging predictions and grid supply and demand balance adjustments. Its expression is as follows:

[0097] ;

[0098] Among them, the short-term output forecast curve is used as the real-time segment, the intraday output forecast curve is used as the follow-up segment, and the long-term output forecast curve is used as the long-term segment. The three are connected along the time axis to form a complete multi-scale clean energy output forecast curve.

[0099] In this solution, instead of simply using existing algorithms for independent prediction, the solution retains the classic algorithm framework and combines the multi-scale operation characteristics of the power grid, the physical constraints of the power grid, and the business requirements of supply and demand balance regulation. It makes targeted structural improvements and collaborative integration designs for the random forest algorithm, time series prediction algorithm, and trend analysis algorithm, forming a combined prediction model of "scale-based improved algorithm + cross-validation fusion mechanism", which is different from the existing single algorithm or simple parallel application mode.

[0100] In short-term forecasting, based on the random forest algorithm, a power grid state-aware feature weighting mechanism is introduced in micro-scale (0-1 hour) forecasting to address the characteristics of strong short-term fluctuations, high nonlinearity, and strong state sensitivity of the power grid. This mechanism uses node voltage, frequency fluctuations, and power fluctuations from historical power grid state data S as dynamic weighting factors to dynamically adjust feature importance in real time, enhancing the model's responsiveness to key features under abnormal fluctuations or disturbances. An embedded equipment operation constraint screening mechanism uses rated capacity, ramp rate, and current carrying capacity from power grid equipment data D as sample screening conditions to eliminate abnormal samples that do not meet equipment operation constraints during the training phase, improving the model's physical consistency. Furthermore, a topology-aware input is added, transforming power grid topology data T into node embedding vectors or adjacency features, giving the random forest model basic spatial correlation capabilities and avoiding the problem of traditional models that only rely on time series data and ignore the influence of power grid structure.

[0101] In medium-term forecasting, based on time-series forecasting algorithms, a multi-source data coupling input structure is constructed for the meso-scale (1-24 hours) forecasting, taking into account the characteristics of obvious intraday periodicity and strong coupling between load and new energy. This structure integrates clean energy output, load classification data, environmental data, and historical grid state data S into a multi-dimensional input, avoiding traditional single-variable or weakly coupled forecasting. A topology partitioning constraint mechanism is also introduced, dividing the power supply area and load partition according to the grid topology data T. A partition-independent forecasting and cross-regional correlation correction mechanism is established in the model to ensure that the forecast results conform to the actual power supply structure of the grid. At the same time, continuous operation constraints of equipment are embedded, transforming transformer capacity, line thermal stability limit, and intraday charging and discharging capacity of energy storage in the grid equipment data D into time-series constraints. This dynamically limits and smooths the forecast curve to avoid situations where intraday scheduling is not feasible.

[0102] In long-term forecasting, based on trend analysis algorithms, in macro-scale (1 day to 12 months) forecasting, considering the characteristics of long-term trend dominance and weak random fluctuations, seasonal statistical features (such as seasonal load average, extreme value distribution, and fluctuation range) from the historical power grid state data S are introduced to calibrate the trend function, improving the stability of trend fitting. Furthermore, the forecasting model is structurally updated by combining power grid topology data T (such as new lines, grid adjustments, and changes in grid connection points), ensuring that long-term forecasts reflect power grid planning and structural changes. Simultaneously, equipment lifecycle and performance degradation data from the power grid equipment data D are integrated to perform long-term corrections to equipment capacity margins and efficiency changes, ensuring that the forecast results conform to long-term equipment operation constraints.

[0103] Furthermore, unlike existing technologies where various prediction algorithms can only run independently and the prediction results are simply spliced ​​or weighted averaged, this application proposes a multi-scale cross-validation fusion mechanism. First, it involves independent modeling at each scale plus cross-scale consistency verification. Based on the improved random forest, time series prediction, and trend analysis models, short-term, medium-term, and long-term prediction results are generated independently, respectively. After the results are output, verification is performed based on cross-scale consistency constraints, requiring that the short-term prediction integral results be consistent with the data in the corresponding time window of the intraday prediction, and that the cumulative trend of the intraday prediction match the direction of the long-term trend. If a deviation occurs between scales, the local prediction results are immediately corrected through a correction function. Second, a multi-model error feedback correction mechanism is built to construct an error evaluation function, statistically analyze the prediction errors of each model's historical backtesting, and then generate an error weight matrix. In the actual prediction process, the output weights of each model are dynamically adjusted by parameters characterizing the power grid's operating state to adapt to different power grid operating conditions, allowing the prediction results to adaptively favor the output of the low-error model. At the same time, a fusion result screening mechanism based on physical constraints is added. All preliminary fusion prediction results must be verified by three types of power grid physical constraints: node voltage, line power flow, and equipment capacity and efficiency. If the fusion result violates the constraints, the output of the compliant model is retained first, and the non-compliant results are limited or replaced to ensure that the final prediction results are feasible for engineering implementation. Finally, a time axis continuity fusion mechanism is used to achieve seamless connection of the time axis of short, medium and long-term prediction curves by using time splicing operators. At the same time, a smooth transition function is embedded in the connection interval to avoid the problem of abrupt changes in prediction values ​​caused by model switching, ensuring the continuity and uniformity of the time sequence of multi-scale prediction curves throughout the entire cycle.

[0104] In contrast, most existing power grid forecasting technologies employ a single algorithm or a simple parallel forecasting mode with multiple algorithms. This not only suffers from the defects of each model operating independently and in isolation, lacking data interaction and linkage, and lacking a closed-loop correction mechanism for results, but also fails to incorporate constraints and verification based on power grid topology and equipment operating limitations. Consequently, the forecast results have poor engineering adaptability. Furthermore, the lack of differentiation between forecasting characteristics at multiple time scales means that a general, unified model cannot account for both short-term load fluctuations and long-term operating trends. This application abandons the traditional extensive fusion approach and adopts an integrated technical solution of algorithm improvement, multi-source data embedding, and cross-validation fusion. On the one hand, it adapts dedicated optimized algorithms for different time scales to accurately improve the accuracy of multi-scale forecasts. On the other hand, it embeds three types of core physical data: historical power grid state data (S), power grid topology data (T), and power grid equipment data (D), ensuring that the forecast results conform to the actual operating patterns of the power grid. Simultaneously, relying on two linkage mechanisms—cross-scale cross-validation and dynamic error feedback—it achieves multi-model collaborative correction, eliminating the inherent forecasting bias of a single model. Ultimately, it establishes an integrated closed-loop multi-scale forecasting framework, outputting continuous, stable, power grid-adaptable, and engineering-applicable full-domain forecast curves.

[0105] S300: Input the historical comprehensive load data into the power grid load simulation model to simulate dynamic load changes and obtain the comprehensive load prediction curve;

[0106] Furthermore, the step of inputting historical comprehensive load data into the power grid load simulation model to simulate dynamic load changes and obtain a comprehensive load forecast curve specifically includes:

[0107] The power grid load simulation model includes a load classification unit and a load prediction unit;

[0108] The comprehensive historical load data is input into the load classification unit for classification, and residential load historical data, industrial load historical data, commercial load historical data and emerging load historical data are obtained.

[0109] Historical data on residential load, industrial load, commercial load, and emerging load are input into the load forecasting unit. Load forecasting is performed through time-series forecasting and scenario adjustment algorithms to obtain residential load forecasting curves, industrial load forecasting curves, commercial load forecasting curves, and emerging load forecasting curves.

[0110] The residential load forecast curve, industrial load forecast curve, commercial load forecast curve, and emerging load forecast curve are combined to form a comprehensive load forecast curve.

[0111] In this embodiment, the power grid load simulation model is based on the node load carrying characteristics of power grid topology data and the rated load threshold of power equipment data to build a load classification unit and a load prediction unit. The load classification unit constructs load classification rules based on power grid topology nodes, accurately dividing the comprehensive load historical data into residential, industrial, commercial, and emerging load data. The load prediction unit is configured with time-series prediction and scenario adjustment algorithms, and combines the load change patterns of historical power grid state data to calibrate parameters, and performs multi-scale prediction of the classified loads. The two units work together to output classified load prediction data in a unified format, providing support for the generation of comprehensive load prediction curves, as detailed below:

[0112] First, the historical load data is split into types using a classification mapping function. Then, the load data of each type is input into the time series forecasting and scenario adjustment function. The forecasting process is constrained by historical power grid operating data (S, T, D) throughout the entire process. The forecast curves for each load type are generated independently and contain short-term, intraday, and long-term multi-scale characteristics. Their expressions are as follows:

[0113] ;

[0114] in: Forecast curve of residential load; Industrial load forecast curve; Commercial load forecast curve; For emerging load forecast curves; , , , These are the mapping functions for residential classification, industrial classification, commercial classification, and emerging classification, respectively, and they satisfy the following conditions: , To integrate historical load data and ensure that load classification is complete and without duplication.

[0115] In this embodiment, instead of using static parameter modeling, a targeted improvement to the algorithm structure is made based on the real-time operating characteristics of the power grid, taking into account the traditional time-series forecasting and scenario adjustment algorithms. An improved load forecasting model is constructed, consisting of "dynamic disturbance adaptation + classified load difference modeling + multi-scale feedback correction," thereby enhancing the model's adaptability to sudden load fluctuation scenarios. The details are as follows:

[0116] I. Adaptive Update Mechanism for Dynamic Disturbance Factor:

[0117] In traditional time-series forecasting models, model parameters are typically trained offline based on historical data, making it difficult to cope with sudden disturbances in power grid operation. This technical solution introduces a dynamic disturbance factor set into the time-series forecasting model. And a real-time update mechanism is constructed, the expression of which is as follows:

[0118] ;

[0119] Wherein, the dynamic disturbance factor This includes, but is not limited to: real-time power grid status data (node ​​voltage deviation, frequency fluctuation, active power surges), surges in renewable energy output (sudden changes in wind speed, rapid changes in solar irradiance), abnormal load-side changes (sudden start-up and shutdown of industrial loads, large-scale changes in electricity consumption behavior), meteorological changes (sudden temperature changes, extreme weather), changes in dispatch instructions, and demand response behavior. Furthermore, based on historical power grid status data S and real-time acquired data, a disturbance detection function is constructed to identify various disturbances, calculate disturbance intensity indices, and form a dynamic disturbance weight vector; simultaneously, disturbance factors are... The parameter update function embedded in the time series forecasting model dynamically adjusts the model's weight coefficients, time window weights, and prediction bias terms, enabling the model to automatically improve its response to abnormal changes when sudden disturbances occur, rather than relying on fixed parameter predictions.

[0120] II. Rolling self-calibration mechanism based on real-time data:

[0121] To address the issue that static calibration cannot adapt to dynamic changes in the power grid, this technical solution introduces a rolling self-calibration mechanism, as detailed below:

[0122] Sliding time window update: The latest running data is continuously introduced in a rolling time window manner (such as the most recent 15 minutes, 1 hour, 24 hours) to update the model online;

[0123] Prediction error feedback correction: By constructing an error feedback function, the predicted value is compared with the actual running data, the error is calculated and the model parameters are adjusted in reverse to achieve closed-loop correction;

[0124] State-driven parameter reconfiguration: When a significant change in the power grid operating state is detected (such as a sudden increase in load or equipment switching), the model parameter reconfiguration mechanism is triggered to quickly reset or retrain key parameters, thereby improving the model's adaptability to new operating conditions.

[0125] III. Improved Differentiated Model Structure for Classification Loads:

[0126] To address the distinct characteristics of different load types, including residential, industrial, commercial, and emerging loads, this technical solution restructures the prediction model:

[0127] The residential load adopts a periodic characteristic enhancement model, which focuses on extracting daily and weekly cycle behavior patterns, introducing influencing factors such as temperature and holidays, and regulating abnormal electricity consumption behavior (such as sudden concentrated power consumption) through disturbance factors.

[0128] The industrial load is used to construct a production behavior driven model, which focuses on incorporating features such as equipment operating status, production plan, and start-up and shutdown patterns, and enhances the sensitivity to sudden start-up and shutdown loads, and dynamically adjusts the model weights.

[0129] Commercial load adopts a time-of-day behavior model, which combines business hours, changes in passenger flow, and the impact of holidays to enhance the modeling of intraday fluctuation characteristics.

[0130] Emerging loads (such as charging piles and energy storage loads) are introduced into a strategy-driven model, which uses charging and discharging strategies, price signals, and scheduling instructions as input variables, so that the prediction results can reflect the impact of scheduling behavior.

[0131] The above-mentioned classification loads all construct independent feature selection rules and prediction sub-models, and map them at the topology node level to avoid error accumulation caused by mixed modeling of different types of loads.

[0132] IV. Data feedback and correction links between multi-scale prediction levels:

[0133] To avoid a disconnect between short-term, medium-term, and long-term forecasts, this technical solution constructs a multi-scale feedback correction mechanism:

[0134] Short-term to medium-term correction: Feedback the statistical results of short-term forecast error to the intraday forecast model to make local corrections to the intraday forecast curve;

[0135] Mid-term calibration to long-term: Based on the cumulative deviation of intraday forecasts, the direction of the long-term trend model is corrected to avoid long-term forecasts deviating from the actual operating trend;

[0136] Long-term constraints on short-term: Using long-term trends as boundary conditions, the rationality of short-term forecast results is verified, avoiding abnormal fluctuations in short-term forecasts that do not conform to long-term trends;

[0137] Unified constraint verification: The prediction results at each scale must be finally corrected through a unified verification mechanism of power grid historical state data S, power grid topology data T, and power grid equipment data D to ensure that the prediction results conform to the power grid operation rules and physical constraints.

[0138] This technical solution achieves highly adaptable prediction for complex power grid operating environments by introducing dynamic disturbance adaptive updates, rolling self-calibration mechanisms, differentiated load modeling, and multi-scale feedback correction links, significantly improving prediction accuracy and engineering application feasibility.

[0139] Based on the power grid topology nodes, through the load superposition operator By overlaying the load forecast curves for residential, industrial, commercial, and emerging industries, and performing final compliance calibration with S, T, and D constraints, a continuous, multi-scale integrated load forecast curve is formed. This provides load data support for subsequent energy storage charging and discharging prediction and power grid supply and demand balance regulation. Its expression is as follows:

[0140] ;

[0141] The comprehensive load forecast curve is formed by superimposing the load forecast values ​​of the corresponding nodes according to the load affiliation of each topological node in the power grid, thus forming a comprehensive load forecast curve at the whole network level. The time granularity and multi-scale characteristics of the curve are consistent with the load classification, and it can be directly matched and analyzed with the multi-scale clean energy output forecast curve.

[0142] Furthermore, the multi-scale clean energy fluctuation simulation model and the power grid load simulation model are constrained by historical power grid operating data;

[0143] The historical power grid operation data includes historical power grid status data, power grid topology data, and power grid equipment data;

[0144] The historical power grid status data includes node voltage, line current, active power, reactive power, system frequency, power factor, line loss rate, and power fluctuation at clean energy grid connection points.

[0145] The power grid topology data includes power grid wiring methods, node and branch connection relationships, line impedance, line admittance, substation main wiring layout, and topology change records.

[0146] The power grid equipment data includes rated capacity, operating efficiency, and current carrying capacity.

[0147] In this embodiment, the historical power grid status data S includes operating parameters such as power grid bus node voltage, line current, active power, reactive power, system frequency, power factor, line loss rate, and power fluctuation at clean energy grid connection points. It is collected at multiple times in millisecond, minute, and daily time series, associated with timestamps and power grid operating condition tags, and outliers are removed by the 3σ principle, missing data is filled by linear interpolation, and normalization is performed. This provides power grid operating condition constraints for multi-scale clean energy fluctuation simulation models and power grid load simulation models, serving as the core basis for judging power grid operating conditions.

[0148] The power grid topology data T includes the power grid wiring method, node and branch connection relationship, line impedance, line admittance, substation main wiring layout, topology change records, etc. Based on the digital collection of the power grid GIS system, a node-branch topology model is established and the topology data after the renovation and expansion is updated in real time. The data connectivity is verified, the coding rules are unified, and the power grid structure boundary is provided for various models to ensure that the simulation and prediction results are consistent with the actual physical transmission characteristics of the power grid.

[0149] The power grid equipment data D includes technical parameters such as rated capacity, operating efficiency, and current carrying capacity of power equipment such as generators, transformers, and switchgear, as well as operating data such as historical losses, maintenance, and failure rates. Files are established according to equipment type and associated with power grid topology nodes. Valid data under rated operating conditions is filtered, and equipment performance degradation parameters are updated. Equipment operating threshold constraints are provided for the power grid load simulation model, and power grid-side matching basis is provided for the adaptation of energy storage equipment operating conditions.

[0150] S400: Input the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to perform energy storage charge and discharge prediction, and obtain the multi-scale energy storage charge and discharge prediction curve.

[0151] Furthermore, the step of inputting the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to predict energy storage charging and discharging, and obtaining the multi-scale energy storage charging and discharging prediction curve, specifically includes:

[0152] The energy storage regulation prediction model includes an energy storage type adaptation unit and a dynamic regulation strategy unit.

[0153] The multi-scale clean energy output prediction curve and the comprehensive load prediction curve are input into the energy storage type adaptation unit to adapt the operating conditions of different types of energy storage devices in the target power grid and obtain the adapted multi-scale output prediction curve and the adapted load prediction curve.

[0154] The adaptive multi-scale output prediction curve and the adaptive load prediction curve are input into the dynamic adjustment strategy unit. The charging and discharging quantities are calculated based on the fluctuation characteristics of the adaptive multi-scale output prediction curve and the change characteristics of the adaptive load prediction curve to obtain the multi-scale energy storage charging and discharging prediction curve.

[0155] In this embodiment, the energy storage regulation and prediction model is based on the characteristics of energy storage access nodes in grid topology data, grid connection parameters in grid equipment data, and the charge and discharge boundaries of energy storage rated performance data, combined with the control logic of energy storage grid configuration control data. It establishes an energy storage type adaptation unit and a dynamic adjustment strategy unit. The adaptation unit establishes a matching rule library between grid multi-scale operating conditions and different types of energy storage devices, achieving accurate adaptation of energy storage devices to grid short-term, intraday, and long-term operating conditions, and outputting the adapted output and load prediction data. The dynamic adjustment strategy unit relies on the supply and demand fluctuation patterns of historical grid status data and the actual charge and discharge characteristics of historical energy storage operation data. It configures multi-scale charge and discharge calculation logic, combined with the adjustment threshold of energy storage grid configuration control data, to achieve multi-scale prediction of energy storage charge and discharge, as detailed below:

[0156] Multi-scale clean energy output prediction curves Comprehensive load forecast curve Input energy storage type adaptation unit, through adaptation rule function The process involves matching energy storage with the grid across multiple operating conditions, influenced by both grid-side S, T, and D data and energy storage-side rated performance data. Energy storage network control data The two-dimensional global constraint outputs the adapted segmental-scale output and load curves, as shown in the following expression:

[0157] ;

[0158] in: , , These are the short-term, intraday, and long-term power output prediction curves included in the multi-scale power output prediction curves. , , These are the short-term, intraday, and long-term load forecast curves included in the adaptive load forecast curve;

[0159] In this embodiment, the adaptation rule function is not a traditional single calculation function, but a rule-based matching mechanism based on multi-source power grid data, used to achieve automatic adaptation between energy storage devices and multi-scale operating conditions of the power grid; the function achieves matching through three types of core information, as follows:

[0160] Grid operation status constraint matching: Based on historical grid status data, the power fluctuation characteristics at different time scales are identified so that the energy storage adaptation results can conform to the fluctuation range and change pattern of the actual grid operation, avoiding deviation from the actual operation status.

[0161] Grid topology matching: Based on grid topology data, energy storage devices are limited to their actual access nodes and the electrical range they can affect, so that the energy storage regulation behavior only acts on the area that it can actually reach in the grid, ensuring spatial feasibility.

[0162] Equipment capacity boundary matching: Based on the rated performance of energy storage equipment and the constraints of grid equipment, the charging and discharging power, response capability and operating range of energy storage are limited so that the matching result does not exceed the upper limit of the physical capacity of the equipment.

[0163] Based on the above three types of constraints, the adaptation rule function automatically selects the most suitable energy storage type by comprehensively judging the time scale, spatial topology and equipment capabilities, and generates the corresponding adaptation relationship to realize the dynamic matching of energy storage and the power grid in short-term, intraday and long-term multi-scale scenarios.

[0164] The adaptive multi-scale output prediction curve and the adaptive load prediction curve are input into the dynamic adjustment strategy unit, and the charging and discharging quantity calculation function is used. By combining grid-scale state data with energy storage-scale operation and control data, the predicted values ​​for energy storage charging and discharging at each scale are calculated. Each scale is constrained by the corresponding dimension of grid and energy storage data, and its expression is as follows:

[0165] ;

[0166] in: , , These are short-term, intraday, and long-term historical power grid status data, respectively. , , These are historical energy storage operation data for short-term, intraday, and long-term periods, respectively. , , These are the energy storage charge and discharge prediction curves for short-term, intraday, and long-term timescales, respectively. Positive values ​​indicate energy storage discharge, and negative values ​​indicate energy storage charging.

[0167] In this embodiment, the charge / discharge calculation function collaboratively completes the charge calculation from four mutually restrictive and coordinated levels, as follows:

[0168] Supply and demand gap-driven calculation mechanism: The charging capacity is first determined by the degree of supply and demand imbalance of the power grid at various time scales; when the output of clean energy exceeds the load demand, the system automatically identifies it as a "surplus energy range" and uses this surplus part as the basic source of energy storage charging; at the same time, the charging intensity is dynamically adjusted according to the degree of imbalance.

[0169] Energy storage state constraint calculation mechanism: When calculating the charging amount, the energy storage's own operating state constraints are introduced, including the current SOC level, the remaining space of the SOC upper limit, the current charging efficiency change and the device's sustainable charging capacity. When the energy storage is close to full charge, the charging amount automatically decreases; when the SOC is low, the charging capacity is prioritized to increase.

[0170] Power grid safety constraint verification mechanism: The calculation of charging amount must meet the power grid operation safety boundary constraints, including line overload constraints, transformer capacity constraints, node voltage stability constraints and system frequency stability constraints. When the charging behavior may cause the local power grid to exceed the limit, the system will automatically reduce or adjust the charging amount.

[0171] Multi-scale time coordination mechanism: The calculation of charging capacity is not a decision made at a single point in time, but is coordinated by combining short-term, intraday and long-term scales. Short-term is used to quickly absorb power fluctuations and prioritize response speed; intraday is used to smooth the load curve and optimize the peak-valley structure; long-term is used for energy migration and operation strategy planning. There is a linkage between different time scales. Short-term charging is prioritized to ensure stability, and intraday and long-term charging are optimized and allocated on this basis.

[0172] Centered on the time axis, a curve integration operator is used. The short-term, intraday, and long-term energy storage charging and discharging forecasts are seamlessly stitched together in chronological order, while the grid-side S, T, D and energy storage-side data are compared. , To achieve final compliance calibration across all constraints, a multi-scale energy storage charge and discharge prediction curve is generated that perfectly matches the time dimension of clean energy output and comprehensive load forecast curves. Its expression is as follows:

[0173] ;

[0174] Multi-scale energy storage charge and discharge prediction curves will , , Continuing along the time axis, the time granularity is completely consistent with the multi-scale clean energy output and comprehensive load forecast curves.

[0175] The S500 inputs the multi-scale clean energy output prediction curve, the comprehensive load prediction curve, and the multi-scale energy storage charging and discharging prediction curve into the supply and demand balance regulation model to perform supply and demand regulation analysis, generate regulation commands, and regulate the output of the target power grid based on the regulation commands.

[0176] Furthermore, the input of multi-scale clean energy output prediction curves, comprehensive load prediction curves, and multi-scale energy storage charging and discharging prediction curves into the supply and demand balance regulation model for supply and demand regulation analysis, and the generation of regulation commands, specifically includes:

[0177] The supply and demand balance adjustment model includes a balance calculation unit, a supply and demand trend prediction unit, and an imbalance optimization adjustment unit.

[0178] The multi-scale clean energy output prediction curve, the comprehensive load prediction curve, and the multi-scale energy storage charging and discharging prediction curve are input into the balance calculation unit to calculate the supply-demand difference and obtain the multi-scale balance difference prediction curve.

[0179] The multi-scale balance difference prediction curve is input into the supply and demand trend prediction unit, and the imbalance level is predicted by the time series trend analysis algorithm to obtain the multi-scale imbalance level prediction curve.

[0180] The multi-scale imbalance level prediction curve is input into the imbalance optimization adjustment unit to solve the adjustment parameters, obtain the real-time adjustment parameters, intraday adjustment parameters and seasonal adjustment parameters, and generate adjustment instructions based on the real-time adjustment parameters, intraday adjustment parameters and seasonal adjustment parameters.

[0181] In this embodiment, the supply and demand balance regulation model uses grid-side S, T, and D data to define the regulation boundary and energy storage network control data to define the regulation threshold as dual constraints, and establishes a balance calculation unit, a supply and demand trend prediction unit, and an imbalance optimization regulation unit. The balance calculation unit quantifies and calculates the multi-scale grid supply and demand difference based on the power transmission rules of the grid topology and the operating efficiency of grid equipment. The supply and demand trend prediction unit is equipped with a time-series trend analysis algorithm, which combines the imbalance patterns of historical grid state data to complete the prediction of multi-scale imbalance levels. The imbalance optimization regulation unit takes minimizing the imbalance level as the core optimization objective, configures an optimization solution algorithm, and solves for multi-scale regulation parameters by combining the regulation capacity of grid equipment and energy storage control rules. At the same time, it presets parameter classification and format specifications, classifies parameters according to grid equipment, energy storage devices, and load-side regulation objects, unifies parameter calculation accuracy, and directly connects to subsequent parameter standardization work, as detailed below:

[0182] Based on the scaled clean energy output, comprehensive load, and energy storage charging and discharging forecasts, a supply-demand difference calculation function is used. The multi-scale equilibrium difference prediction curve is calculated, and the entire process is strictly constrained by the power grid topology data T and the power grid equipment data D to ensure that the difference calculation closely matches the physical topology of the power grid and the operating limits of the equipment. Its expression is as follows:

[0183] ;

[0184] in: , , These are the balance difference prediction curves for short-term, intraday, and long-term periods, respectively; taking the short-term period as an example... It directly reflects the degree of matching between power grid output and load at various scales.

[0185] Input the scaled supply and demand difference into the time series trend analysis function. Analyze the trend of the difference, and then use the imbalance level determination function. Output power grid imbalance levels at various scales; time-series trend analysis is based on corresponding short-term, intraday, and long-term historical power grid status data. , , The constraints ensure that the rating determination conforms to the operating patterns of the power grid at different times, and its expression is as follows:

[0186] ;

[0187] in: , , These are the imbalance level prediction curves for short-term, intraday, and long-term periods; the imbalance level determination function. Built-in power grid operation standard thresholds, such as voltage fluctuation range, frequency deviation limit, power deficit and surplus ratio, convert the supply and demand difference after trend analysis into discrete quantitative imbalance level, which facilitates subsequent precise adjustment.

[0188] In this embodiment, the supply-demand difference calculation function calculates the supply-demand difference through the following steps:

[0189] Multi-source data alignment and unified benchmark processing: First, the multi-scale clean energy output prediction curve, comprehensive load prediction curve, and energy storage charging and discharging prediction curve are aligned in time dimension and unified to the standard time scale of the power grid; at the same time, data from different sources are standardized to make them comparable under a unified power benchmark, ensuring the consistency and accuracy of subsequent difference calculations.

[0190] Multi-factor supply and demand composite calculation: Under a unified time base, clean energy output and energy storage discharge are considered together as the "supply side", and load demand and energy storage charging are considered together as the "demand side". Among them, the supply side represents the total available power supply capacity of the power grid; the demand side represents the actual consumption and absorption capacity of the power grid. By performing time-by-time composite calculations on both sides, the overall supply and demand status of the power grid at each time point is obtained.

[0191] Power grid physical constraint correction: Based on the preliminary supply and demand calculation results, power grid topology constraints and equipment capacity constraints are introduced for correction, including: line power flow transmission capacity limits, node power balance constraints, upper limit constraints on transformer and key equipment capacity, and local voltage and frequency stability constraints. The supply and demand results are corrected through the above constraints to avoid physically infeasible supply and demand differences.

[0192] Multi-scale difference generation mechanism: After completing the constraint correction, the supply and demand difference is decomposed according to the time scale to form short-term supply and demand difference (for rapid fluctuation adjustment), intraday supply and demand difference (for peak and valley adjustment optimization), and long-term supply and demand difference (for operation strategy planning). The differences at each scale maintain independent analysis capabilities while maintaining continuous connection in the time series.

[0193] Centered on the time axis, a curve integration operator is used. The short-term, intraday, and long-term imbalance levels are seamlessly stitched together in chronological order to form a continuous time-series level sequence. Simultaneously, historical grid state data (S), grid topology data (T), and grid equipment data (D) are used to impose full-domain constraints on the grid, ultimately outputting a multi-scale imbalance level prediction curve. Its expression is as follows:

[0194] ;

[0195] The multi-scale imbalance level prediction curve is decomposed into imbalance level sequences of corresponding scales according to real-time, intraday, and seasonal scales. These sequences are then input into the imbalance optimization and regulation unit. The core optimization objective is to minimize the imbalance level at each scale. Combining the scale constraints of the grid side and the energy storage side, the built-in particle swarm optimization algorithm is used to solve the regulation parameters at each scale. The solution process at each scale is independent of each other and matches the grid regulation demand of the corresponding time period.

[0196] Real-time parameter adjustment Solve for real-time imbalance levels With the goal of combining real-time power grid status data Real-time operating limits of power grid topology data, power grid equipment data, and energy storage network control data. The real-time adjustment threshold is determined, and the adjustment parameters that adapt to short-term imbalances of 0-1 hours are calculated, including the real-time output correction of conventional power sources in the grid, the real-time charging and discharging power adjustment of energy storage, and the optimization parameters of power flow distribution in the grid.

[0197] Intraday adjustment parameters Solve for intraday imbalance levels With the goal of combining intraday power grid status data Continuous daily operation constraints of power grid equipment and rated performance of energy storage Given the intraday charging and discharging capacity limitations, we solve for the adjustment parameters to adapt to intraday imbalances from 1 to 24 hours, including the planned output adjustment values ​​of the power grid for each period, the intraday charging and discharging sequence optimization parameters of energy storage, and the load-side demand response adjustment coefficient.

[0198] Seasonal adjustment parameters Solve by seasonal imbalance level With the goal of combining seasonal power grid status data Considering the long-term transmission capacity of the power grid topology, the seasonal operating characteristics of equipment, and the long-term cycle life constraints of energy storage, we solve for the adjustment parameters that can adapt to long-term imbalances ranging from 1 day to 12 months, including the optimized value of clean energy grid-connected capacity, the power grid cross-seasonal power dispatch parameters, and the adjustment amount of energy storage seasonal charging and discharging strategies.

[0199] The imbalance optimization adjustment unit calculates the real-time adjustment parameters obtained from the solution. Intraday adjustment parameters Seasonal adjustment parameters The data is integrated and standardized to generate hierarchical and executable multi-scale adjustment commands in accordance with power grid dispatching specifications.

[0200] Furthermore, the energy storage regulation prediction model and the supply-demand balance regulation model are constrained by the energy storage device data;

[0201] The energy storage device data includes rated energy storage performance data, historical energy storage operating status data, and energy storage network control data.

[0202] The rated performance data of the energy storage includes rated capacity, charge and discharge power, conversion efficiency, upper and lower limits of SOC, response time, grid-connected rated voltage, and grid-connected rated frequency;

[0203] The historical operating status data of the energy storage includes historical SOC changes of the energy storage device, actual charging and discharging power and capacity, start-stop records, fault data, grid connection switching records and off-grid switching records;

[0204] The energy storage network control data includes virtual inertia, voltage and frequency support thresholds, charging and discharging triggering conditions, multi-energy storage collaborative control strategies, and scheduling communication parameters.

[0205] In this embodiment, the rated performance data of energy storage includes the factory calibration parameters of energy storage devices, such as rated capacity, charging and discharging power, conversion efficiency, SOC upper and lower limits, response time, grid-connected rated voltage, and grid-connected rated frequency. The data is collected and clustered according to device type, invalid calibration values ​​are removed, and the grid topology nodes and voltage levels of the devices are associated with the devices. This provides a basic basis for the adaptation of energy storage type adaptation units, determines the charging and discharging performance boundaries of energy storage devices, and avoids operation beyond rated conditions.

[0206] Historical operational status data for energy storage includes historical SOC changes of energy storage devices, actual charging and discharging power and capacity, start-up and shutdown records, fault data, grid connection switching records, and off-grid switching records. Data is collected sequentially along a time dimension consistent with grid data, correlated with grid load and clean energy output data during the same period, completes missing data, removes abnormal data under fault conditions, and splits multi-scale operational data. This provides operational characteristic support for dynamic adjustment strategy units, analyzes energy storage charging and discharging efficiency and SOC response patterns, and provides historical references for multi-scale charging and discharging quantity calculations.

[0207] The energy storage grid-based control data includes grid-based control parameters such as the virtual inertia of energy storage devices and voltage / frequency support thresholds, as well as charging and discharging triggering conditions, multi-energy storage collaborative control strategies, and dispatch communication parameters. Control parameters and strategy files are collected from the energy storage EMS system and the power grid dispatch system. The control parameters of different manufacturers and types of equipment are standardized, and the units and judgment rules are unified. This provides a basis for the control logic matching of the energy storage regulation and prediction model, ensuring that the charging and discharging prediction curve is consistent with the actual grid-based control capability of energy storage, and realizing precise regulation of power grid supply and demand.

[0208] The constraints imposed by the energy storage device data are manifested as "prediction constraints" and "decision constraints," as detailed below:

[0209] I. Prediction constraint logic for the energy storage regulation prediction model:

[0210] The core objective of the energy storage regulation prediction model is to generate predicted energy storage charge and discharge curves. Its constraints, primarily based on energy storage device data, are reflected in the following three aspects:

[0211] Rated performance constraints (hard boundary constraints): Rated performance data of energy storage directly constitute the physical boundary conditions of the model, including: upper limit of charging and discharging power, upper and lower limits of SOC, conversion efficiency range, and response time constraints. During the model prediction process, all charging and discharging prediction results must first be filtered through this boundary. Any excess will be automatically reduced or redistributed to ensure that the prediction results do not exceed the physical capabilities of the equipment.

[0212] Historical operating status constraints (behavioral consistency constraints): Historical operating status data of energy storage is used to constrain the model output to conform to the actual operating behavior characteristics of the equipment, including: the inertial law of SOC change, the charging and discharging switching frequency limit, the actual response lag characteristics and historical fault and unavailability intervals; when generating prediction results, the model will perform historical consistency verification on the predicted behavior to avoid abnormal charging and discharging behaviors that are obviously inconsistent with the historical operating mode.

[0213] Network-based control constraints (control logic constraints): Network-based control data for energy storage is used to constrain energy storage to comply with grid dispatch and control logic, including: voltage / frequency support priority, virtual inertia response rules, charge and discharge triggering conditions, and multi-energy storage collaborative control strategies. This constraint ensures that energy storage charge and discharge prediction is not only "feasible" but also "compliant with control strategy requirements," avoiding conflicts between prediction results and actual dispatch logic.

[0214] Energy storage device data takes effect in the prediction model through physical upper limit constraints on charging and discharging results, historical pattern constraints on predicted behavior, and control logic constraints on output strategies, thereby ensuring that the energy storage charging and discharging prediction curve is always within the range of "executable, schedulable, and safe operation".

[0215] II. Decision Constraint Logic for the Supply-Demand Balance Adjustment Model:

[0216] The core objective of the supply-demand balance adjustment model is to generate adjustment commands. Its constraints, primarily based on energy storage device data, are reflected in "adjustment feasibility constraints" and "execution safety constraints," as detailed below:

[0217] Regulation capability constraints (executability constraints): Rated performance data of energy storage is used to limit the executable range of regulation commands, including: maximum charge and discharge capacity per unit time, sustainable regulation duration, and response speed limit; all regulation commands must match the actual regulation capability of energy storage during the generation stage, otherwise they will be automatically downgraded or executed in segments.

[0218] Operational status constraints (real-time safety constraints): Historical operational status data of energy storage is used to constrain adjustment behavior to be adapted to the current operational status, including: whether the current SOC allows charging or discharging, whether it is currently in a fault or power-limited state, and whether there is a risk of frequent switching. This constraint is used to prevent adjustment commands from causing abnormal operation of energy storage at the execution level.

[0219] Control strategy constraints (dispatch consistency constraints): Control data for energy storage grid-based systems is used to constrain regulation commands to comply with grid control strategies, including: grid frequency priority support strategy, voltage support priority rules, multi-energy storage collaborative allocation mechanism, and dispatch communication and execution timing requirements; ensuring that regulation commands are consistent with the grid dispatch system and avoiding strategy conflicts.

[0220] In the regulation model, data from energy storage devices is used to limit the maximum execution boundary of regulation commands, correct regulation behaviors that do not conform to the current operating state, and align grid control strategies with energy storage response logic, thereby ensuring that regulation commands are executable, safe, and coordinated.

[0221] The constraint effect of energy storage device data in both models is uniformly reflected in a "three-layer closed-loop constraint system". Physical boundary constraints (rated performance) prevent overcapacity and overpower operation, behavioral consistency constraints (historical state) prevent abnormal prediction and abnormal adjustment, and control strategy constraints (grid control) ensure consistency with grid dispatch logic.

[0222] The adjustment parameters at various scales are classified according to the adjustment objects of power grid equipment, energy storage devices, and load side, and the dimensions, command formats, and execution accuracy of the parameters are standardized to ensure that the adjustment parameters can be directly identified by the power grid dispatch system and the energy storage management system.

[0223] Based on the principles of real-time priority, intraday coordination, and seasonal planning, adjustment instructions are divided into real-time execution instructions, intraday scheduling instructions, and seasonal planning instructions. Real-time instructions are executed immediately at the minute level, intraday instructions are executed at the hour level, and seasonal instructions are planned in advance at the daily level.

[0224] The integrated multi-scale adjustment instructions are generated into standardized scheduling files, which are then precisely issued according to grid topology nodes, energy storage access locations, and load zones. At the same time, the instructions include adjustment boundaries and execution feedback requirements for each scale to ensure that the adjustment actions are implemented and to achieve precise optimization of the grid's multi-scale supply and demand imbalance.

[0225] Example 2:

[0226] See Figure 2 A multi-scale power grid supply and demand balance regulation system, which is applied to the method described in Example 1, the system comprising:

[0227] Data acquisition module 1 is used to acquire historical clean energy data, historical comprehensive load data, and historical environmental data of the target power grid;

[0228] The multi-scale clean energy output prediction curve acquisition module 2 is used to input historical clean energy data and environmental historical data into the multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain the multi-scale clean energy output prediction curve.

[0229] The comprehensive load forecasting acquisition module 3 is used to input historical comprehensive load data into the power grid load simulation model to simulate dynamic load changes and obtain a comprehensive load forecasting curve.

[0230] The multi-scale energy storage charge and discharge prediction curve acquisition module 4 is used to input the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to perform energy storage charge and discharge prediction and obtain the multi-scale energy storage charge and discharge prediction curve.

[0231] The output regulation module 5 is used to input the multi-scale clean energy output prediction curve, the comprehensive load prediction curve and the multi-scale energy storage charging and discharging prediction curve into the supply and demand balance regulation model for supply and demand regulation analysis, generate regulation commands, and regulate the output of the target power grid based on the regulation commands.

[0232] Furthermore, for the specific implementation steps of the above modules, please refer to the corresponding description in Example 1, which will not be repeated here.

[0233] Example 3:

[0234] See Figure 3 A multi-scale power grid supply and demand balance regulation device, the device including a processor 6 and a memory 7;

[0235] The memory 7 is used to store computer program code 71 and to transmit the computer program code 71 to the processor 6;

[0236] The processor 6 is used to execute the multi-scale power grid supply and demand balance regulation method described in Embodiment 1 according to the instructions in the computer program code 71.

[0237] This embodiment also includes a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are executed on a computer, the multi-scale power grid supply and demand balance regulation method described in Embodiment 1 is implemented.

[0238] Generally, computer instructions for implementing the method of the present invention can be carried on any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media can include any computer-readable medium except for the signal itself which is temporarily propagating.

[0239] Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0240] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. In particular, Python, suitable for neural network computation, and platform frameworks based on TensorFlow, PyTorch, etc., can be used. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet using an Internet service provider).

[0241] The aforementioned equipment and non-transitory computer-readable storage media can be found in the detailed description of a multi-scale power grid supply and demand balance regulation method and its beneficial effects, which will not be repeated here.

[0242] Although embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A multi-scale power grid supply and demand balance regulation method, characterized in that, include: Acquire historical clean energy data, comprehensive load data, and environmental data of the target power grid; Historical clean energy data and environmental data are input into a multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain multi-scale clean energy output prediction curves. The historical data of the comprehensive load is input into the power grid load simulation model to simulate the dynamic changes of the load and obtain the comprehensive load prediction curve. The multi-scale clean energy output prediction curve and the comprehensive load prediction curve are input into the energy storage regulation prediction model to predict the energy storage charge and discharge, and the multi-scale energy storage charge and discharge prediction curve is obtained. The multi-scale clean energy output forecast curve, the comprehensive load forecast curve, and the multi-scale energy storage charging and discharging forecast curve are input into the supply and demand balance regulation model for supply and demand regulation analysis, generating regulation commands, and regulating the output of the target power grid based on the regulation commands.

2. The multi-scale power grid supply and demand balance regulation method according to claim 1, characterized in that: The process of inputting historical clean energy data and historical environmental data into a multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain a multi-scale clean energy output prediction curve specifically includes: The multi-scale clean energy fluctuation simulation model includes micro-scale units, meso-scale units, and macro-scale units. Acquire real-time output data, daily output data, and seasonal output data from historical clean energy data, as well as real-time sunlight, real-time wind speed, daily sunlight, daily wind speed, seasonal sunlight, and seasonal wind speed from historical environmental data. Real-time power output data, real-time illumination, and real-time wind speed are input into microscale units, and short-term power output fluctuations are simulated using the random forest algorithm to obtain short-term power output prediction curves. Intraday power output data, intraday sunshine and intraday wind speed are input into the mesoscale unit, and intraday power output fluctuations are simulated through time series prediction algorithms to obtain intraday power output prediction curves. Seasonal power output data, seasonal sunlight, and seasonal wind speed are input into macro-scale units, and seasonal power output fluctuations are simulated through trend analysis algorithms to obtain long-term power output prediction curves. The short-term output forecast curve, intraday output forecast curve, and long-term output forecast curve are combined to form a multi-scale clean energy output forecast curve.

3. The multi-scale power grid supply and demand balance regulation method according to claim 1, characterized in that: The step of inputting historical comprehensive load data into a power grid load simulation model to simulate dynamic load changes and obtain a comprehensive load forecast curve specifically includes: The power grid load simulation model includes a load classification unit and a load prediction unit; The comprehensive historical load data is input into the load classification unit for classification, and residential load historical data, industrial load historical data, commercial load historical data and emerging load historical data are obtained. Historical data on residential load, industrial load, commercial load, and emerging load are input into the load forecasting unit. Load forecasting is performed through time-series forecasting and scenario adjustment algorithms to obtain residential load forecasting curves, industrial load forecasting curves, commercial load forecasting curves, and emerging load forecasting curves. The residential load forecast curve, industrial load forecast curve, commercial load forecast curve, and emerging load forecast curve are combined to form a comprehensive load forecast curve.

4. The multi-scale power grid supply and demand balance regulation method according to claim 1, characterized in that: The multi-scale clean energy fluctuation simulation model and the power grid load simulation model are constrained by historical power grid operation data. The historical power grid operation data includes historical power grid status data, power grid topology data, and power grid equipment data; The historical power grid status data includes node voltage, line current, active power, reactive power, system frequency, power factor, line loss rate, and power fluctuation at clean energy grid connection points. The power grid topology data includes power grid wiring methods, node and branch connection relationships, line impedance, line admittance, substation main wiring layout, and topology change records. The power grid equipment data includes rated capacity, operating efficiency, and current carrying capacity.

5. The multi-scale power grid supply and demand balance regulation method according to claim 1, characterized in that: The process of inputting the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to predict energy storage charging and discharging, and obtaining the multi-scale energy storage charging and discharging prediction curve, specifically includes: The energy storage regulation prediction model includes an energy storage type adaptation unit and a dynamic regulation strategy unit. The multi-scale clean energy output prediction curve and the comprehensive load prediction curve are input into the energy storage type adaptation unit to adapt the operating conditions of different types of energy storage devices in the target power grid and obtain the adapted multi-scale output prediction curve and the adapted load prediction curve. The adaptive multi-scale output prediction curve and the adaptive load prediction curve are input into the dynamic adjustment strategy unit. The charging and discharging quantities are calculated based on the fluctuation characteristics of the adaptive multi-scale output prediction curve and the change characteristics of the adaptive load prediction curve to obtain the multi-scale energy storage charging and discharging prediction curve.

6. The multi-scale power grid supply and demand balance regulation method according to claim 1, characterized in that: The process of inputting multi-scale clean energy output prediction curves, comprehensive load prediction curves, and multi-scale energy storage charging and discharging prediction curves into the supply and demand balance regulation model for supply and demand regulation analysis and generating regulation commands specifically includes: The supply and demand balance adjustment model includes a balance calculation unit, a supply and demand trend prediction unit, and an imbalance optimization adjustment unit. The multi-scale clean energy output prediction curve, the comprehensive load prediction curve, and the multi-scale energy storage charging and discharging prediction curve are input into the balance calculation unit to calculate the supply-demand difference and obtain the multi-scale balance difference prediction curve. The multi-scale balance difference prediction curve is input into the supply and demand trend prediction unit, and the imbalance level is predicted by the time series trend analysis algorithm to obtain the multi-scale imbalance level prediction curve. The multi-scale imbalance level prediction curve is input into the imbalance optimization adjustment unit to solve the adjustment parameters, obtain the real-time adjustment parameters, intraday adjustment parameters and seasonal adjustment parameters, and generate adjustment instructions based on the real-time adjustment parameters, intraday adjustment parameters and seasonal adjustment parameters.

7. The multi-scale power grid supply and demand balance regulation method according to claim 1, characterized in that: The energy storage regulation prediction model and the supply-demand balance regulation model are constrained by the energy storage device data. The energy storage device data includes rated energy storage performance data, historical energy storage operating status data, and energy storage network control data. The rated performance data of the energy storage includes rated capacity, charge and discharge power, conversion efficiency, upper and lower limits of SOC, response time, grid-connected rated voltage, and grid-connected rated frequency; The historical operating status data of the energy storage includes historical SOC changes of the energy storage device, actual charging and discharging power and capacity, start-stop records, fault data, grid connection switching records and off-grid switching records; The energy storage network control data includes virtual inertia, voltage and frequency support thresholds, charging and discharging triggering conditions, multi-energy storage collaborative control strategies, and scheduling communication parameters.

8. A multi-scale power grid supply and demand balance regulation system, characterized in that, The system is applied to the method according to any one of claims 1-7, the system comprising: The data acquisition module (1) is used to acquire historical clean energy data, historical comprehensive load data and historical environmental data of the target power grid; The multi-scale clean energy output prediction curve acquisition module (2) is used to input historical clean energy data and environmental historical data into the multi-scale clean energy fluctuation simulation model to perform multi-scale fluctuation simulation and obtain the multi-scale clean energy output prediction curve. The comprehensive load forecasting acquisition module (3) is used to input the comprehensive load historical data into the power grid load simulation model to simulate the dynamic changes of load and obtain the comprehensive load forecasting curve. The multi-scale energy storage charge and discharge prediction curve acquisition module (4) is used to input the multi-scale clean energy output prediction curve and the comprehensive load prediction curve into the energy storage regulation prediction model to perform energy storage charge and discharge prediction and obtain the multi-scale energy storage charge and discharge prediction curve. The output regulation module (5) is used to input the multi-scale clean energy output prediction curve, the comprehensive load prediction curve and the multi-scale energy storage charging and discharging prediction curve into the supply and demand balance regulation model for supply and demand regulation analysis, generate regulation commands, and regulate the output of the target power grid based on the regulation commands.

9. A multi-scale power grid supply and demand balancing regulation device, characterized in that: The device includes a processor (6) and a memory (7); The memory (7) is used to store computer program code (71) and to transmit the computer program code (71) to the processor (6). The processor (6) is used to execute the multi-scale power grid supply and demand balance regulation method according to any one of claims 1-7 according to the instructions in the computer program code (71).

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed on a computer, implement the multi-scale power grid supply and demand balance regulation method as described in any one of claims 1-7.

Images