Coordination control method and device based on multiple time scales, medium and product

Abstract

The embodiment of the invention provides a coordination control method and device based on multiple time scales, a medium and a product, and is applied to a multi-source heterogeneous energy system, and the method comprises the steps: obtaining multi-time scale prediction data corresponding to the multi-source heterogeneous energy system; the multi-time scale prediction data comprises prediction data of the multi-source heterogeneous energy system under multiple different time scales, and each prediction data is generated through different prediction models; carrying out feature fusion on the multiple prediction data to obtain a prediction feature vector; and in combination with a preset hierarchical objective function, obtaining a cross-time-scale control strategy based on the predicted feature vector. The method is used for solving the problems of asynchronous response and cross-scale control conflict of multi-source equipment in a network-forming type micro-grid, so that the operation stability, the economical efficiency and the cooperative control precision of a system under complex working conditions are improved.

Description

Technical Field

[0001] This application relates to the field of power system control technology, and in particular to methods, devices, media and products based on multi-timescale coordinated control. Background Technology

[0002] With the widespread application of distributed energy resources, grid-connected photovoltaic-storage-charging-thermal microgrids have become a key system for improving energy utilization efficiency. However, this system involves heterogeneous devices from multiple sources, such as photovoltaics, energy storage, charging piles, and heat pumps, and faces challenges such as large output fluctuations, high load uncertainty, and significant differences in equipment response delays. In actual operation, it is necessary to simultaneously meet the control requirements of multiple time scales, including long-term energy optimization, medium-term power smoothing, and short-term voltage and frequency stability. Traditional single control modes are difficult to balance economy and stability.

[0003] Existing technical solutions often employ a hierarchical control architecture, which separates long-term, medium-term, and short-term controls, resulting in a lack of coordination among objectives at different time scales. For example, long-term energy storage planning does not consider short-term power fluctuation constraints, which can easily lead to conflicts in control strategies; prediction models are usually built for a single time scale and lack cross-scale information fusion, which limits prediction accuracy.

[0004] Furthermore, existing technologies largely rely on centralized cloud platforms for prediction and control, which suffers from high communication latency and insufficient real-time performance, making it difficult to meet the microgrid's requirement for second-level response. Therefore, a coordinated control method is urgently needed to address these technical bottlenecks. Summary of the Invention

[0005] This application provides a multi-timescale coordinated control method, device, medium, and product to solve the problems of asynchronous response of multi-source devices and cross-scale control conflicts in grid-type microgrids, so as to improve the system's operational stability, economy, and coordinated control accuracy under complex operating conditions.

[0006] In a first aspect, embodiments of this application provide a multi-timescale coordinated control method applied to a multi-source heterogeneous energy system; the method includes:

[0007] Acquire multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different timescales, and each prediction data is generated by a different prediction model.

[0008] The prediction data from multiple sources are fused to obtain a prediction feature vector.

[0009] Combining a preset hierarchical objective function, a control strategy across time scales is obtained based on the predicted feature vector. The hierarchical objective function includes at least an energy storage state optimization objective corresponding to the first time scale, a power fluctuation smoothing objective corresponding to the second time scale, and a voltage frequency stability objective corresponding to the third time scale; and the time spans corresponding to the first time scale, the second time scale, and the third time scale decrease sequentially.

[0010] The control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

[0011] In one possible implementation, the step of combining a preset hierarchical objective function with the predicted feature vector to obtain a control strategy across time scales includes:

[0012] Based on the predicted feature vector and the energy storage state optimization objective, the first sub-strategy is obtained;

[0013] Based on the predicted feature vector and the power fluctuation smoothing objective, the second sub-strategy is obtained;

[0014] Based on the predicted feature vector and the voltage frequency stability target, the third sub-strategy is obtained.

[0015] In one possible implementation, the multi-source heterogeneous energy system includes at least one energy production end and at least one energy consumption end;

[0016] After obtaining the control strategy across time scales, the method further includes:

[0017] Based on the delay characteristics of the energy production end and the energy consumption end, a distribution advance time is set for each of the energy production ends and each of the energy consumption ends;

[0018] According to the priority order preset by the first sub-strategy, the second sub-strategy, and the third sub-strategy, and according to the distribution advance time corresponding to each energy production end and each energy consumption end, the first sub-strategy, the second sub-strategy, and the third sub-strategy are distributed in sequence, so that the equipment in the multi-source heterogeneous energy system can adjust its state according to the first sub-strategy, the second sub-strategy, or the third sub-strategy.

[0019] In one possible implementation, the feature fusion of multiple predicted data to obtain a predicted feature vector includes:

[0020] Based on the changing trends corresponding to the various predicted data, weights are dynamically assigned to the time scales corresponding to the various predicted data.

[0021] By combining the weights corresponding to each time scale, feature fusion is performed on the various prediction data to obtain the prediction feature vector.

[0022] In one possible implementation, the step of dynamically assigning weights to the time scales corresponding to the various predicted data based on the changing trends of the various predicted data includes:

[0023] Based on the forecast data corresponding to multiple time scales, the correlation between forecast data at different time scales is calculated.

[0024] Based on the magnitude of the correlation, corresponding weights are assigned to the predicted data corresponding to different time scales; wherein, in descending order of the time span, the weights of the predicted data corresponding to different time scales change from positively correlated to negatively correlated with the magnitude of the correlation.

[0025] In one possible implementation, the method further includes:

[0026] Obtain environmental data of the environment in which the multi-source heterogeneous energy system is located, as well as the operational data corresponding to at least one energy production end and at least one energy consumption end in the multi-source heterogeneous energy system;

[0027] Based on the environmental data and the operational data, the multi-timescale prediction data and the control strategy are revised.

[0028] Secondly, embodiments of this application provide a multi-timescale coordinated control device applied to a multi-source heterogeneous energy system, the device comprising:

[0029] The first acquisition module is used to acquire multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different timescales, and each prediction data is generated by a different prediction model.

[0030] The fusion module is used to perform feature fusion on multiple prediction data to obtain a prediction feature vector;

[0031] The second acquisition module is used to combine a preset hierarchical objective function with the predicted feature vector to acquire a cross-timescale control strategy. The hierarchical objective function includes at least an energy storage state optimization objective corresponding to the first timescale, a power fluctuation smoothing objective corresponding to the second timescale, and a voltage frequency stability objective corresponding to the third timescale. The time spans corresponding to the first timescale, the second timescale, and the third timescale decrease sequentially. The control strategy includes at least a first sub-strategy corresponding to the first timescale, a second sub-strategy corresponding to the second timescale, and a third sub-strategy corresponding to the third timescale.

[0032] In one possible implementation, the second acquisition module is further configured to acquire the first sub-strategy based on the predicted feature vector and the energy storage state optimization objective;

[0033] The second acquisition module is further configured to acquire the second sub-strategy based on the predicted feature vector and the power fluctuation smoothing target;

[0034] The second acquisition module is further configured to acquire the third sub-strategy based on the predicted feature vector and the voltage frequency stability target.

[0035] In one possible implementation, the multi-source heterogeneous energy system includes at least one energy production end and at least one energy consumption end, and the device further includes: a setting module and a distribution module;

[0036] The setting module is used to set the distribution advance time corresponding to each of the energy production terminals and each of the energy consumption terminals based on the delay characteristics corresponding to the energy production terminals and the energy consumption terminals;

[0037] The distribution module is used to distribute the first sub-strategy, the second sub-strategy, and the third sub-strategy in a pre-set priority order according to the distribution advance time corresponding to each energy production end and each energy consumption end, so that the devices in the multi-source heterogeneous energy system can adjust their state according to the first sub-strategy, the second sub-strategy, or the third sub-strategy.

[0038] In one possible implementation, the fusion module is further configured to dynamically assign weights to the time scales corresponding to the multiple predicted data based on the changing trends corresponding to the multiple predicted data.

[0039] The fusion module is also used to combine the weights corresponding to each time scale to perform feature fusion on the various prediction data to obtain the prediction feature vector.

[0040] In one possible implementation, the fusion module is also used to calculate the correlation between prediction data at different time scales based on prediction data corresponding to multiple time scales.

[0041] The fusion module is also used to assign corresponding weights to the prediction data corresponding to different time scales based on the magnitude of the correlation; wherein, in order of decreasing time span, the weight of the prediction data corresponding to different time scales changes from positive to negative correlation with the magnitude of the correlation.

[0042] In one possible implementation, the second acquisition module is further configured to acquire environmental data of the environment in which the multi-source heterogeneous energy system is located, as well as operational data corresponding to at least one energy production end and at least one energy consumption end in the multi-source heterogeneous energy system.

[0043] The second acquisition module is further configured to correct the multi-timescale prediction data and the control strategy based on the environmental data and the operational data.

[0044] Thirdly, embodiments of this application provide a multi-source heterogeneous energy system, which includes at least one energy production end and at least one energy consumption end;

[0045] The multi-source heterogeneous energy system is used to acquire multi-time-scale prediction data corresponding to the multi-source heterogeneous energy system by employing the multi-time-scale coordinated control method based on the first aspect and / or any of the possible implementations of the first aspect as described above. The multi-time-scale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different time scales, and each prediction data is generated by a different prediction model. Feature fusion is performed on the multiple prediction data to obtain a prediction feature vector. Based on the prediction feature vector, a cross-time-scale control strategy is obtained by combining a preset hierarchical objective function. The hierarchical objective function includes at least an energy storage state optimization objective corresponding to the first time scale, a power fluctuation smoothing objective corresponding to the second time scale, and a voltage frequency stability objective corresponding to the third time scale. The time spans corresponding to the first time scale, the second time scale, and the third time scale decrease sequentially.

[0046] The control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

[0047] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0048] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0049] The multi-timescale coordinated control method, device, medium, and product provided in this application are applied to multi-source heterogeneous energy systems. The method includes: acquiring multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different timescales, and each prediction data is generated through different prediction models; performing feature fusion on multiple prediction data to obtain a prediction feature vector; and combining a preset hierarchical objective function with the prediction feature vector to obtain a cross-timescale control strategy to solve the problems of asynchronous response of multi-source devices and cross-scale control conflicts in grid-type microgrids, so as to improve the system's operational stability, economy, and coordinated control accuracy under complex operating conditions. Attached Figure Description

[0050] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0051] Figure 1 A schematic diagram of a scenario for a multi-timescale coordinated control method provided in this application;

[0052] Figure 2 A flowchart illustrating a multi-timescale coordinated control method provided in this application. Figure 1 ;

[0053] Figure 3 A flowchart illustrating a multi-timescale coordinated control method provided in this application. Figure 2 ;

[0054] Figure 4 A schematic diagram of a multi-timescale coordinated control device provided in this application;

[0055] Figure 5 A schematic diagram of the structure of the electronic device provided in this application.

[0056] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0057] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0058] This application provides a multi-timescale-based coordinated control method, which can be applied to, for example... Figure 1 The application environment is shown. The system can be divided into a monitoring layer, a data acquisition layer, a control center, and an energy utilities channel layer. These are connected via DC, AC, and communication lines to achieve device interconnection and collaborative control. Local weather stations and smart meters constitute the data acquisition layer, collecting real-time environmental and electricity consumption data to provide a basis for system decision-making.

[0059] Specifically, the monitoring layer can be a cloud platform / web / app monitoring platform, providing functions such as data visualization, alarm management, and remote control, allowing users to monitor system dynamics anytime, anywhere. The data acquisition layer can include, for example, a local weather station responsible for collecting local meteorological data such as light intensity, temperature, and wind speed. This data is crucial for predicting photovoltaic power generation and optimizing system operation strategies. It can also include smart meters, which can be installed at various key nodes of the microgrid to collect electrical data in real time, such as voltage, current, power, and electricity consumption, providing fundamental information for the system's energy management and optimized scheduling.

[0060] The control center can be implemented through an Artificial Intelligence (AI) microgrid coordinator, which receives various types of data from the data acquisition layer and performs data analysis and decision-making based on AI algorithms. It is responsible for coordinating and controlling the energy flow between the energy supply side, energy storage system, charging load, and thermal load to achieve stable system operation and optimized scheduling. Simultaneously, it communicates and interacts with other components through the energy computer's common channel layer.

[0061] On the energy supply side, photovoltaic arrays can output electrical energy via photovoltaic inverters to connect and interact with the microgrid's AC system. The energy storage system uses storage cabinets and controllers to regulate charging and discharging. The storage cabinets store excess electrical energy, using battery packs (such as lithium-ion or lead-acid batteries) as the storage medium. When sunlight is abundant, the cabinets store surplus energy generated by photovoltaic power generation; when sunlight is insufficient or load demand is high, the stored energy is released to meet the load. The energy storage controller is responsible for controlling the charging and discharging process of the storage cabinets, precisely adjusting the charging and discharging power and status of the cabinets according to the system's operating strategy and real-time needs to ensure the safe and efficient operation of the energy storage system.

[0062] Correspondingly, the charging load can be, for example, a charging pile array and electric vehicles, while the thermal load can include heat pumps, building air conditioning, etc.

[0063] Understandably, this method can, for example, collect meteorological and electrical data in real time through local weather stations and smart meters, and transmit it to the AI ​​microgrid coordinating controller. The controller, combining the operational goals and strategies set by the cloud platform / Web / App monitoring platform with real-time meteorological and load data, predicts photovoltaic power generation and formulates charging and discharging plans for the energy storage system, charging strategies for electric vehicles, and heating schemes for thermal loads. By coordinating and controlling equipment such as photovoltaic inverters, energy storage controllers, charging piles, and heat pumps, the energy balance and optimized operation of the photovoltaic-storage-charging-heating microgrid are achieved, improving energy utilization efficiency, reducing operating costs, and ensuring system stability and reliability.

[0064] For example, after receiving various types of data, the AI ​​microgrid coordinating controller performs intelligent analysis and decision-making. When the photovoltaic power generation exceeds the current load demand, the controller sends a command to the energy storage controller. The energy storage controller then controls the energy storage cabinet to enter charging mode, safely and efficiently storing the excess electrical energy. When it is cloudy or at night, and the photovoltaic power generation is insufficient, the energy storage controller, according to the controller's command, releases the stored electrical energy from the energy storage cabinet, supplying power to the load through the energy computer's common channel layer, ensuring the stable operation of the system.

[0065] In one embodiment, a multi-timescale-based coordinated control method is provided. This embodiment illustrates the application of this multi-timescale-based coordinated control method to a multi-source heterogeneous energy system. It is understood that the multi-source heterogeneous energy system includes at least one energy production end and at least one energy consumption end. The energy production end equipment may include, for example, traditional energy power generation equipment, renewable energy power generation equipment, energy storage and replenishment equipment, external power grid interfaces, etc. The energy consumption end equipment may include, for example, industrial equipment, commercial building equipment, and residential equipment, etc.

[0066] Figure 2 A flowchart illustrating a multi-timescale coordinated control method provided in this application. Figure 1 .like Figure 2 As shown, the multi-time-scale coordinated control method provided in this embodiment is applied to a multi-source heterogeneous energy system. This multi-time-scale coordinated control method includes:

[0067] S201. Obtain multi-timescale prediction data corresponding to multi-source heterogeneous energy systems.

[0068] The multi-timescale forecast data includes forecasts of multi-source heterogeneous energy systems at various time scales, each generated using different forecasting models. The forecast data can be used to refer to the predicted future states of each production and consumption end within the multi-source heterogeneous energy system across multiple time spans. Specifically, the forecast data may include, but is not limited to, forecasts of power generation at the energy production end, load demand at the energy consumption end, and forecasts of the impact of environmental factors (such as solar irradiance, wind speed, and temperature) on the energy system.

[0069] Specifically, multi-timescale forecast data can be generated using different forecasting models, each optimized for a specific timescale to ensure the accuracy and applicability of the forecast results. For example, for short-term forecasts (such as the next few minutes to hours), forecasting models based on real-time data and machine learning algorithms can be used to capture rapid changes in the system state; for medium-term forecasts (such as the next few hours to days), time series analysis or statistical methods can be used, taking into account historical data and periodic patterns; and for long-term forecasts (such as the next few days to weeks or longer), macroeconomic factors such as weather forecasts and energy market trends can be combined to make more accurate predictions.

[0070] Optionally, raw data can be collected in real time through the data acquisition layer (weather station, smart meter, equipment status sensor, etc.) corresponding to the multi-source heterogeneous energy system, and transmitted to the edge AI computing unit of the control center. After parallel inference by each prediction model, structured prediction data at each time scale is output, providing a foundation for subsequent feature fusion steps.

[0071] S202. Perform feature fusion on multiple prediction data to obtain the prediction feature vector.

[0072] Among them, the prediction feature vector is a comprehensive data representation obtained by fusing the features of prediction data of multi-source heterogeneous energy systems at multiple time scales.

[0073] Understandably, due to the "information silos" and "dimensional mismatch" problems existing between data from different time scales, physical dimensions, and sources, directly using these raw data for control strategy formulation may be affected by the differences between the data, impacting the control effect. Therefore, it is possible to effectively integrate and extract features from the acquired prediction data of multi-source heterogeneous energy systems at various time scales.

[0074] Specifically, the forecast data at different time scales can first be unified and normalized on a time axis. For example, using the smallest time unit of the control period (such as seconds or milliseconds) as a benchmark, the trend data of long time scales (such as daily average temperature and daily electricity price) can be extended to a high-resolution time axis through interpolation or preservation methods. The instantaneous data of short time scales can be mapped to a unified time window through moving average or aggregation to ensure that all data are strictly synchronized on the timestamp. For data with different units (such as power in kW, temperature in ℃, and electricity price in yuan / kWh), the Min-Max Normalization or Z-Score standardization method can be used to map all values ​​to the [0, 1] interval or standard normal distribution to eliminate the influence of unit differences on the weight of subsequent model calculations.

[0075] Subsequently, for the aligned multi-source heterogeneous data, key features reflecting different operating characteristics of the system can be extracted. These can include: time-series dynamic features, using convolutional neural networks (CNN) or long short-term memory networks (LSTM) to extract the volatility, ramp rate, periodicity, and trend and residual terms of photovoltaic and wind power output; coupling and correlation features, extracting correlation features between source, load, and storage, such as "positive correlation between solar intensity and air conditioning load", "negative correlation between peak electricity price and electric vehicle charging demand", and "constraint relationship between energy storage state of charge (SOC) and adjustable capacity"; and environmental constraint features, extracting external boundary condition features such as meteorological conditions (wind speed, irradiance, temperature), equipment operating constraints (maximum output limit, start-stop status), and market signals (real-time electricity price, demand response instructions).

[0076] After identifying the key features, a multimodal fusion mechanism can be used to integrate these features. Specific fusion methods include, but are not limited to, splicing fusion, attention mechanism fusion, and tensor fusion. The fused feature vector can output a predicted feature vector, which may include, for example, specific numerical predictions of photovoltaic, wind power, and load, as well as their uncertainty range, changing trends, mutual coupling relationships, and controllability potential.

[0077] S203. Combining the preset hierarchical objective function, based on the predicted feature vector, obtain the control strategy across time scales.

[0078] The hierarchical objective function includes at least the energy storage state optimization objective corresponding to the first time scale, the power fluctuation smoothing objective corresponding to the second time scale, and the voltage and frequency stability objective corresponding to the third time scale; and the time spans corresponding to the first, second, and third time scales decrease sequentially. The control strategy includes at least the first sub-strategy corresponding to the first time scale, the second sub-strategy corresponding to the second time scale, and the third sub-strategy corresponding to the third time scale.

[0079] Understandably, the first time scale can be a long-term scale, which is the longest time span among the three time scales. It is used to formulate hierarchical targets for optimizing the state of energy storage, and is employed to develop long-term energy plans, optimize equipment operating periods, and manage economic control throughout the entire lifecycle of multi-source heterogeneous energy systems. It forms the basis and provides top-level guidance for medium- and short-term control strategies, with a typical time span of 24-72 hours (1-3 days). This can be flexibly adjusted to longer periods such as 7 days depending on different application scenarios such as industrial parks, commercial complexes, and islands. Long-term scale data mainly consists of trend data, including macroscopic data sampling. Long-term scale controls the basic operating parameters and fixed operating rules of the system's core equipment (such as the energy storage SOC target curve, the fixed operating period of the heat pump, and the long-term power limit of the charging pile).

[0080] Correspondingly, the second time scale is the medium-term scale, an intermediate level bridging the long-term and short-term scales. Its time span lies between the two, and its core objective is to smooth out power fluctuations in a hierarchical manner. Its function is to dynamically adjust equipment operating limits within the basic parameter boundaries established in the long-term scale, based on medium-term trend fluctuation predictions, to smooth out phased fluctuations in system power, thus laying a solid foundation for power balance in short-term grid stability control. The typical time span is 1-4 hours, which can be flexibly fine-tuned according to the operating fluctuation characteristics of multi-source heterogeneous energy systems. The medium-term scale forecast data consists of fluctuating and phased data (such as the decrease / increase in photovoltaic output within 1-4 hours, the batch connection quantity / total power of charging piles in the next 2 hours, and the phased start-up and shutdown demand of heat pumps). The data update frequency is on the order of minutes (e.g., once every 5-15 minutes), higher than the long-term scale but lower than the short-term scale. Meanwhile, the medium-term scale does not change the basic operating rules of the equipment established in the long-term scale (such as the target curve of energy storage SOC and the core operating period of heat pump), but only dynamically adjusts the power limit / adjustment coefficient of the equipment within its parameter boundaries (such as the dynamic limit of energy storage charging and discharging power and the upper limit of the batch access power of charging piles), which belongs to "dynamic optimization within the boundary".

[0081] The third time scale is the short-term scale, the shortest time span among the three time scales. Its core objective is to match the hierarchical target of stable voltage and frequency (highest priority). It is used to address instantaneous operating disturbances in multi-source heterogeneous energy systems, rapidly optimize core grid parameters, and ensure system voltage and frequency stability within the rated range. Its typical time span is 10 seconds to 5 minutes, and can be flexibly fine-tuned based on the real-time response characteristics of system equipment and the instantaneous nature of operating disturbances. The determination of the short-term scale is based on the highest priority of the control target and the instantaneous nature of the time span, combined with data and control characteristics, equipment response matching, and hierarchical support characteristics to form the judgment criteria. This creates a clear gradient in span with the long-term and medium-term scales while accurately matching the instantaneous stability control requirements of the system. Its prediction data consists of instantaneous and real-time high-frequency sampling data (such as millisecond-level fluctuations in photovoltaic output, instantaneous trends in voltage and frequency changes, and instantaneous loads from sudden charging pile connections). The data update frequency is at the millisecond / second level (such as 10kHz, 1 second / time), which is the highest sampling frequency among the three-level scales. It can accurately capture every instantaneous operating condition disturbance. Furthermore, without changing the long-term basic planning or the medium-term power limit boundaries, it only makes rapid, small-scale, and temporary adjustments to the core grid parameters (virtual inertia, damping, virtual impedance) and the emergency operating status of the equipment (instantaneous charging and discharging power of energy storage, power cut-off of charging piles, and emergency shutdown of heat pumps). After the disturbance is eliminated, the equipment parameters automatically return to the set values ​​at the medium and long-term scales.

[0082] Specifically, by using a predicted feature vector that integrates long-, medium-, and short-scale information as the sole input, and substituting it into a pre-defined hierarchical objective function (priority: short-term voltage and frequency stability > medium-term power fluctuation smoothing > long-term energy storage state optimization), a multi-objective optimization algorithm is used to solve the problem and generate cross-timescale control strategies (first / second / third sub-strategies) that correspond one-to-one with the three timescales. Each sub-strategy guides the adjustment of the operating status of various equipment in the multi-source heterogeneous energy system, and the sub-strategies are coordinated without conflict, thus achieving a balance between the system's long-term economic efficiency, medium-term stability, and short-term stability.

[0083] Assuming a multi-source heterogeneous energy system comprises a photovoltaic array, wind turbine generators, a lithium-ion battery energy storage system, and a central air conditioning load, its long-term control involves predicting the next day's photovoltaic power generation and load demand based on weather forecasts and historical load data, and formulating battery charging and discharging plans to avoid insufficient energy storage during nighttime peak load periods. Mid-term control involves real-time monitoring of photovoltaic power generation fluctuations and smoothing power output by adjusting battery charging and discharging rates to reduce the impact on the power grid. Short-term control involves rapidly increasing diesel generator output power or releasing battery energy when the system frequency suddenly drops to maintain frequency stability. Through these hierarchical control strategies, the system can achieve comprehensive optimization of economy, stability, and synergy at different time scales.

[0084] For example, suppose the predicted feature vector contains the following information: Long-term photovoltaic output will be high from 10:00 to 14:00 the following day, and the energy storage SOC target will decrease from 90% to 30%; medium-term photovoltaic output will decrease by 30% at 11:00; short-term, three charging piles will suddenly connect at 11:05. Based on the solution of the hierarchical objective function, the three-level sub-strategies will adjust the equipment parameters in a coordinated manner, which can be expressed as:

[0085] The first sub-strategy sets the energy storage SOC target curve to ≥80% before 11:00 and discharge to 30% from 11:00 to 14:00; the heat pump operates at full power from 10:00 to 14:00.

[0086] The second sub-strategy, at 10:45, increases the limit of energy storage discharge power from 150kW to 200kW, thereby increasing the power smoothing coefficient of photovoltaic inverters.

[0087] The third sub-strategy, at 11:04.5 (0.5 minutes in advance), issued the following instructions: the virtual inertia of the Virtual Synchronous Generator (VSG) was increased from 12 to 18, the Power Converter System (PCS) immediately discharged at 200kW, and the power of the three charging piles that were suddenly connected was reduced to 10kW / unit.

[0088] The control effects achieved above enable the system to smoothly cope with photovoltaic fluctuations and sudden charging pile access, maintain stable voltage and frequency, and ensure that energy storage discharge still conforms to the long-term SOC target curve, thus achieving cross-scale collaborative control.

[0089] In some optional embodiments, the method further includes: acquiring environmental data of the environment in which the multi-source heterogeneous energy system is located, as well as operational data corresponding to at least one energy production end and at least one energy consumption end in the multi-source heterogeneous energy system; and revising multi-timescale prediction data and control strategies based on the environmental data and operational data.

[0090] Environmental data acquisition can be achieved by using environmental sensors (such as illuminance meters, anemometers, temperature / humidity sensors, and barometers) deployed at the system site to collect environmental parameters that affect the operation of the energy system in real time. For example, in photovoltaic power generation, illuminance and temperature directly affect power generation efficiency; in wind power generation, wind speed and direction determine the output power of the wind turbine; and the performance of energy storage systems is significantly affected by ambient temperature (e.g., lithium battery efficiency varies with temperature).

[0091] Operational data acquisition can collect real-time operational data from both the energy production end (such as photovoltaic inverters, wind power converters, and energy storage converters) and the energy consumption end (such as smart meters and industrial load controllers). This includes: production end: power generation, equipment status (such as fault codes), and efficiency parameters; consumption end: load power, electricity consumption mode (such as peak and off-peak periods), and interruptible load status.

[0092] Understandably, the introduction of external data (such as weather forecasts) and equipment status data (such as energy storage aging coefficients) enables the correction model to dynamically adjust the predicted feature vector and control strategy based on future operating conditions. For example, during the energy storage aging stage, the correction model will prioritize adjusting the SOC target curve to avoid over-discharge leading to a shortened equipment lifespan; when the charging pile load suddenly increases, the correction model will dynamically increase the energy storage discharge power limit, thereby further improving the system's adaptability to sudden scenarios.

[0093] Specifically, deviation analysis is performed on environmental data, operational data, and the original multi-timescale prediction data to calculate the deviation rate between the predicted and actual values ​​at each scale. Based on the deviation rate and equipment operating characteristics, a dynamic correction model is constructed to iteratively correct the long-term, medium-term, and short-term prediction data. Finally, based on the corrected prediction data, the generated cross-timescale control strategy and the control parameters of each device are adaptively adjusted within a hierarchical objective function framework to achieve closed-loop optimization of prediction and control, thereby improving the adaptability and control accuracy of the strategy under complex operating conditions.

[0094] For example, when the actual photovoltaic output is lower than the predicted value, if the weather forecast indicates an increase in solar irradiance the next day, the correction model will reduce the correction magnitude of the short-term forecast and prioritize the stability of the long-term forecast. If the energy storage aging coefficient is high, the correction model will prioritize adjusting the SOC target curve to avoid over-discharge that could shorten the equipment's lifespan. Through joint analysis of external data and equipment status data, the correction model can dynamically adjust the prediction feature vector and control strategy.

[0095] For example, taking a photovoltaic power output scenario, environmental data is collected by a light intensity meter and a temperature sensor deployed next to the photovoltaic array. When the real-time light intensity deviates from the historical prediction data by more than 15% or the ambient temperature deviates from the standard operating temperature by ±5℃, the short-term photovoltaic power output prediction data is iteratively corrected. Based on the corrected prediction data, the short-term energy storage charging and discharging power limit is adjusted synchronously to ensure the system power balance under photovoltaic power output fluctuations.

[0096] For example, in a tiered execution scenario, a scaled differential correction strategy can be adopted. Long-term forecast data is based on daily operational data as the core correction basis, and adjustments are only made when the weekly meteorological data deviation exceeds 20%. Medium-term forecast data is mainly corrected based on hourly operational data. Short-term forecast data is corrected in real time based on second-level real-time operational data, and the corresponding level of control strategy is updated synchronously after correction.

[0097] Optionally, corrections can be made in the event of equipment failure to ensure fault tolerance. For example, when abnormal operating data such as photovoltaic inverter fault codes are collected, the weight of the prediction data corresponding to the faulty equipment is reduced, and the output prediction data of the backup equipment (energy storage, wind power) is corrected based on environmental data. The control strategy is also adjusted to transfer the control tasks of the faulty equipment to the backup equipment to ensure continuous system operation.

[0098] This application provides a multi-timescale coordinated control method applied to a multi-source heterogeneous energy system. The method includes: acquiring multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at various time scales, each prediction data being generated through different prediction models; fusing features of the multiple prediction data to obtain a prediction feature vector; and, based on the prediction feature vector and a preset hierarchical objective function, obtaining a cross-timescale control strategy to solve the problems of asynchronous response of multi-source devices and cross-scale control conflicts in grid-type microgrids, thereby improving the system's operational stability, economy, and coordinated control accuracy under complex operating conditions.

[0099] Figure 3 A flowchart illustrating the multi-timescale coordinated control method provided in this application embodiment. Figure 2 .like Figure 3 As shown, this embodiment is... Figure 2 Based on the embodiments, a multi-timescale coordinated control method is described in detail, which includes:

[0100] S301. Obtain multi-timescale prediction data corresponding to multi-source heterogeneous energy systems.

[0101] Step S301 is similar to step S201 described above, and will not be repeated here.

[0102] S302. Based on the changing trends corresponding to multiple forecast data, dynamically assign weights to the time scales corresponding to the multiple forecast data.

[0103] In this way, by quantifying the changing trends of predicted data at different time scales (such as stability, volatility, and trend), and dynamically adjusting their weights, the control strategy can achieve a balance between short-term response and long-term optimization, thereby improving the robustness of the system.

[0104] Understandably, differentiating weight allocation improves the adaptability of the predicted feature vector to the actual operating conditions of the system. The weight ratio of each scale is dynamically adjusted based on the fluctuation characteristics and correlation of long-term, medium-term, and short-term predicted data: when long-term data trends are stable but short-term power fluctuations are severe, the weight of short-term predicted data is increased, allowing the control strategy to prioritize responses to short-term fluctuations; when long-term energy storage SOC trends are significant but short-term fluctuations are small, the weight of long-term predicted data is increased, allowing the control strategy to prioritize long-term energy storage optimization objectives. For example, under conditions of sudden drop in photovoltaic output, increasing the weight of short-term data ensures that the control strategy prioritizes adjusting the energy storage discharge power limit, improving the accuracy and adaptability of the control strategy.

[0105] Specifically, for example, when a flat long-term forecast curve (low uncertainty) is detected, but short-term forecast data shows high-frequency and large fluctuations, the weight of short-term forecast data is significantly increased, while the weight of long-term data is appropriately reduced, so that the generated control strategy prioritizes responding to short-term fluctuations. The system will quickly switch from "economic dispatch mode" to "power smoothing mode," instructing the energy storage system or rapid response load to act immediately to smooth out instantaneous disturbances and prevent the power limit at the grid connection point from being exceeded.

[0106] In one optional embodiment, weights are dynamically assigned to the time scales corresponding to the multiple prediction data based on the changing trends of the multiple prediction data, including: calculating the correlation between prediction data at different time scales based on the prediction data corresponding to the multiple time scales; and assigning corresponding weights to the prediction data corresponding to the time scales of different time spans based on the magnitude of the correlation.

[0107] In this model, the weights and correlations of the predicted data at different time scales change from positive to negative, following the order from longest to shortest time span. That is, the greater the correlation, the greater the weight allocation for predicted data at longer time scales, and the smaller the weight allocation for predicted data at shorter time scales, thus achieving a differentiated and precise allocation of weights.

[0108] Optionally, the correlation can be calculated in real time using an algorithm that fuses dynamic statistical correlation analysis with model confidence. Specifically, the system can construct a system state response vector for the current moment, which includes real-time collected frequency change rate, voltage deviation, and bus power fluctuation values. Simultaneously, it acquires predicted data sequences for long-term, medium-term, and short-term time scales. Using a sliding window mechanism (with a window length of W, e.g., data points from the past 15 minutes), the system calculates the dynamic Pearson correlation coefficient between the predicted sequences at each scale and the actual system state response vector. This coefficient quantifies the explanatory power or fit of the predicted data at a certain time scale to the actual operating state of the system within a recent historical window. Subsequently, to eliminate interference from the uncertainty of the prediction model itself, the system introduces a prediction confidence factor. This factor is obtained by normalizing the inverse of the prediction variance output by the corresponding prediction model: the smaller the variance of the prediction interval output by a model, the higher its confidence, and the closer the prediction confidence factor is to 1; conversely, it approaches 0. Finally, by combining statistical correlation and model confidence, the comprehensive correlation index for each scale is calculated, and then transformed into the final dynamic weight allocation through a normalization function (such as the Softmax function).

[0109] Understandably, if long-term forecast data (such as 24-hour photovoltaic power output) is strongly correlated with short-term forecast data, it indicates that short-term fluctuations conform to the long-term trend. In this case, the long-term weight should be increased (to optimize decision-making by utilizing trend certainty) and the short-term weight should be decreased (to avoid over-responding to noise). If short-term forecast data (such as 15-minute power output) is weakly correlated with long-term forecast data, it indicates that short-term fluctuations deviate from the long-term trend. In this case, the short-term weight should be increased (to prioritize responding to sudden changes) and the long-term weight should be decreased (to avoid being guided by erroneous trends).

[0110] By introducing correlation analysis as the triggering mechanism for weight allocation, intelligent adaptation of the control strategy is achieved. Specifically, under steady-state conditions (high correlation), the system automatically "trusts" long-term planning, focusing on economy and avoiding ineffective regulation; under disturbed conditions (low correlation), the system automatically "trusts" short-term measurements, focusing on safety and stability and quickly smoothing out fluctuations.

[0111] S303. Combining the weights corresponding to each time scale, feature fusion is performed on multiple prediction data to obtain a prediction feature vector.

[0112] Specifically, the system can fuse the dynamic weighting coefficients for the first time scale (long-term), the second time scale (medium-term), and the third time scale (short-term). Since the prediction data at different time scales have different time resolutions (e.g., long-term predictions may be at the hour level, while short-term predictions may be at the second or minute level) and data dimensions, the system first uses the current moment as a benchmark to map the prediction sequences of each time scale to a unified future prediction time window.

[0113] Understandably, for low-resolution data, upsampling is performed using interpolation algorithms (such as linear interpolation or cubic spline interpolation); for high-resolution data, downsampling is performed using moving average or key feature point extraction methods to ensure that all predicted data are strictly synchronized on the time axis. Subsequently, the predicted data for each physical dimension are normalized to eliminate the impact of dimensional differences on subsequent weighted calculations.

[0114] Secondly, a feature weighted fusion based on dynamic weights can be performed. By using dynamic weight coefficients, the preprocessed multi-scale prediction data is weighted to generate fused features. Subsequently, the trend feature sequence, volatility feature index, comprehensive confidence index, and other derived features (such as the evolution trend of energy storage state of charge) obtained by the above fusion are concatenated and encoded according to a preset dimensional order to construct a high-dimensional prediction feature vector.

[0115] Specifically, when long-term photovoltaic (PV) output trends are stable but short-term power fluctuations are drastic, dynamic weight allocation increases the weight of short-term forecast data, causing the control strategy to prioritize responses to short-term fluctuations. Conversely, when long-term energy storage SOC trends are significant but short-term fluctuations are small, dynamic weight allocation increases the weight of long-term forecast data, causing the control strategy to prioritize long-term energy storage optimization objectives. For example, in scenarios involving a sudden drop in PV output, dynamic weight allocation ensures that the control strategy prioritizes adjusting the energy storage discharge power limit, thereby further improving the accuracy and adaptability of the control strategy.

[0116] S304. Based on the predicted feature vector and the energy storage state optimization objective, obtain the first sub-strategy.

[0117] The first sub-strategy is designed for the first time scale (i.e., the long-term time scale). The system reads the generated prediction feature vector and extracts the long-term trend components (such as the net load curve for the next 24 hours and the expected value of photovoltaic output) as well as the comprehensive confidence index.

[0118] For example, by combining the energy storage state optimization objectives (i.e., minimizing electricity costs, maximizing peak-valley arbitrage profits, or equalizing the lifespan and wear of energy storage batteries throughout the entire daily operating cycle), the system can construct a long-term optimization model, such as Mixed Integer Programming (MILP).

[0119] Specifically, the long-term power trajectory in the predicted feature vector is used to predict the energy surplus or deficit for each future period. If the feature vector shows that there is sustained high photovoltaic output and low electricity price at midday, and peak load in the evening, the optimization model will calculate an optimal energy storage state of charge (SOC) reference trajectory.

[0120] The output of this first sub-strategy is manifested as the energy storage charging and discharging plan instructions for each hourly segment within the next 24 hours (e.g., charging at 50kW from 02:00 to 06:00 and discharging at 80kW from 18:00 to 22:00), and the SOC target value is set at the end of each time period.

[0121] S305. Based on the predicted feature vector and the power fluctuation smoothing objective, obtain the second sub-policy.

[0122] Among them, the second sub-strategy, which is geared towards the second time scale (i.e., the medium-term time scale, such as the minute to ten-minute scale), is used to analyze the predicted feature vector again, focusing on extracting the fusion fluctuation intensity index and the medium-term power change trend.

[0123] For example, in conjunction with the power fluctuation smoothing objective (i.e., suppressing the power ramp-up rate at the grid connection point, meeting the grid dispatch's restrictions on the power change rate, and reducing assessment penalties caused by new energy fluctuations), the system constructs a medium-term rolling optimization model (such as model predictive control).

[0124] When the intensity of fusion fluctuations in the predicted feature vector increases, it indicates a significant risk of power drop or surge in the near future (such as a sudden drop in photovoltaic power due to cloud cover). In this case, the optimization model no longer simply pursues economic efficiency, but prioritizes calculating energy storage power limit adjustment schemes or inverter output correction curves that can mitigate the fluctuations.

[0125] The resulting second sub-strategy manifests as a dynamic power reference value and upper and lower limits for charging and discharging power within the next hour (e.g., dynamically increasing the maximum discharge power limit of energy storage from 100kW to 150kW within the next 15 minutes to compensate for the predicted photovoltaic power shortfall; or adjusting the active power setpoint of the photovoltaic inverter to output power along a smooth curve). This strategy focuses on power balance and grid connection compliance.

[0126] S306. Based on the predicted feature vector and the voltage frequency stability target, obtain the third sub-strategy.

[0127] Among them, with the third sub-strategy oriented towards the third time scale (short-term time scale, such as second to millisecond), the system can deeply mine high-frequency transient features (such as maximum ramp rate, instantaneous disturbance markers) and short-term confidence markers in the predicted feature vector.

[0128] Specifically, a hierarchical objective function design is employed to achieve coordinated control of long-term energy storage SOC optimization, medium-term power fluctuation smoothing, and short-term voltage and frequency stability objectives. The long-term energy storage SOC optimization objective formulates energy storage charging and discharging plans by predicting long-term photovoltaic output and energy storage SOC trends in the eigenvector. The medium-term power smoothing objective adjusts energy storage charging and discharging power limits by predicting medium-term load fluctuation information in the eigenvector. The short-term voltage and frequency stability objective optimizes VSG parameters by predicting short-term instantaneous power fluctuation information in the eigenvector. For example, in the long term, a SOC adjustment margin is reserved to cope with short-term fluctuations; in the medium term, the maximum power point tracking (MPPT) strategy of the photovoltaic inverter is used to adjust and smooth the power; and in the short term, VSG parameters are optimized to maintain voltage and frequency stability, thereby further improving the synergy and adaptability of the control strategy.

[0129] For example, in conjunction with the voltage and frequency stability objective (i.e., maintaining the voltage and frequency of the microgrid bus within a safe range and providing inertia support when sudden load switching or faults occur), the system can construct short-term fast response rules or parameter adaptive algorithms.

[0130] If the predicted eigenvector indicates an impending high-power surge (such as the connection of a large charging pile or the start-up of a motor), and the short-term confidence flag of the short-term weights is high, the system will immediately trigger the virtual synchronous machine (VSG) parameter adaptive mechanism. Based on the predicted disturbance magnitude, key parameters in the VSG control loop, including virtual inertia, damping coefficient, and virtual impedance, are calculated and updated in real time.

[0131] The resulting third sub-strategy manifests as the set of underlying control parameters for the energy storage converter or grid-connected inverter. This strategy does not directly set the power value, but rather alters the system's dynamic characteristics to enhance its disturbance rejection capability, focusing on system stability and power quality.

[0132] For example, the energy storage state optimization objective is used for long-term scale regulation. By optimizing the energy storage charging and discharging schedule, long-term operating costs are reduced and the energy storage life is extended. An example is setting a daily target curve for energy storage SOC: charging to 90% from 00:00 to 06:00 and discharging to 30% from 12:00 to 14:00. The power fluctuation smoothing objective is used for medium-term scale regulation. By adjusting the photovoltaic inverter MPPT strategy and the energy storage charging and discharging power limits, power fluctuations are reduced. An example is increasing the energy storage discharge limit from 100kW to 150kW when photovoltaic output decreases by 30%. The voltage and frequency stability objective is used for short-term scale regulation. By optimizing parameters such as VSG virtual inertia, damping, and virtual impedance, voltage and frequency fluctuations are suppressed. An example is adjusting the VSG virtual inertia from 12 to 18 when charging piles suddenly connect. These three types of sub-strategies operate collaboratively within a hierarchical objective function framework, ensuring the consistency and synergy of control strategies at different time scales.

[0133] It should be noted that the above strategies can guide the medium term in the long term. That is, the energy storage SOC operating range of the medium-term strategy is constrained by the SOC target value set by the long-term strategy, ensuring that short-term adjustments do not deviate from the long-term economic plan. The medium term can limit the short term; the power output upper limit of the short-term strategy is constrained by the dynamic limit calculated by the medium-term strategy, preventing frequent adjustments from depleting the energy storage capacity. The short term can be adjusted for execution. At the moment of actual execution, the underlying controller uses the parameters optimized by the short-term strategy as a benchmark to execute the power command given by the medium-term strategy, and monitors in real time whether the safety boundary set by the long-term strategy has been reached.

[0134] S307. Based on the delay characteristics of the energy production end and the energy consumption end, set the distribution advance time for each energy production end and each energy consumption end.

[0135] The energy production and consumption ends can include, for example, system equipment such as photovoltaic inverters, energy storage PCS, charging piles, and heat pumps. After generating a control strategy across time scales, the inherent response delay characteristics of the system equipment are first identified.

[0136] Understandably, control strategies should be prioritized and their issuance time adjusted based on the response delay characteristics of the equipment. Different devices exhibit varying response delays due to their construction, operating principles, and other factors. For example, some large energy production equipment may require a considerable amount of time from receiving a control command to actually starting to adjust its production output; while some small energy-consuming devices may respond quickly. Based on these differences, appropriate control strategy priorities should be set for different devices to ensure that critical equipment or equipment with high response time requirements executes control strategies first.

[0137] At the same time, taking into account the equipment's response delay characteristics, control commands are issued in advance to ensure that the equipment can perform the corresponding operations in a timely and accurate manner when a response is required, so as to maintain the stable operation of the energy system.

[0138] Optionally, the system can adopt an asynchronous distribution and synchronous execution mechanism. Specifically, asynchronous distribution means that the control center sends instructions to the corresponding devices at different times based on the distribution lead time of different devices. For example, if power balance needs to be achieved at t=10:00:00, the system may send instructions to the diesel engine at t=09:59:50 (10 seconds in advance), and send instructions to the energy storage at t=10:00:00 (0 seconds in advance).

[0139] Synchronous activation is achieved by including a unified timestamp tag or activation trigger condition in the instruction. Upon receiving the instruction, if the current time has not yet reached the activation time, the device stores the instruction in a buffer. Once the designated activation time arrives, all devices, regardless of when they received the instruction, execute the action simultaneously. This ensures physical consistency in the actions of multi-source heterogeneous devices, avoiding instantaneous power surges caused by "first-to-last" actions due to differences in response speed.

[0140] Assume the system predicts a load surge at t=14:00:00, requiring both energy storage and diesel generators to operate simultaneously. The energy storage device has a delay of approximately 50ms, with a distribution lead time set to 100ms. The system issues a command at 13:59:59.900. The diesel generator has a delay of approximately 8s (including start-up and ramp-up), with a distribution lead time set to 10s. The system issues a command at 13:59:50.000. Although the command issuance times differ by nearly 10 seconds, both reach their expected output at 14:00:00, achieving perfect power complementarity and minimizing bus frequency fluctuations.

[0141] S308. According to the priority order of the first sub-strategy, the second sub-strategy, and the third sub-strategy, and according to the distribution advance time corresponding to each energy production end and each energy consumption end, the first sub-strategy, the second sub-strategy, and the third sub-strategy are distributed in sequence, so that the equipment in the multi-source heterogeneous energy system can adjust its state according to the first sub-strategy, the second sub-strategy, or the third sub-strategy.

[0142] The three sub-strategies are sequentially issued and executed according to the preset priority order of the first, second, and third sub-strategies, combined with the distribution lead time configured for each energy production and consumption end. In the multi-source heterogeneous energy system, the photovoltaic inverters, energy storage PCS, charging piles, heat pumps and other equipment adjust their operating status according to the received corresponding sub-strategy instructions, ensuring that each device performs control actions synchronously based on its own response characteristics, so as to achieve accurate implementation of cross-time scale control strategies and global coordinated control of the system.

[0143] Understandably, the system can pre-maintain a policy priority queue. Based on the urgency of the control objective and its impact on system security, the priority order of three types of sub-policies can be pre-set. That is, when commands at different scales conflict, a fuzzy decision algorithm can be used to coordinate command parameters with network stability as the core, and set the corresponding policy priorities. For example, the highest priority is the third sub-policy (short-term voltage and frequency stability), because it involves the system's transient stability, with a response window of milliseconds to seconds. Execution delays may lead to frequency collapse or voltage exceeding limits. The second highest priority is the second sub-policy (medium-term power fluctuation smoothing), because it involves grid connection compliance and power quality, with a response window of minutes, and must be executed under the premise that short-term stability is guaranteed. The basic priority is the first sub-policy (long-term energy storage state optimization), because it involves economic efficiency and long-term energy balance, with a response window of hours, allowing it to be temporarily overridden or corrected by higher-priority policies in emergency situations.

[0144] Specifically, when different policies generate conflicting instructions for the same device at the same time (for example, a long-term policy requires charging while a short-term policy requires discharging to support the frequency), the system forcibly locks and executes the instruction of the high-priority policy, while the low-priority policy automatically enters a "suspended" or "corrected" state and resumes execution after the high-priority event ends.

[0145] After prioritizing, the system can utilize the distribution lead time of each device to construct a dynamic asynchronous distribution timeline. For example, the distribution of the third sub-strategy is executed first. This is for fast-response devices such as energy storage converters and static var generators. When the system detects a high-frequency disturbance signal or reaches a second-level control cycle, it immediately reads their extremely short distribution lead time (typically tens to hundreds of milliseconds). Within this tiny time window before the target takes effect, the system sends the third sub-strategy packet, containing parameter adjustment instructions such as virtual inertia and damping coefficient, to the device with the highest communication priority. This ensures that the device can complete parameter reconstruction the instant the disturbance occurs, providing immediate inertia support.

[0146] Next, the second sub-strategy is executed. For medium-speed response devices such as photovoltaic inverters, diesel generators, and adjustable loads, the system reads the distribution lead time (in the range of seconds to tens of seconds) based on medium-term power fluctuation predictions. At the corresponding time point before the target takes effect, the system issues the second sub-strategy package containing instructions such as power limits and ramp rate constraints. During this process, if a short-term strategy is detected being executed, the system will automatically limit the medium-term power instructions within the safety boundaries defined by the short-term strategy to prevent frequency stability from being disrupted by sudden power changes.

[0147] Finally, the first sub-strategy is executed. For slow-moving devices and energy management layers with thermal inertia or chemical slow-release characteristics, the system reads their minute-level distribution lead time based on the long-term optimization plan. A considerable time before the target takes effect, the system issues the first sub-strategy package containing instructions such as the State of Charge (SOC) target curve and the day-ahead charge / discharge plan. This strategy sets the system's operating baseline for the entire day; short- and medium-term strategies will be fine-tuned based on this, rather than being completely overhauled.

[0148] For example, command timing compensation driven by device response latency characteristics can improve the synchronization of control strategy execution. Specifically, short-term control strategies, due to their high urgency, are executed first and issued 10ms in advance to ensure voltage and frequency stability; medium-term control strategies, due to their lower response latency, are executed second-highest and issued 1 second in advance to ensure power smoothing; and long-term control strategies, due to their higher response latency, are executed last and issued 30 seconds in advance to ensure the achievement of energy storage SOC optimization goals. For instance, when both the heat pump and energy storage need to adjust their power simultaneously, command timing compensation ensures that energy storage adjustments are executed first, thereby further improving system stability.

[0149] Optionally, during strategy execution, the system collects real-time actual operating data of the equipment (such as actual power, bus voltage, frequency deviation, etc.) and compares it with the predicted target. If the actual response deviates from the expectation (such as instruction delay due to network jitter or insufficient response due to equipment failure), the system will immediately trigger a recalculation mechanism to reassess the latency characteristics and generate corrected strategy instructions, and re-enter the distribution process, thus forming a complete closed-loop control of "prediction-decision-distribution-execution-feedback".

[0150] This application provides a multi-timescale coordinated control method. It acquires long-, medium-, and short-term multi-timescale prediction data from a multi-source heterogeneous energy system, calculates inter-scale correlations based on data trends, dynamically allocates weights according to the scale from largest to smallest span and the rule of positive to negative correlation, and obtains a predicted feature vector by combining the weighted data. Based on this vector and three hierarchical objectives—energy storage optimization, power smoothing, and voltage-frequency stability—corresponding three-level sub-strategies are generated. Then, a distribution lead time is set according to the response delay characteristics of source-grid-load equipment. Finally, strategies are distributed sequentially according to preset priorities and lead times, driving system equipment to complete state adjustments and achieving cross-timescale coordinated control. Through dynamic weight fusion and hierarchical strategy time-sequential precise distribution, the method solves the problems of asynchronous response of multi-source equipment and cross-scale control conflicts in grid-type microgrids, significantly improving the operational stability, economy, and coordinated control accuracy under complex system conditions.

[0151] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0152] Based on the same inventive concept, this application also provides a multi-timescale coordinated control device for implementing the multi-timescale coordinated control method described above. The solution provided by this multi-timescale coordinated control device is similar to the solution described in the multi-timescale coordinated control method above. Therefore, the specific limitations in one or more device embodiments provided below can be found in the limitations of the multi-timescale coordinated control method described above, and will not be repeated here.

[0153] In one embodiment, such as Figure 4 As shown, a multi-timescale coordinated control device 400 is provided for use in a multi-source heterogeneous energy system. The device 400 includes:

[0154] The first acquisition module 401 is used to acquire multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at various time scales, and each prediction data is generated by a different prediction model.

[0155] The fusion module 402 is used to perform feature fusion on multiple prediction data to obtain a prediction feature vector;

[0156] The second acquisition module 403 is used to combine a preset hierarchical objective function and, based on the predicted feature vector, acquire a control strategy across time scales. The hierarchical objective function includes at least an energy storage state optimization objective corresponding to the first time scale, a power fluctuation smoothing objective corresponding to the second time scale, and a voltage frequency stability objective corresponding to the third time scale; and the time spans corresponding to the first time scale, the second time scale, and the third time scale decrease sequentially; the control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

[0157] In one possible implementation, the hierarchical objective function includes at least the energy storage state optimization objective corresponding to the first time scale, the power fluctuation smoothing objective corresponding to the second time scale, and the voltage frequency stability objective corresponding to the third time scale; and the time spans corresponding to the first time scale, the second time scale, and the third time scale decrease sequentially.

[0158] The control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

[0159] The second acquisition module 403 is also used to acquire the first sub-strategy based on the predicted feature vector and the energy storage state optimization target.

[0160] The second acquisition module 403 is also used to acquire a second sub-strategy based on the predicted feature vector and the power fluctuation smoothing target.

[0161] The second acquisition module 403 is also used to acquire a third sub-strategy based on the predicted feature vector and the voltage frequency stability target.

[0162] In one possible implementation, the multi-source heterogeneous energy system includes at least one energy production end and at least one energy consumption end; the device also includes: a setting module and a distribution module;

[0163] The configuration module is used to set the distribution advance time for each energy production end and each energy consumption end based on the corresponding delay characteristics of the energy production end and energy consumption end.

[0164] The distribution module is used to distribute the first sub-strategy, the second sub-strategy, and the third sub-strategy in a pre-set priority order according to the distribution advance time corresponding to each energy production end and each energy consumption end, so that the equipment in the multi-source heterogeneous energy system can adjust its state according to the first sub-strategy, the second sub-strategy, or the third sub-strategy.

[0165] In one possible implementation, the fusion module 402 is also used to dynamically assign weights to the time scales corresponding to the multiple prediction data based on the changing trends corresponding to the multiple prediction data.

[0166] The fusion module 402 is also used to combine the weights corresponding to each time scale to perform feature fusion on multiple prediction data to obtain a prediction feature vector.

[0167] In one possible implementation, the fusion module 402 is further configured to calculate the correlation between prediction data at different time scales based on prediction data corresponding to multiple time scales.

[0168] The fusion module 402 is also used to assign corresponding weights to the prediction data corresponding to different time scales based on the magnitude of the correlation; wherein, in order of decreasing time span, the weight of the prediction data corresponding to different time scales changes from positive correlation to negative correlation with the magnitude of the correlation.

[0169] In one possible implementation, the second acquisition module 403 is further configured to acquire environmental data of the environment in which the multi-source heterogeneous energy system is located, as well as operational data corresponding to at least one energy production end and at least one energy consumption end in the multi-source heterogeneous energy system.

[0170] The second acquisition module 403 is also used to correct multi-timescale prediction data and control strategies based on environmental data and operational data.

[0171] Each module in the above-mentioned device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0172] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Figure 5 As shown, the electronic device 500 provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device 500 further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0173] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0174] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0175] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0176] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0177] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0178] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0179] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0180] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0181] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0182] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0183] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0184] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0185] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0186] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0187] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A multi-timescale coordinated control method, characterized in that, Applied to multi-source heterogeneous energy systems; The method includes: Acquire multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different timescales, and each prediction data is generated by a different prediction model. The prediction data from multiple sources are fused to obtain a prediction feature vector. Combining a preset hierarchical objective function, a control strategy across time scales is obtained based on the predicted feature vector. The hierarchical objective function includes at least an energy storage state optimization objective corresponding to the first time scale, a power fluctuation smoothing objective corresponding to the second time scale, and a voltage frequency stability objective corresponding to the third time scale; and the time spans corresponding to the first time scale, the second time scale, and the third time scale decrease sequentially. The control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

2. The method according to claim 1, characterized in that, The step of combining a preset hierarchical objective function with the predicted feature vector to obtain a cross-timescale control strategy includes: Based on the predicted feature vector and the energy storage state optimization objective, the first sub-strategy is obtained; Based on the predicted feature vector and the power fluctuation smoothing objective, the second sub-strategy is obtained; Based on the predicted feature vector and the voltage frequency stability target, the third sub-strategy is obtained.

3. The method according to claim 2, characterized in that, The multi-source heterogeneous energy system includes at least one energy production end and at least one energy consumption end; After obtaining the control strategy across time scales, the method further includes: Based on the delay characteristics of the energy production end and the energy consumption end, a distribution advance time is set for each of the energy production ends and each of the energy consumption ends; According to the priority order preset by the first sub-strategy, the second sub-strategy, and the third sub-strategy, and according to the distribution advance time corresponding to each energy production end and each energy consumption end, the first sub-strategy, the second sub-strategy, and the third sub-strategy are distributed in sequence, so that the equipment in the multi-source heterogeneous energy system can adjust its state according to the first sub-strategy, the second sub-strategy, or the third sub-strategy.

4. The method according to claim 1, characterized in that, The step of fusing features from multiple predicted data to obtain a predicted feature vector includes: Based on the changing trends corresponding to the various predicted data, weights are dynamically assigned to the time scales corresponding to the various predicted data. By combining the weights corresponding to each time scale, feature fusion is performed on the various prediction data to obtain the prediction feature vector.

5. The method according to claim 4, characterized in that, The step of dynamically assigning weights to the time scales corresponding to the various predicted data based on their changing trends includes: Based on the forecast data corresponding to multiple time scales, the correlation between forecast data at different time scales is calculated. Based on the magnitude of the correlation, corresponding weights are assigned to the predicted data corresponding to different time scales; wherein, in descending order of the time span, the weights of the predicted data corresponding to different time scales change from positively correlated to negatively correlated with the magnitude of the correlation.

6. The method according to claim 1, characterized in that, The method further includes: Obtain environmental data of the environment in which the multi-source heterogeneous energy system is located, as well as the operational data corresponding to at least one energy production end and at least one energy consumption end in the multi-source heterogeneous energy system; Based on the environmental data and the operational data, the multi-timescale prediction data and the control strategy are revised.

7. A multi-timescale coordinated control device, characterized in that, Applied to multi-source heterogeneous energy systems; The device includes: The first acquisition module is used to acquire multi-timescale prediction data corresponding to the multi-source heterogeneous energy system; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different timescales, and each prediction data is generated by a different prediction model. The fusion module is used to perform feature fusion on multiple prediction data to obtain a prediction feature vector; The second acquisition module is used to combine a preset hierarchical objective function and, based on the predicted feature vector, acquire a cross-timescale control strategy. The hierarchical objective function includes at least an energy storage state optimization objective corresponding to the first timescale, a power fluctuation smoothing objective corresponding to the second timescale, and a voltage frequency stability objective corresponding to the third timescale; and the time spans corresponding to the first timescale, the second timescale, and the third timescale decrease sequentially. The control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

8. A multi-source heterogeneous energy system, characterized in that, The multi-source heterogeneous energy system includes at least one energy production end and at least one energy consumption end; The multi-source heterogeneous energy system is used to acquire multi-timescale prediction data corresponding to the multi-source heterogeneous energy system by employing the multi-timescale coordinated control method as described in any one of claims 1-6; the multi-timescale prediction data includes prediction data of the multi-source heterogeneous energy system at multiple different timescales, each prediction data being generated by a different prediction model; and feature fusion is performed on multiple prediction data to obtain a prediction feature vector; and a cross-timescale control strategy is obtained based on the prediction feature vector by combining a preset hierarchical objective function, wherein the hierarchical objective function includes at least an energy storage state optimization objective corresponding to a first timescale, a power fluctuation smoothing objective corresponding to a second timescale, and a voltage frequency stability objective corresponding to a third timescale; and the time spans corresponding to the first timescale, the second timescale, and the third timescale decrease sequentially. The control strategy includes at least a first sub-strategy corresponding to the first time scale, a second sub-strategy corresponding to the second time scale, and a third sub-strategy corresponding to the third time scale.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images