Multi-energy system and control method of multi-energy system

Abstract

The embodiment of the application provides a kind of multi-energy system and the control method of multi-energy system, belong to energy management technical field.The method is in response to the day-ahead energy scheduling strategy generated by multi-energy system, obtains the multi-source heterogeneous data collected by edge energy management unit, the current time and instantaneous running data of multi-energy system;Determine the target preset period corresponding to the current time, and the day-ahead load prediction data in the period;Multi-source heterogeneous data is input into edge prediction model to infer, and real-time load prediction data is obtained;Based on real-time load prediction data and day-ahead load prediction data, the day-ahead time-sharing scheduling data corresponding to target preset period is corrected, and obtains intra-day time-sharing scheduling data;When instantaneous running data does not satisfy abnormal operating condition, control multi-energy system in target preset period according to intra-day time-sharing scheduling data runs.The method reduces the dependence of multi-energy system collaborative control on communication, improves the real-time performance and accuracy of energy scheduling.

Description

Technical Field

[0001] This application relates to the field of energy management technology, and in particular to a multi-energy system and a control method for the multi-energy system. Background Technology

[0002] In home energy management, cloud-based centralized scheduling is typically used to coordinate the control of devices such as photovoltaics, energy storage, heat pumps, and home appliances, in order to achieve unified management of distributed energy.

[0003] However, the above-mentioned solutions generally rely on the cloud to continuously complete forecasts and issue commands, resulting in insufficient local autonomous decision-making capabilities. When communication is interrupted or latency increases, it is difficult to respond in a timely manner to photovoltaic fluctuations and load changes. At the same time, the coordination of multi-timescale scheduling is insufficient, and forecast errors are easy to accumulate, leading to a decrease in control accuracy and operational stability.

[0004] Therefore, how to reduce the dependence of home energy system collaborative control on communication and improve the real-time performance, accuracy and stability of scheduling under dynamic operating conditions has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application provides a multi-energy system and a control method for the multi-energy system, which reduces the dependence of home energy system collaborative control on communication and improves the real-time performance, accuracy and stability of scheduling under dynamic operating conditions.

[0006] In a first aspect, embodiments of this application provide a control method for a multi-energy system, applied to an edge energy management unit of the multi-energy system, the edge energy management unit being deployed on the user side of the multi-energy system; the method includes:

[0007] In response to the day-ahead energy dispatch strategy generated by the multi-energy system, the system acquires multi-source heterogeneous data collected by the edge energy management unit, and obtains the current time and instantaneous operation data of the multi-energy system; the day-ahead energy dispatch strategy includes day-ahead time-sharing dispatch data corresponding to multiple preset periods;

[0008] Determine the target preset period corresponding to the current time, as well as the day-ahead load forecast data within the target preset period, and input the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system;

[0009] Based on real-time load forecast data and day-ahead load forecast data, the day-ahead time-sharing scheduling data corresponding to the target preset period is corrected to obtain intraday time-sharing scheduling data.

[0010] If the instantaneous operating data does not meet the abnormal operating conditions, the control system will operate according to the intraday time-sharing scheduling data within the target preset period.

[0011] In one possible implementation, the multi-energy system further includes a combined heat and power (CHP) system, a photovoltaic (PV) system, and an energy storage system; the day-ahead energy dispatch strategy includes a day-ahead energy supply strategy for the CHP system, a day-ahead energy consumption and allocation strategy for the PV system, and a day-ahead charge and discharge timing strategy for the energy storage system; prior to responding to the day-ahead energy dispatch strategy of the multi-energy system, the method further includes:

[0012] In response to the day-ahead energy supply strategy sent by the energy coordination unit, the power constraint conditions corresponding to the day-ahead energy supply strategy are determined; the power constraint conditions are used to characterize the available power range of the heat pump tri-generation system.

[0013] The historical multi-source heterogeneous data collected by the edge energy management unit is determined, and the historical multi-source heterogeneous data is input into the edge prediction model to obtain the day-ahead photovoltaic power output prediction data and the day-ahead electricity load prediction data;

[0014] The operating status of the energy storage system is determined, and based on the operating status, day-ahead photovoltaic output forecast data, day-ahead electrical load forecast data, and power constraints, a thermoelectric coupling scheduling optimization solution is performed with the optimization direction of a preset multi-objective function as the optimization direction. This yields day-ahead consumption allocation strategies and day-ahead charge-discharge timing strategies corresponding to multiple preset periods. The thermoelectric coupling scheduling optimization solution is implemented using at least one of the following: weighted summation multi-objective optimization algorithm, particle swarm optimization algorithm, and genetic algorithm.

[0015] In one possible implementation, based on real-time load forecast data and day-ahead load forecast data, the day-ahead time-of-day scheduling data corresponding to the target preset period is corrected to obtain intraday time-of-day scheduling data, including:

[0016] By comparing real-time load forecast data with day-ahead load forecast data, the scheduling deviation value of the corresponding multi-energy system is obtained.

[0017] The correction strategy that matches the scheduling deviation value is determined as the target correction strategy;

[0018] The daily time-sharing scheduling data corresponding to the target preset period is corrected according to the target correction strategy to obtain the intraday time-sharing scheduling data.

[0019] In one possible implementation, the correction strategy includes a first correction strategy; determining the correction strategy that matches the scheduling deviation value as the target correction strategy includes:

[0020] If the scheduling deviation is less than the preset deviation threshold, the first correction strategy will be determined as the target correction strategy.

[0021] The daily time-sharing scheduling data corresponding to the target preset period is corrected according to the target correction strategy to obtain the intraday time-sharing scheduling data, including:

[0022] Based on real-time load forecast data, determine the system load status corresponding to the target preset period;

[0023] Based on scheduling deviations and system load status, the compensation priority and adjustable margin of different energy systems are determined.

[0024] Based on the compensation priorities and adjustable margins of different energy systems, the daytime time-sharing data corresponding to the target preset cycle is corrected to obtain the intraday time-sharing data.

[0025] In one possible implementation, the correction strategy further includes a second correction strategy; determining the correction strategy that matches the scheduling deviation value as the target correction strategy also includes:

[0026] If the scheduling deviation is not less than the preset deviation threshold, the second correction strategy will be determined as the target correction strategy.

[0027] Real-time load forecast data includes intraday heat load forecast data, intraday photovoltaic output forecast data, and intraday electricity load forecast data; intraday time-of-use scheduling data is obtained by correcting the day-ahead time-of-use scheduling data corresponding to the target preset period according to the target correction strategy, and also includes:

[0028] Determine the equipment boundary constraints and system configuration data for multi-energy systems;

[0029] Based on system configuration data, intraday heat load forecast data, intraday photovoltaic output forecast data, intraday electricity load forecast data, and equipment boundary constraints, the optimization direction is to optimize the thermoelectric coupling scheduling, and the real-time energy scheduling data corresponding to the target preset period is obtained.

[0030] Based on real-time energy dispatch data, the daytime time-sharing dispatch data corresponding to the target preset cycle is updated to obtain intraday time-sharing dispatch data.

[0031] In one possible implementation, abnormal operating conditions include a preset abnormal threshold and a preset allowable range; after correcting the daytime time-sharing scheduling data corresponding to the target preset period based on real-time load forecast data and daytime load forecast data to obtain intraday time-sharing scheduling data, the method further includes:

[0032] Based on instantaneous operating data, the instantaneous power deviation of the multi-energy system is determined; the instantaneous power deviation is used to reflect the degree of deviation between power supply and demand in the multi-energy system at the current instant.

[0033] If the instantaneous power deviation does not exceed the preset abnormal threshold and the instantaneous operating data does not exceed the preset allowable range, it is determined that the instantaneous operating data does not meet the abnormal operating conditions.

[0034] If the instantaneous power deviation exceeds the preset abnormal threshold, and / or the instantaneous operating data exceeds the preset allowable range, the instantaneous operating data is determined to meet the abnormal operating conditions.

[0035] The instantaneous operating data includes at least one of the following: photovoltaic instantaneous power, energy storage instantaneous charging and discharging power, heat pump instantaneous power, load instantaneous power, grid connection point frequency detection data, and grid connection point voltage detection data.

[0036] In one possible implementation, after determining that the instantaneous running data meets the abnormal running conditions, the method further includes:

[0037] Based on instantaneous operational data, determine the type and degree of abnormal disturbances in the multi-energy system;

[0038] Based on the type and degree of abnormal disturbance, determine the target compensation strategy for the multi-energy system;

[0039] According to the target compensation strategy, the intraday time-sharing scheduling data is corrected to obtain the corrected time-sharing scheduling data, and the multi-energy system is controlled to operate according to the corrected time-sharing scheduling data within the target preset period.

[0040] The target compensation strategy includes at least one of the following: energy storage compensation control strategy, heat pump power regulation control strategy, and load reduction control strategy.

[0041] Secondly, embodiments of this application provide a control method for a multi-energy system, applied to an energy coordination unit of the multi-energy system, wherein the energy coordination unit is deployed on the service side of the multi-energy system; the method includes:

[0042] Acquire historical multi-source heterogeneous data sent by the edge power management unit;

[0043] Historical multi-source heterogeneous data is input into a cloud-based prediction model to obtain day-ahead prediction data for multi-energy systems; the day-ahead prediction data includes heat load prediction data.

[0044] Based on heat load forecast data, with the optimization direction of the preset multi-objective function, the thermoelectric coupling scheduling optimization is performed to obtain the daytime energy supply strategy corresponding to multiple preset cycles.

[0045] The daytime energy supply strategy is sent to the edge energy management unit.

[0046] In one possible implementation, the cloud-based prediction model includes a photovoltaic power generation prediction branch, a building thermal environment prediction branch, a domestic hot water load prediction branch, a household appliance load prediction branch, and a heat pump performance prediction branch; the method further includes:

[0047] Extract photovoltaic power output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data from historical multi-source heterogeneous data;

[0048] The data types of photovoltaic output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data are determined respectively, and for each data type, a target modeling algorithm corresponding to the data type is matched;

[0049] Different target modeling algorithms corresponding to different data types are used to perform differentiated training on historical multi-source heterogeneous data to obtain multiple prediction branches;

[0050] By integrating multiple prediction branches, a cloud-based prediction model is obtained;

[0051] Among them, the photovoltaic power generation prediction branch was trained using a long short-term memory network model; the building thermal environment prediction branch was trained using an RC building thermal environment model; the domestic hot water load prediction branch was trained using an ARIMA model; the household appliance load prediction branch was trained using a random forest model; and the heat pump performance prediction branch was trained using an SVR model.

[0052] In one possible implementation, after integrating multiple prediction branches to obtain a cloud-based prediction model, the method further includes:

[0053] Knowledge distillation is performed on the cloud-based prediction model to obtain a lightweight edge prediction model;

[0054] The edge prediction model is sent to the edge energy management unit.

[0055] Thirdly, embodiments of this application provide a control device for a multi-energy system, applied to an edge energy management unit of a multi-energy system, comprising:

[0056] The acquisition module is used to acquire multi-source heterogeneous data collected by the edge energy management unit in response to the day-ahead energy dispatch strategy generated by the multi-energy system, and to acquire the current time and instantaneous operation data of the multi-energy system; the day-ahead energy dispatch strategy includes day-ahead time-sharing dispatch data corresponding to multiple preset periods.

[0057] The processing module is used to determine the target preset period corresponding to the current time, as well as the day-ahead load forecast data within the target preset period, and input the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system.

[0058] The processing module is also used to correct the daytime time-sharing scheduling data corresponding to the target preset period based on real-time load forecast data and daytime load forecast data, so as to obtain intraday time-sharing scheduling data.

[0059] The processing module is also used to control the multi-energy system to operate according to the intraday time-sharing scheduling data within the target preset period when the instantaneous operating data does not meet the abnormal operating conditions.

[0060] In one possible implementation, the processing module is further configured to determine the power constraint conditions corresponding to the day-ahead energy supply strategy in response to the day-ahead energy supply strategy sent by the energy coordination unit; the power constraint conditions are used to characterize the available power range of the heat pump tri-generation system.

[0061] The processing module is also used to determine the historical multi-source heterogeneous data collected by the edge energy management unit, and input the historical multi-source heterogeneous data into the edge prediction model to obtain day-ahead photovoltaic power output prediction data and day-ahead electrical load prediction data.

[0062] The processing module is also used to determine the operating status of the energy storage system, and based on the operating status, day-ahead photovoltaic output forecast data, day-ahead electrical load forecast data, and power constraints, it performs thermoelectric coupling scheduling optimization with the optimization direction of the preset multi-objective function as the optimization direction, and obtains day-ahead consumption allocation strategy and day-ahead charge and discharge timing strategy corresponding to multiple preset periods; the thermoelectric coupling scheduling optimization is implemented by at least one of the following: weighted summation multi-objective optimization algorithm, particle swarm optimization algorithm, and genetic algorithm.

[0063] In one possible implementation, the processing module is further configured to compare the real-time load forecast data with the day-ahead load forecast data to obtain the scheduling deviation value of the corresponding multi-energy system.

[0064] The processing module is also used to determine the correction strategy that matches the scheduling deviation value as the target correction strategy.

[0065] The processing module is also used to correct the day-ahead time-sharing scheduling data corresponding to the target preset period according to the target correction strategy, so as to obtain the intraday time-sharing scheduling data.

[0066] In one possible implementation, the processing module is further configured to determine the first correction strategy as the target correction strategy if the scheduling deviation value is less than a preset deviation threshold.

[0067] The processing module is also used to determine the system load status corresponding to the target preset period based on real-time load forecast data.

[0068] The processing module is also used to determine the compensation priority and adjustable margin of different energy systems based on scheduling deviations and system load status.

[0069] The processing module is also used to correct the daytime time-sharing scheduling data corresponding to the target preset cycle based on the compensation priority and adjustable margin of different energy systems, so as to obtain the intraday time-sharing scheduling data.

[0070] In one possible implementation, the processing module is further configured to determine the second correction strategy as the target correction strategy if the scheduling deviation value is not less than a preset deviation threshold.

[0071] The processing module is also used to determine the equipment boundary constraints and system configuration data of multi-energy systems.

[0072] The processing module is also used to perform thermal-electric coupling scheduling optimization based on system configuration data, intraday heat load forecast data, intraday photovoltaic power output forecast data, intraday electricity load forecast data and equipment boundary constraints, with the optimization direction of the preset multi-objective function, to obtain real-time energy scheduling data corresponding to the preset target period.

[0073] The processing module is also used to update the daytime time-sharing data corresponding to the target preset cycle based on real-time energy dispatch data, so as to obtain the intraday time-sharing data.

[0074] In one possible implementation, the processing module is further configured to determine the instantaneous power deviation of the multi-energy system based on instantaneous operating data; the instantaneous power deviation is used to reflect the degree of deviation between power supply and demand of the multi-energy system at the current instantaneous moment.

[0075] The processing module is also used to determine that the instantaneous operating data does not meet the abnormal operating conditions when the instantaneous power deviation does not exceed the preset abnormal threshold and the instantaneous operating data does not exceed the preset allowable range.

[0076] The processing module is also used to determine whether the instantaneous operating data meets the abnormal operating conditions when the instantaneous power deviation exceeds the preset abnormal threshold and / or the instantaneous operating data exceeds the preset allowable range.

[0077] The instantaneous operating data includes at least one of the following: photovoltaic instantaneous power, energy storage instantaneous charging and discharging power, heat pump instantaneous power, load instantaneous power, grid connection point frequency detection data, and grid connection point voltage detection data.

[0078] In one possible implementation, the processing module is also configured to determine the type and degree of abnormal disturbances in the multi-energy system based on instantaneous operating data.

[0079] The processing module is also used to determine the target compensation strategy for the multi-energy system based on the type and degree of abnormal disturbance.

[0080] The processing module is also used to correct the intraday time-sharing scheduling data according to the target compensation strategy, obtain the corrected time-sharing scheduling data, and control the multi-energy system to operate according to the corrected time-sharing scheduling data within the target preset period.

[0081] The target compensation strategy includes at least one of the following: energy storage compensation control strategy, heat pump power regulation control strategy, and load reduction control strategy.

[0082] Fourthly, embodiments of this application provide a control device for a multi-energy system, applied to an energy coordination unit of a multi-energy system, comprising:

[0083] The acquisition module is used to acquire historical multi-source heterogeneous data sent by the edge energy management unit.

[0084] The processing module is used to input historical multi-source heterogeneous data into the cloud-based prediction model to obtain day-ahead prediction data for multi-energy systems; the day-ahead prediction data includes heat load prediction data.

[0085] The processing module is also used to perform thermoelectric coupling scheduling optimization based on heat load forecast data, with the optimization direction of preset multi-objective functions as the optimization direction, to obtain daytime energy supply strategies corresponding to multiple preset periods.

[0086] The processing module is also used to send day-ahead energy supply strategies to the edge energy management unit.

[0087] In one possible implementation, the processing module is also used to extract photovoltaic power output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data from historical multi-source heterogeneous data.

[0088] The processing module is also used to determine the data types of photovoltaic output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data, and to match the target modeling algorithm corresponding to any data type.

[0089] The processing module is also used to perform differentiated training on historical multi-source heterogeneous data using target modeling algorithms corresponding to different data types, and obtain multiple prediction branches.

[0090] The processing module is also used to integrate multiple prediction branches to obtain a cloud-based prediction model;

[0091] Among them, the photovoltaic power generation prediction branch was trained using a long short-term memory network model; the building thermal environment prediction branch was trained using an RC building thermal environment model; the domestic hot water load prediction branch was trained using an ARIMA model; the household appliance load prediction branch was trained using a random forest model; and the heat pump performance prediction branch was trained using an SVR model.

[0092] In one possible implementation, the processing module is further configured to perform knowledge distillation on the cloud-based prediction model to obtain a lightweight edge prediction model.

[0093] The processing module is also used to send the edge prediction model to the edge energy management unit.

[0094] Fifthly, embodiments of this application provide a multi-energy system, including an edge energy management unit and an energy coordination unit;

[0095] The edge energy management unit is deployed on the user side of the multi-energy system; the energy coordination unit is deployed on the service side of the multi-energy system; the energy coordination unit communicates with the edge energy management unit.

[0096] The edge energy management unit is used to perform the first aspect and / or various possible implementations of the first aspect as described above; the energy coordination unit is used to perform the second aspect and / or various possible implementations of the second aspect.

[0097] Sixthly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0098] The memory stores the instructions that the computer executes;

[0099] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect or the second aspect and / or various possible implementations of the second aspect.

[0100] In a seventh aspect, 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, or the second aspect and / or various possible implementations of the second aspect.

[0101] Eighthly, 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, or the second aspect and / or various possible implementations of the second aspect.

[0102] The multi-energy system and its control method provided in this application embodiment combine day-ahead energy dispatching strategies, multi-source heterogeneous data, current time and instantaneous operating data at the user side through an edge energy management unit to determine a target preset period and form a real-time load forecast result matching the current operating conditions. This method can specifically correct the day-ahead time-sharing dispatching data corresponding to the target preset period and control the coordinated operation of photovoltaics, energy storage, heat pumps, and home appliances when no abnormal operating conditions are triggered. This reduces the dependence of multi-energy system coordinated control on continuous cloud communication and command issuance, improving the real-time performance, accuracy, and operational stability of dispatching under dynamic operating conditions. This method solves the problems of existing home energy management systems, such as strong reliance on centralized cloud dispatching, insufficient connection between day-ahead planning and intraday operation, and easy accumulation of prediction errors, leading to poor real-time performance, stability, and economy. Simultaneously, it enhances the local autonomous control capability and responsiveness to dynamic operating conditions of the home multi-energy system under multi-timescale coordination. Attached Figure Description

[0103] 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.

[0104] Figure 1 A flowchart illustrating the control method for a multi-energy system on the edge energy management unit side provided in this application embodiment;

[0105] Figure 2 This is a schematic diagram of the system architecture of the energy coordination unit and the edge energy management unit in a multi-energy system provided in the embodiments of this application;

[0106] Figure 3 A schematic diagram of signaling interaction for a control method of a multi-energy system provided in an embodiment of this application;

[0107] Figure 4 Schematic diagram of the structure of the control device for the multi-energy system provided in the embodiments of this application Figure 1 ;

[0108] Figure 5 Schematic diagram of the structure of the control device for the multi-energy system provided in the embodiments of this application Figure 2 ;

[0109] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0110] 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

[0111] 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.

[0112] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0113] First, let me explain the terms used in this application:

[0114] Long Short-Term Memory (LSTM) network model: This LSTM model is an improved variant of recurrent neural networks. Through the gating structure of memory cells, forget gates, input gates, and output gates, it solves the long-term gradient vanishing problem of traditional recurrent neural networks and can accurately capture the temporal dependencies of long-term time-series data.

[0115] RC building thermal environment model: Resistance-Capacitance, abbreviated as RC; This RC building thermal environment model equates the building envelope, indoor air and other components to series and parallel circuits of thermal resistance and thermal capacity. It simulates the dynamic changes of indoor temperature and heat flow through heat conduction differential equations, and can achieve accurate prediction of building thermal environment without a large amount of training data.

[0116] ARIMA model: Autoregressive Integrated Moving Average, abbreviated as ARIMA; This ARIMA model transforms non-stationary time series into stationary series through differencing operations, and combines autoregressive and moving average terms to fit the historical changes in the data, thus completing a classic statistical model for short-term time series trend prediction.

[0117] Random forest model: It generates a large number of independent decision trees by randomly sampling samples and features, and outputs the final prediction value based on the voting results of all trees. It has strong anti-overfitting ability and can efficiently handle nonlinear prediction problems with multiple features coupled, such as household appliance load.

[0118] SVR model: Support Vector Regression, abbreviated as SVR; The SVR model maps low-dimensional nonlinear data to a high-dimensional space through a kernel function, constructs an optimal hyperplane to minimize the prediction error, and has high prediction accuracy in small-sample, nonlinear heat pump performance parameter prediction scenarios.

[0119] Most existing home energy management solutions adopt a cloud-based centralized dispatch model. The cloud uses historical energy consumption data, environmental data, and equipment status data to predict load and output, forming a daily time-of-use dispatch plan. Then, it sends control commands to user-side equipment to make photovoltaic, energy storage, heat pumps, and home appliances operate according to the predetermined strategy.

[0120] The above solutions can achieve unified management under static operating conditions, but they are highly dependent on the communication link and the continuous online status of the cloud. Furthermore, the local device mainly undertakes data acquisition and command execution functions, and when there are sudden changes in photovoltaic output, temporary changes in load, or increased network latency, it is difficult for the user side to make timely corrective decisions.

[0121] Meanwhile, existing scheduling schemes typically use the forecast results from the previous moment as the basis for subsequent control. There is a lack of close connection between the day-ahead plan and the intraday operation. The forecast error will continue to accumulate during the execution process, which can easily lead to the control results deviating from the actual load demand, resulting in unreasonable energy storage charging and discharging arrangements and inaccurate timing of heat pump operation, thereby affecting the real-time performance, stability and economic operation of the system.

[0122] Therefore, how to reduce the dependence of home energy system collaborative control on communication and improve the real-time performance, accuracy and stability of scheduling under dynamic operating conditions has become an urgent technical problem to be solved.

[0123] To address the aforementioned issues, this application provides a control method for a multi-energy system, applicable to a multi-energy system comprising an edge energy management unit and an energy coordination unit. By employing a control approach combining cloud-based training and user-side inference, the predictive capability is retained locally in the edge energy management unit. Upon receiving the day-ahead energy scheduling strategy from the energy coordination unit, the edge management unit combines collected multi-source heterogeneous data, current time, and instantaneous operational data. Based on the current time, it determines a target preset period, inputs the multi-source heterogeneous data into the edge prediction model to obtain real-time load prediction data, and compares this real-time load prediction data with the day-ahead load prediction data to correct the corresponding day-ahead time-of-day scheduling data, forming intraday time-of-day scheduling data. When the instantaneous operational data does not trigger abnormal operating conditions, the multi-energy system is controlled to operate according to the corrected scheduling data, thereby reducing communication dependence and improving the real-time performance, accuracy, and stability of scheduling.

[0124] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0125] Figure 1 This is a flowchart illustrating the control method for a multi-energy system provided in an embodiment of this application. The executing entity in this embodiment can, for example, be an edge energy management unit within the multi-energy system. Figure 1 As shown, the control method for a multi-energy system provided in this embodiment includes:

[0126] S101: In response to the day-ahead energy dispatch strategy generated by the multi-energy system, acquire multi-source heterogeneous data collected by the edge energy management unit, and acquire the current time and instantaneous operation data of the multi-energy system.

[0127] The day-ahead energy dispatch strategy includes day-ahead time-sharing dispatch data corresponding to multiple preset cycles; this day-ahead energy dispatch strategy is used to characterize the time-sharing quantitative dispatch data of multiple energy systems generated for each preset cycle within a single day, based on the predicted operating conditions.

[0128] In this embodiment, the multi-energy system is deployed on the user side as a controlled object, and may specifically include at least one of a power generation unit, an energy storage unit, an energy-consuming device, and a controllable load; the edge energy management unit in the multi-energy system is deployed on the user side as the execution subject of this embodiment, and the edge energy management unit is communicatively / electrically connected to each device in the multi-energy system; at the same time, the edge energy management unit is communicatively connected to the energy coordination unit deployed on the service side, and performs bidirectional data transmission with the energy coordination unit.

[0129] Specifically, the multi-energy system includes a heat pump tri-generation system, a photovoltaic system, and an energy storage system. The system is equipped with an edge energy management unit and an energy coordination unit. The edge energy management unit is deployed on the user side of the multi-energy system, and the energy coordination unit is deployed on the service side. The edge energy management unit is communicatively / electrically connected to the heat pump tri-generation system, the photovoltaic system, and the energy storage system, respectively. The heat pump tri-generation system is an energy-consuming device capable of simultaneously providing cooling, heating, and domestic hot water production, and is an energy-consuming load within the multi-energy system. Its operating principle is to consume electrical energy and output cooling and heating to meet building temperature control and domestic hot water supply needs.

[0130] The edge energy management unit receives the day-ahead energy supply strategy of the combined heat and power system from the energy coordination unit, and uploads the system operation data collected daily to the energy coordination unit so that the energy coordination unit can complete the iterative optimization of the prediction model deployed in the cloud.

[0131] Understandably, the day-ahead energy dispatch strategy refers to the time-sharing quantitative dispatch data of multiple energy systems for each preset cycle within a single day, generated by the energy coordination unit on the service side based on the predicted operating conditions on the day before the multi-energy system's operating day (day-ahead period). This day-ahead energy dispatch strategy includes the day-ahead energy supply strategy for the combined heat and power (CHP) system, the day-ahead consumption and allocation strategy for the photovoltaic system, and the day-ahead charge and discharge timing strategy for the energy storage system. Among them, the day-ahead energy supply strategy for the CHP system is generated by the energy coordination unit on the service side on the day before the target operating day. The day-ahead consumption and allocation strategy for the photovoltaic system and the day-ahead charge and discharge timing strategy for the energy storage system are obtained by the edge energy management unit after receiving the day-ahead energy supply strategy of the CHP system, using the actual power usage constraints corresponding to the strategy as input constraints, and combining the locally acquired electricity consumption forecast data.

[0132] For example, the day-ahead energy supply strategy of the heat pump tri-generation system is generated by the energy coordination unit on the service side the day before the target operating day. This strategy is a time-division quantified heat energy supply data divided into preset periods (e.g., 15 minutes) for the entire 24 hours of the target operating day, representing the heat-side output demand that the heat pump tri-generation system needs to meet under each preset period. The energy coordination unit then distributes the day-ahead energy supply strategy of the heat pump to the edge energy management unit on the user side.

[0133] After receiving the day-ahead energy supply strategy of the combined heat pump and tri-generation system, the edge energy management unit uses the actual power usage constraint corresponding to the strategy as the input constraint, and combines it with the locally acquired prediction data to solve for the day-ahead consumption allocation strategy of the photovoltaic system and the day-ahead charge and discharge timing strategy of the energy storage system, thereby obtaining a complete day-ahead energy dispatch strategy.

[0134] Meanwhile, the complete day-ahead energy dispatch strategy is stored locally to form a time-sharing strategy table arranged along the time axis of the target operating day, which serves as the benchmark for intraday rolling correction and real-time control on the target operating day. Multiple preset periods in this time-sharing strategy table can be divided according to preset time lengths, and each preset period corresponds to day-ahead time-sharing dispatch data related to the operation of multiple energy systems.

[0135] Understandably, multi-source heterogeneous data is used to support the predictive inference of subsequent edge prediction models. Its sources may include local storage data of edge energy management units, real-time collected data, and external environmental data. Its data content may include at least one of the following: historical load data, historical output data of distributed energy resources, environmental data, user setting data, energy storage status data, equipment operation status data, and electricity price information.

[0136] The current time is used to map the operating status of the multi-energy system to a specific location within a single day. The edge energy management unit can obtain the current time and determine the time period in which the operation is taking place by combining the period boundaries in the strategy table. Instantaneous operating data is used to characterize the real-time status of different devices and energy systems within the current system.

[0137] The edge energy management unit collects the above data and uses the collection results as the input basis for subsequent periodic positioning, load inference, scheduling correction and anomaly judgment.

[0138] In one possible implementation, the edge energy management unit can preprocess the multi-source heterogeneous data after receiving it and organize it into feature data that can be read by the model according to the target prediction window.

[0139] In this embodiment, the multi-source heterogeneous data and instantaneous operation data corresponding to the photovoltaic, energy storage, and heat pump combined supply system and environmental monitoring points are continuously collected and cached in the local data buffer by the acquisition module of the edge energy management unit; the data acquisition process at the hardware level is always running and is not controlled by the scheduling steps.

[0140] For example, the preset period can be 15 minutes. After the day-ahead energy dispatch strategy is successfully generated, the edge energy management unit performs the data acquisition step at the beginning of each preset period. Among them, the multi-source heterogeneous data are the historical observation data of various types of devices and environment collected by the edge energy management unit through the local bus and sensor interface. The current time is obtained by the edge energy management unit reading the local system clock to determine the corresponding target preset period. The instantaneous operation data are the real-time measurement data of the photovoltaic, energy storage, heat pump tri-generation system and grid connection point, which are used to characterize the current actual operating conditions of the system and provide input for subsequent model inference, disturbance identification and strategy correction.

[0141] Among them, multi-source heterogeneous data can provide full-dimensional data support for subsequent global data processing, model training, and time-of-use energy allocation strategy generation for different energy systems such as photovoltaic power generation systems, energy storage battery packs, and combined heat and power systems in home energy management systems, including real-time operating status, equipment parameters, environmental perception information, user energy demand, and grid interaction characteristics.

[0142] Understandably, this step incorporates both day-ahead energy dispatch strategies and real-time data collected from the user side into the edge-side processing flow, enabling subsequent control to no longer rely solely on a single day-ahead forecast result, but instead establish a local correction basis for the current operating state.

[0143] In this embodiment, the photovoltaic energy dispatch strategy is a power supply-side strategy, used to characterize the photovoltaic power consumption and distribution scheme; the energy storage energy dispatch strategy is a bidirectional power supply-side strategy, used to characterize the charge and discharge sequence scheme of the energy storage battery pack in each preset cycle; the heat pump energy dispatch strategy is a load-side strategy, used to characterize the constraint scheme of the allowable power consumption of the heat pump combined cycle power system in each preset cycle; wherein, the heat pump energy dispatch strategy does not refer to the heat pump distributing and dispatching energy, but rather the power consumption constraint obtained by the energy dispatch acting on the load side.

[0144] The energy coordination unit and the edge energy management unit are logically functional units that cooperate and collaborate equally. They can be deployed on independent hardware platforms; there is no master-slave relationship between them. The energy coordination unit focuses on global data processing, model training, lightweight model generation, and the generation of next-day energy dispatch predictions for the combined heat and power (CHP) system. It provides the edge energy management unit with the prediction model and the energy dispatch strategy for the CHP system, but does not output specific equipment control commands. The edge energy management unit is deployed in scenarios where multiple energy systems operate collaboratively, such as in residential or industrial settings, and independently completes local real-time planning, periodic strategy correction, and equipment management.

[0145] Meanwhile, the time slice used in the next-day energy scheduling strategy of the heat pump tri-generation system generated by the energy coordination unit is consistent with the preset cycle granularity of the edge energy management unit; in specific implementation, the energy coordination unit can be deployed on a cloud server or a private remote server.

[0146] S102: Determine the target preset period corresponding to the current time, as well as the day-ahead load forecast data within the target preset period, and input the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system.

[0147] In this embodiment of the application, when the currently executing preset period is about to end, the scheduling strategy optimization logic within the corresponding preset period is triggered, and the data acquisition in step S101 is executed; and the data acquisition refers to the algorithm level calling and reading the locally cached actual test data, that is, extracting the actual multi-source heterogeneous data and instantaneous running data accumulated within the current preset period, for the edge prediction model to perform inference calculations, and to solve the scheduling strategy for the next target preset period.

[0148] Understandably, the target preset period refers to the next preset period of the current operating cycle; the day-ahead load forecast data is the load baseline value used when the day-ahead plan is generated, and it is the load forecast data generated using the corresponding forecast model; the edge forecast model is deployed in the user-side edge energy management unit. This edge forecast model is trained and distributed by the service-side energy coordination unit, and the edge energy management unit loads the edge forecast model to complete the forecast inference calculation locally; the real-time load forecast data is used to reflect the reassessment of the load demand within the target preset period under the current conditions.

[0149] After receiving and preprocessing the data, the edge energy management unit determines the target preset period based on the correspondence between the current time and the daily time-of-use table, and reads the day-ahead load forecast data associated with the target preset period from the day-ahead energy dispatch strategy.

[0150] The edge prediction model receives processed multi-source heterogeneous data and outputs real-time load prediction data corresponding to the target preset period.

[0151] For example, if "00:00-00:15" is the first preset period, data acquisition of S101 is triggered at 00:12 (3 minutes before the end of the current period, the period is nearing its end); at this time, the current time is within the first preset period, and the calculated target preset period is the subsequent second preset period (00:15-00:30). The daytime load forecast data associated with the target preset period is read from the daytime energy dispatch strategy, and the real-time load forecast data corresponding to the target preset period is obtained using the edge prediction model.

[0152] Understandably, this step involves locating the current moment to a specific target preset period, extracting the day-ahead load forecast benchmark within that period, and then using a local edge prediction model to combine real-time multi-source heterogeneous data for inference to form real-time load forecast data corresponding to the current operating conditions, thereby establishing a dynamic connection between the day-ahead plan and intraday operation.

[0153] S103: Based on real-time load forecast data and day-ahead load forecast data, the day-ahead time-sharing scheduling data corresponding to the target preset period is corrected to obtain intraday time-sharing scheduling data.

[0154] After obtaining real-time load forecast data and day-ahead load forecast data, the edge energy management unit can adjust the day-ahead time-of-day scheduling data for the target preset period based on the difference between the two, so as to obtain intraday time-of-day scheduling data that is compatible with the current operating status.

[0155] In one possible implementation, after obtaining real-time load forecast data and day-ahead load forecast data, the edge energy management unit calculates the forecast deviation between the real-time load forecast data and the day-ahead load forecast data; within the operating constraints of each preset device in the multi-energy system, it adaptively adjusts the power supply side scheduling parameters in the day-ahead time-of-use scheduling data corresponding to the target preset period to obtain intraday time-of-use scheduling data adapted to the actual operating conditions of the multi-energy system.

[0156] Among them, the power constraint of the heat pump on the load side remains unchanged, and only the scheduling parameters of the photovoltaic system and the energy storage system are adjusted.

[0157] For example: if the real-time load forecast is higher than the day-ahead load forecast, it indicates that the actual energy demand corresponding to the target preset cycle has increased. At this time, it is necessary to prioritize increasing the local consumption scale of photovoltaic power. When the photovoltaic output is insufficient to make up for the corresponding load gap, it is compensated by adjusting the charging and discharging sequence of energy storage. All adjustment results do not exceed the active power regulation range of the photovoltaic inverter, the charging and discharging power threshold of the energy storage battery, and the safe state of charge range.

[0158] If the real-time load forecast is lower than the day-ahead load forecast, it indicates that the actual energy demand corresponding to the target preset cycle has decreased. In this case, the local photovoltaic power consumption will be reduced first, and the surplus power will be allocated to the energy storage battery for charging. When the energy storage battery can no longer be charged, the photovoltaic output power will be reduced.

[0159] Based on the above adaptive adjustments, intraday time-sharing scheduling data corresponding to the target preset cycle is generated for subsequent control of the multi-energy system.

[0160] In this embodiment, the correction process satisfies system operation constraints and device constraints. Intraday time-of-use scheduling data is thus generated, which is essentially a local execution plan for a target preset period. The edge energy management unit can repeatedly execute this correction process in a periodic rolling manner; that is, it re-acquires day-ahead load forecast data and the latest multi-source heterogeneous data each time a new target preset period begins, thereby continuously generating updated intraday time-of-use scheduling data.

[0161] Understandably, this step compares and corrects real-time load forecast data with day-ahead load forecast data, transforming the execution plan from the original forecast plan into an intraday plan that matches the current operating status. Forecast errors do not accumulate directly along subsequent control processes, but are reabsorbed into the scheduling results in each target preset cycle.

[0162] At the same time, this correction process enables multi-energy systems to adapt to photovoltaic fluctuations and load changes on a minute-level timescale, significantly improving the real-time performance and accuracy of energy dispatch.

[0163] S104: If the instantaneous operating data does not meet the abnormal operating conditions, control the multi-energy system to operate according to the intraday time-sharing scheduling data within the target preset period.

[0164] Understandably, abnormal operating conditions refer to various operating conditions during the operation of a multi-energy system, such as when the equipment's measuring points, electrical parameters, or thermal parameters exceed the equipment's preset safe operating thresholds, or when the equipment malfunctions or the communication link becomes abnormal.

[0165] In this embodiment of the application, by determining whether the instantaneous operating data meets the abnormal operating conditions, it can be determined whether the current multi-energy system is in an abnormal operating state. The abnormal operating state refers to the working state of at least one controlled device or communication link in the multi-energy system after triggering the above-mentioned abnormal operating conditions. In this state, a secondary adaptation and correction of the scheduling data is triggered to generate control instructions adapted to the current abnormal operating conditions, thereby realizing safe control under abnormal operating conditions.

[0166] Before outputting control commands corresponding to the target preset cycle, the edge energy management unit first uses instantaneous operating data to determine whether the current multi-energy system is in an abnormal operating state.

[0167] When the instantaneous operating data does not meet the abnormal operating conditions, it indicates that the system is in an executable state. At this time, the intraday time-sharing scheduling data is converted into a set of control instructions for the relevant equipment to control the multi-energy system to operate according to the intraday time-sharing scheduling data within the target preset period.

[0168] When the instantaneous operating data meets the abnormal operating conditions, it indicates that the system is not in an executable state. At this time, it is necessary to adapt and correct the currently obtained intraday time-sharing scheduling data to generate control instructions adapted to the current abnormal operating conditions, so as to control the multi-energy system to operate according to the second-corrected intraday time-sharing scheduling data when it is in an abnormal operating condition within the target preset period.

[0169] For example, abnormal operating states include abnormal states of the equipment itself, abnormal states of the electrical grid connection point, abnormal states of communication, and abnormal states of thermal operation; among them, abnormal states of photovoltaic systems include: "photovoltaic inverter fault alarm, sudden change in output power, overvoltage, overcurrent"; abnormal states of energy storage systems include: "battery overcharge, over-discharge, SOC (State of Charge) exceeding the safe upper and lower limits, charging and discharging power exceeding the limit, and BMS (Battery Management System) reporting a fault"; abnormal states of heat pump tri-generation systems include: "compressor fault, outlet water temperature exceeding the limit, and equipment reporting a fault code"; abnormal states of grid connection point electrical systems include: "grid voltage exceeding the limit, frequency exceeding the limit, and islanding alarm"; abnormal states of communication links include: "interruption of local bus communication between the edge energy management unit and photovoltaic, energy storage, and heat pump, and inability to read measurement points or issue commands".

[0170] In one possible implementation, after issuing control commands, the edge energy management unit can continuously monitor the execution results of instantaneous operating data and corresponding intraday time-sharing scheduling data; if the instantaneous operating data during the execution process continuously meets the abnormal operating conditions for a preset duration, the current normal execution process will be exited.

[0171] Understandably, this detection process runs through the entire target preset cycle; that is, the actual operating condition of the system is continuously detected before the current cycle ends and before the scheduling optimization of the next target preset cycle is triggered.

[0172] If the instantaneous operating data within the preset time period continuously meets the abnormal operating conditions, it indicates that there is a continuous abnormal operating state within the preset time period corresponding to the current target preset cycle. The abnormal operating condition of the system cannot be eliminated by the energy distribution control of the multi-energy system alone. At this time, the current normal execution process is exited and the preset fault protection action is triggered.

[0173] If instantaneous operating data within a preset time period is detected, and there are at least a preset number (e.g., 5) instantaneous moments that meet the abnormal operating conditions, it indicates that the detected event is an instantaneous disturbance or short-term operating condition fluctuation, not a fault in the equipment itself. In this case, the disturbance can be mitigated by controlling the energy distribution of the photovoltaic system and the energy storage system without triggering the fault protection and handling process. The system maintains the current routine execution process and continues to execute the control commands corresponding to the intraday time-sharing scheduling data.

[0174] For example, the preset duration can be set to a range of 30s-3min to filter out sensor acquisition glitches and instantaneous electrical disturbances, and to avoid false triggering of fault protection.

[0175] Abnormal operating conditions may include at least one of the following: the state of charge of the energy storage battery exceeds the safe range, the photovoltaic inverter has a fault alarm, the thermal parameters of the combined heat and power system exceed the limits, the voltage / frequency of the grid connection point exceeds the limits, and the communication between the edge energy management unit and the controlled equipment is interrupted.

[0176] The preset fault protection actions include shutdown, power-limited operation, and fault alarm.

[0177] Understandably, this step translates the corrected intraday time-sharing scheduling data into the actual operational behavior of the multi-energy system, thereby forming a complete edge closed loop from policy reception, load reassessment, scheduling correction to equipment execution.

[0178] In one possible implementation, the multi-source heterogeneous data collected by the edge energy management unit on the same day is periodically transmitted back to the energy coordination unit deployed on the service side.

[0179] Understandably, the multi-source heterogeneous data transmitted back to the energy coordinating unit deployed on the service side includes operational measurement data from photovoltaic systems, energy storage systems, and combined heat and power (CHP) systems, as well as environmental monitoring data and load operation data.

[0180] For example, the multi-source heterogeneous data transmitted back to the energy collaboration unit deployed on the service side includes the actual output power of photovoltaics, the charging and discharging power of energy storage batteries, the SOC reported by the BMS, the operating power of the heat pump, the building's heating and cooling load, and the ambient temperature.

[0181] In one possible implementation, the edge energy management unit uploads the locally cached and organized operating data to the energy coordination unit according to a preset backhaul cycle; the backhaul cycle can be set on a daily basis, i.e., data is reported once a day, or it can be set to an hourly reporting cycle.

[0182] In this embodiment, the edge energy management unit incorporates multi-source heterogeneous data, day-ahead load forecast data, and instantaneous operation data into a local closed-loop control process, using the target preset period corresponding to the current moment as the granularity. This enables timely correction of the day-ahead time-sharing scheduling data for the target preset period and control of the multi-energy system operation. Therefore, the control basis no longer rests on a single day-ahead forecast result, but rather regenerates the execution plan within each target preset period based on real-time operating conditions, reducing reliance on continuous online and low-latency external communication links.

[0183] The control method for a multi-energy system provided in this embodiment, in response to the day-ahead energy dispatch strategy generated by the multi-energy system, acquires multi-source heterogeneous data collected by the edge energy management unit, and obtains the current moment and instantaneous operating data of the multi-energy system; determines the target preset period corresponding to the current moment, and the day-ahead load forecast data within the target preset period, and inputs the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system; based on the real-time load forecast data and the day-ahead load forecast data, corrects the day-ahead time-sharing dispatch data corresponding to the target preset period to obtain intraday time-sharing dispatch data; when the instantaneous operating data does not meet the abnormal operating conditions, the multi-energy system is controlled to operate according to the intraday time-sharing dispatch data within the target preset period. This method reduces the dependence of multi-energy system collaborative control on continuous cloud communication and command issuance, improves the local autonomous control capability and response capability to dynamic operating conditions of the home multi-energy system under multi-timescale collaboration, and improves the real-time performance, accuracy, and stability of energy dispatch under dynamic operating conditions.

[0184] The control method for a multi-energy system in this application is applied to the coordinated optimization scheduling and control of a photovoltaic system, an energy storage system, and a combined heat pump system in a home energy management scenario. Specifically, the multi-energy system includes an energy coordination unit on the service side and an edge energy management unit deployed on the home energy user side. The two types of units communicate with each other and cooperate to jointly realize the intelligent optimization control of the home multi-energy system.

[0185] Figure 2 This is a schematic diagram of the system architecture of the energy coordination unit and the edge energy management unit in a multi-energy system provided in an embodiment of this application. Figure 2 As shown, the energy coordination unit is deployed on the cloud server 201 and has the capabilities of global data statistics, model training iteration, strategy generation and model distribution; the edge energy management unit is deployed on the local server 202 on the home energy user side and can independently complete local data collection, edge model prediction, strategy correction and device closed-loop control.

[0186] In other words, the energy coordination unit is used to train the prediction model based on global multi-energy data, generate a lightweight prediction model, and distribute the lightweight prediction model to the edge energy management unit; the edge energy management unit is used to receive the lightweight prediction model, use the lightweight prediction model to perform local real-time planning, output the controllable parameters of the multi-energy system, and complete the photovoltaic-storage-thermal coupling optimization control.

[0187] For example, the controlled objects of this multi-energy system control method are various energy-consuming and energy-supplying devices in the system, specifically including photovoltaic systems, energy storage battery packs, and combined heat and power systems; at the same time, it can also link with terminal energy-consuming devices such as smart home appliances to form a complete home photovoltaic-storage-thermal coupling multi-energy application system.

[0188] In one possible implementation, the edge energy management unit further integrates an AI energy management robot that operates in conjunction with energy management strategies through task scheduling algorithms.

[0189] Specifically, the AI ​​robot dynamically plans the execution time and path of its own tasks (such as inspection and cleaning) based on current photovoltaic output forecast data, energy storage SOC, and grid electricity price signals. For example, when photovoltaic output is sufficient and energy storage SOC is high, the AI ​​robot prioritizes high-energy-consuming tasks (such as deep cleaning) and uses surplus photovoltaic power to charge it; when photovoltaic output is insufficient and grid electricity price is at its peak, the AI ​​robot prioritizes low-energy-consuming tasks (such as basic inspection) and reduces energy storage discharge to lower electricity purchase costs.

[0190] Understandably, this task scheduling strategy is linked with the real-time load forecast data of the edge energy management unit, and uses a multi-objective optimization algorithm to achieve dynamic matching between operational energy consumption and system energy supply and demand, thereby further improving the overall energy efficiency of the system.

[0191] Figure 3 This is a schematic diagram of the signaling interaction of the control method for a multi-energy system provided in an embodiment of this application. Figure 3 As shown, in this embodiment... Figure 1 Based on the single-sided control embodiment of the edge energy management unit shown, combined with Figure 2 The system architecture shown further details the bidirectional signaling interaction logic and execution process between the edge energy management unit and the energy coordination unit, and provides a detailed explanation of the control method for the multi-energy system. The control method for the multi-energy system shown in this embodiment includes:

[0192] S301: The energy coordination unit acquires historical multi-source heterogeneous data sent by the edge energy management unit.

[0193] The main entity executing this step is the energy coordination unit. As a collaborative control node on the service side, the energy coordination unit establishes a data communication connection with the edge energy management unit deployed on the residential or park side, and receives historical multi-source heterogeneous data periodically uploaded or triggered by tasks by the edge energy management unit.

[0194] In this embodiment, the multi-energy system is used to carry the control method of this application. During operation, each device generates interconnected thermal and electrical data. The edge energy management unit is the management node of the edge side (i.e., the user side) of the multi-energy system. It is responsible for collecting on-site equipment operation information, control execution information and environmental correlation information, and sending them to the service side according to the preset interface protocol.

[0195] Understandably, historical multi-source heterogeneous data refers to data sets with different sources, sampling frequencies, and data structures. It includes at least historical operating data related to heat load forecasting, historical status data related to the operating status of multiple energy systems, and historical control data related to the generation of day-ahead energy supply strategies.

[0196] In practice, the edge energy management unit can send time-organized data to the energy coordination unit through an interface or communication link. After receiving the data, the energy coordination unit can preprocess the historical multi-source heterogeneous data and form a data set that can be called later.

[0197] Understandably, this step completes the process of data collection and transmission from the edge side, and data reception and preparation on the server side, providing an input basis for subsequent cloud-based prediction models and enabling a unified expression of the temporal correlation between thermal load, electrical load, and equipment operating status.

[0198] S302: The energy coordination unit inputs historical multi-source heterogeneous data into the cloud prediction model to obtain day-ahead prediction data for multiple energy systems.

[0199] The day-ahead forecast data includes heat load forecast data.

[0200] Understandably, the cloud-based prediction model is deployed in the server-side computing environment where the energy coordination unit is located, and is used to transform historical multi-source heterogeneous data into day-ahead prediction results for multiple preset future periods.

[0201] The cloud-based forecasting model is used to output day-ahead forecast data based on historical multi-source heterogeneous data. Among them, heat load forecast data is a component of the day-ahead forecast data, which is used to characterize the changing trend of heat demand on the residential side or the park side in each preset period in the future. At the same time, the heat load forecast data is used to support the generation of day-ahead scheduling strategy for the cloud-based heat pump tri-generation system, and to provide a benchmark for load-side power consumption constraints.

[0202] In this embodiment of the application, the day-ahead forecast data may include other forecast results related to subsequent thermoelectric coupling scheduling, in addition to the heat load forecast data.

[0203] It should be noted that the prediction data of the multi-energy system control method of this application includes two dimensions: day-ahead prediction and intraday real-time prediction. The prediction data involved in the entire process is not limited to heat load prediction data. Specifically, the prediction data includes heat load prediction data, photovoltaic output prediction data, and electrical load prediction data. Among them, the energy coordination unit deployed on the cloud server 201 completes the day-ahead heat load prediction, thereby generating the baseline scheduling strategy of the combined heat and power system. Meanwhile, the user-side edge energy management unit relies on the local lightweight edge prediction model to complete the real-time prediction of photovoltaic output and electrical load in the intraday real-time dimension, providing complete data support for the rolling correction of the intraday time-sharing scheduling strategy and the coordinated optimization and control of photovoltaic and energy storage, forming a two-layer prediction architecture of day-ahead baseline prediction and intraday precise correction.

[0204] For example, the energy coordination unit can extract prediction-related feature information from historical multi-source heterogeneous data and input it into the cloud prediction model for inference, and output the prediction values ​​for each preset period of the next day at a preset time granularity. The preset period can be 1 hour, 30 minutes or 15 minutes.

[0205] Heat load forecast data can be specifically represented as a series. ,in, Indicates the first Predicted heat load demand values ​​for a preset period This indicates the number of preset cycles covered by the current day's scheduling; this sequence is used to characterize the temporal changes in future heat demand.

[0206] Understandably, this step transforms the temporal variation patterns, equipment correlations, and control feedback information in historical multi-source heterogeneous data into day-ahead forecast data that can be directly used for scheduling. In particular, it enables the heat load forecast data to reflect the temporal changes in heat demand within multiple preset cycles, providing a quantitative basis for optimizing thermoelectric coupling.

[0207] S303: Based on heat load forecast data, the energy coordination unit performs thermoelectric coupling scheduling optimization with the optimization direction of preset multi-objective functions as the optimization direction, and obtains day-ahead energy supply strategies corresponding to multiple preset periods.

[0208] S304: The energy coordination unit sends the day-ahead energy supply strategy to the edge energy management unit.

[0209] Understandably, the energy coordination unit organizes the thermoelectric coupling scheduling solution process based on the heat load forecast data in the day-ahead forecast data, and jointly optimizes the heat and electricity resources in the multi-energy system.

[0210] Among them, the preset multi-objective function is a set of objective functions used in scheduling solution, which is used to uniformly express multiple optimization objectives, specifically including supply and demand satisfaction objectives, energy balance objectives, and operation coordination objectives.

[0211] Thermoelectric coupling scheduling is a coordinated scheduling process that simultaneously considers the thermal energy supply link and the electrical energy supply link. Its core lies in determining the operation arrangement and energy distribution relationship of different energy units within each preset cycle based on heat load forecast data and other day-ahead forecast data.

[0212] The current energy supply strategy outputs the results of this step, which are used to characterize the supply arrangements of the heat pump tri-generation system over multiple consecutive preset cycles.

[0213] In practical implementation, the energy coordination unit establishes a thermo-electric coupling scheduling model based on heat load forecast data and other day-ahead forecast data, and performs optimization under the direction of the optimal multi-objective function to obtain day-ahead energy supply strategies corresponding to multiple preset periods. The solution results can be output in the form of time-sharing strategy tables or other strategy data, and each preset period corresponds to a corresponding energy supply arrangement.

[0214] The current energy supply strategy, as a service-side scheduling result, includes at least the strategy application date, preset period index, and supply arrangements corresponding to each preset period. The edge energy management unit, as the strategy recipient, can convert the supply arrangements corresponding to each preset period into field equipment control plans or local execution references after receiving the strategy.

[0215] For example, the energy coordination unit can generate a structured file or message message for the day-ahead energy supply strategy and send it to the edge energy management unit by actively pushing or pulling from the edge side; after receiving it, the edge energy management unit saves the received day-ahead energy supply strategy for subsequent execution and invocation.

[0216] Understandably, this step introduces heat load forecast data into a multi-objective optimization process to achieve joint allocation of heat-side demand and electricity-side resources, enabling multiple preset periodic supply methods to be arranged executable under unified constraints, thereby forming a day-ahead energy supply strategy for the heat pump combined cycle power system with time continuity, and then transmitting this day-ahead energy supply strategy to the edge energy management unit.

[0217] At the same time, the scheduling results obtained from the day-ahead forecast and thermoelectric coupling optimization are transmitted to the edge side, so that the energy coordination unit and the edge energy management unit can form a connection from strategy generation to strategy distribution.

[0218] In one possible implementation, when the energy coordination unit deployed on the cloud server 201 generates day-ahead forecast data using the cloud forecast model, the day-ahead forecast data includes not only heat load forecast data, but also day-ahead photovoltaic output data and day-ahead electrical load forecast data on the server side. The day-ahead photovoltaic output data and day-ahead electrical load forecast data on the server side are used to assist in constructing a thermo-electric coupling scheduling model when performing thermo-electric coupling scheduling optimization, so as to solve the current day-ahead energy supply strategy.

[0219] In this embodiment, the energy coordination unit deployed on the cloud server 201 generates a daytime energy supply strategy for the combined heat and power (CHP) system based on daytime forecast data. This strategy is a baseline scheduling scheme for the CHP system with a preset daily cycle in a multi-energy system. The edge energy management unit deployed on the local server 202 generates a daytime energy scheduling strategy based on the daytime energy supply strategy and the daytime forecast data (daytime photovoltaic output data and daytime electricity load forecast data) generated on the user side. This strategy is a baseline scheduling scheme for the preset daily cycle in a multi-energy system. It is a top-level time-sharing control strategy that coordinates the operation of photovoltaic, energy storage, and CHP equipment, providing an initial baseline for subsequent intraday real-time rolling corrections.

[0220] Specifically, day-ahead energy dispatch strategies include day-ahead energy supply strategies for combined heat and power (CHP) systems, day-ahead energy consumption and distribution strategies for photovoltaic systems, and day-ahead charge and discharge timing strategies for energy storage systems. These three sub-strategies correspond to the equipment dispatch and control requirements of the load side, power supply side, and bidirectional power supply side of multi-energy systems, respectively.

[0221] Among them, the daytime energy supply strategy of the heat pump tri-generation system is a load-side constrained scheduling strategy, which is obtained by solving the daytime heat load forecast data output from the cloud. It is used to preset the allowable power consumption constraint range of the heat pump tri-generation system for each cycle of the day. This strategy does not adjust the energy distribution logic, but limits the upper limit of the operating power of the heat pump from the energy consumption end to match the energy demand of building heating and cooling and domestic hot water, so as to ensure the comfort of end-user energy and the stability of equipment operation.

[0222] The day-ahead consumption allocation strategy for photovoltaic systems is a power-side allocation dispatch strategy. It is obtained by combining day-ahead photovoltaic output forecast data and electricity load demand. It is used to plan the local self-consumption and surplus allocation scheme of photovoltaic power in each cycle of the day. The strategy aims to maximize local photovoltaic consumption and reduce grid purchase costs. It plans the output power flow and consumption power of the photovoltaic system in a time-sharing manner to adapt to the renewable energy consumption needs of residential users.

[0223] The day-ahead charge / discharge timing strategy of the energy storage system is a bidirectional power supply-side timing scheduling strategy. It is obtained by combining the day-ahead load fluctuation pattern and photovoltaic power output timing characteristics. It is used to preset the charge / discharge timing and power benchmark of the energy storage battery pack for each cycle of the day. This strategy takes into account peak and valley electricity prices, photovoltaic power output fluctuations, and load supply and demand differences, and rationally plans the timing nodes for energy storage charging and discharging to smooth system power fluctuations and optimize the economy and stability of multi-energy system operation.

[0224] In one possible implementation, the cloud-based prediction model includes a photovoltaic power generation prediction branch, a building thermal environment prediction branch, a domestic hot water load prediction branch, a household appliance load prediction branch, and a heat pump performance prediction branch. Training the cloud-based prediction model includes: extracting photovoltaic output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data from historical multi-source heterogeneous data; determining the data types of the photovoltaic output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data respectively, and matching the target modeling algorithm corresponding to each data type; using the target modeling algorithms corresponding to different data types to perform differentiated training on the historical multi-source heterogeneous data to obtain multiple prediction branches; and integrating the multiple prediction branches to obtain the cloud-based prediction model.

[0225] Among them, the photovoltaic power generation prediction branch was trained using an LSTM model; the building thermal environment prediction branch was trained using an RC building thermal environment model; the domestic hot water load prediction branch was trained using an ARIMA model; the household appliance load prediction branch was trained using a random forest model; and the heat pump performance prediction branch was trained using an SVR model.

[0226] In this embodiment, the RC building thermal environment model is used to predict changes in building interior temperature and heat load demand. Based on the thermal resistance-heat capacity principle, the model simulates the heat transfer characteristics of the building envelope by inputting parameters such as outdoor temperature, user-set room temperature, and heat pump output, generating an indoor temperature prediction curve. For example, when a sudden drop in outdoor temperature is detected, the model predicts that the indoor temperature will gradually decrease, and dynamically adjusts the heat pump power output based on the heat pump performance prediction results to ensure that the indoor temperature remains within the user-set comfort range. Furthermore, the model also supports moderately reducing heat pump power during peak electricity price periods, utilizing building thermal inertia to maintain temperature fluctuations, thereby reducing the need for high-priced electricity and improving system economy.

[0227] In practice, historical multi-source heterogeneous data can be formed by cloud server 201 receiving sampling records, equipment status records and environmental observation records uploaded by edge energy management unit, and then forming a training sample set after timestamp alignment, outlier removal and missing value completion.

[0228] For photovoltaic power output data, it can be constructed into a sequence input LSTM model according to a set time granularity to learn the output correlation caused by irradiance changes and weather disturbances; for building thermal environment parameters, indoor temperature, building envelope heat flux, and external meteorological parameters can be input into the RC building thermal environment model to characterize the dynamic response under the combined effects of thermal resistance and heat capacity; for domestic hot water usage data, an ARIMA model can be established based on historical water consumption and time period distribution to output the hot water load for the next prediction period; for household appliance load data, the historical power, start / stop status, and usage time of multiple household appliances can be input into a random forest model to obtain the corresponding load prediction results; for heat pump performance data, features such as ambient temperature, supply and return water temperature, and operating frequency can be input into an SVR model to estimate heat pump performance parameters.

[0229] After each prediction branch is trained, its outputs are merged to form a unified cloud prediction model for day-ahead scheduling and intraday correction.

[0230] In this embodiment, the output results of different prediction branches are integrated into unified real-time load prediction data through feature fusion algorithms (such as weighted average or attention mechanism). This integration method not only retains the modeling advantages of each branch for specific data types, but also improves the overall prediction accuracy through inter-model collaboration, providing a more reliable data foundation for scheduling correction.

[0231] Understandably, this cloud-based prediction model selects appropriate modeling algorithms for different data types, allowing each prediction object to complete training and output in an independent branch, and then integrates the results of each branch. This enables the cloud-based prediction model to simultaneously cover multiple types of operational information such as photovoltaic output, building thermal environment, domestic hot water, household appliance load, and heat pump performance, and to provide a consistent prediction basis for the coordinated energy scheduling of multi-energy systems.

[0232] By adopting the above method, different types of historical data can obtain training models that match their characteristics, the prediction accuracy and stability of branch outputs are improved, the cloud prediction model's adaptability to multi-source heterogeneous data is enhanced, and thus the integrity of the current-day prediction data and the reliability of the scheduling basis are improved.

[0233] In one possible implementation, after integrating multiple prediction branches to obtain a cloud prediction model, the cloud prediction model is subjected to knowledge distillation to obtain a lightweight edge prediction model; the edge prediction model is then sent to the edge energy management unit.

[0234] In practice, after integrating multiple prediction branches, the cloud-based prediction model retains the prediction capabilities of each branch for the corresponding data type and inputs the integrated prediction output as distillation source information into the distillation training module.

[0235] The distillation training module can run on a cloud server 210. It uses a pre-configured distillation loss function to iteratively optimize the student model. The distillation loss function consists of the true label loss and the distillation constraint loss, which is used to constrain the student model output to be close to the teacher model output.

[0236] After training, the resulting edge prediction model retains key parameters related to local inference and compresses unnecessary connections and redundant feature channels, making the model size adaptable to the storage and computing power conditions of the edge energy management unit.

[0237] When a lightweight edge prediction model is sent to the edge energy management unit, the cloud server 210 sends the model file, version number, and verification information to the edge side via a network communication link. After receiving the data, the edge energy management unit completes local model loading and cache updates, and then directly calls the edge prediction model to predict multi-source heterogeneous data collected in real time during subsequent operation.

[0238] Understandably, through the above method, the edge side can independently complete local predictions based on a lightweight model and form a basis for intraday correction in conjunction with the day-ahead scheduling data. This method combines the cloud prediction model training capability with the edge prediction model inference capability, reduces the computational burden on the edge side, and improves the prediction continuity and control response speed of the edge energy management unit in communication-constrained scenarios.

[0239] In one possible implementation, after the energy coordination unit receives the multi-source heterogeneous data (e.g., multi-source heterogeneous data collected within a single day) periodically returned by the edge energy management unit, it adds the newly added operational data to the dataset for iterative training and parameter updates of the prediction model in the cloud. After training is completed, the updated model is sent to the edge energy management unit, realizing a two-end data closed loop from edge data collection to iterative updates of the prediction model.

[0240] Understandably, based on this system architecture, the energy coordination unit can complete model training and lightweight processing based on global multi-source heterogeneous data, generate a lightweight model adapted to edge computing power, and send it to the local area; after receiving the lightweight model, the edge energy management unit can perform local real-time load forecasting and energy planning based on the local real-time multi-source heterogeneous data, dynamically optimize and output controllable parameters such as photovoltaic consumption, energy storage charging and discharging, and heat pump power consumption constraints, so as to realize refined, adaptive, and collaborative optimization control of the household multi-energy system.

[0241] S305: The edge energy management unit determines the power constraints corresponding to the day-ahead energy supply strategy.

[0242] S306: The edge energy management unit determines the historical multi-source heterogeneous data collected by the edge energy management unit, and inputs the historical multi-source heterogeneous data into the edge prediction model to obtain the day-ahead photovoltaic power output prediction data and the day-ahead electricity load prediction data.

[0243] S307: The edge energy management unit determines the operating status of the energy storage system, and based on the operating status, day-ahead photovoltaic output forecast data, day-ahead electrical load forecast data, and power constraints, it performs thermoelectric coupling scheduling optimization with the optimization direction of the preset multi-objective function as the optimization direction, and obtains day-ahead consumption allocation strategy and day-ahead charge and discharge timing strategy corresponding to multiple preset cycles.

[0244] Among them, the power constraint condition is used to characterize the available power range of the heat pump tri-generation system; the day-ahead energy supply strategy is a time-division quantified heat energy supply strategy of the heat pump tri-generation system under each preset cycle of a single day, generated based on predicted operating conditions; the day-ahead consumption and allocation strategy is a time-division quantified photovoltaic energy consumption and allocation strategy of the photovoltaic system under each preset cycle of a single day, generated based on predicted operating conditions; the day-ahead charge and discharge timing strategy is a time-division quantified charge and discharge timing allocation strategy of the energy storage system under each preset cycle of a single day, generated based on predicted operating conditions; the thermoelectric coupling scheduling optimization solution is implemented using at least one of the following: weighted summation multi-objective optimization algorithm, particle swarm optimization algorithm, and genetic algorithm.

[0245] In practical applications, the power constraints can be determined by the rated power of the heat pump unit, the start-stop threshold, and the temperature setpoint. The edge energy management unit then limits the upper and lower limits of the adjustable power of the heat pump tri-generation system in each preset cycle.

[0246] Historical multi-source heterogeneous data can consist of photovoltaic irradiance, ambient temperature, indoor and outdoor load records, and equipment operation logs, and after normalization and time alignment, it is input into the edge prediction model.

[0247] The operating status of an energy storage system can include the energy storage state of charge, charging and discharging power, and temperature status. When solving the problem, the edge energy management unit combines this status with the predicted data and power constraints and inputs them into the optimization model to output a time-sharing quantitative day-ahead consumption allocation strategy and a day-ahead charging and discharging timing strategy.

[0248] Understandably, thermoelectric coupling scheduling optimization refers to the process of optimizing the controllable parameters of the equipment in each cycle by taking the optimality of a preset multi-objective function as the optimization direction and combining the operating constraints of the photovoltaic and thermal energy storage equipment. The thermoelectric coupling scheduling optimization can be executed in the edge energy management unit, and its results are cached and distributed to the photovoltaic system, the combined heat and power system and the energy storage system in the form of control sequences corresponding to multiple preset cycles.

[0249] For example, thermoelectric coupling scheduling optimization is a constrained multi-objective optimization process for thermoelectric coupling multi-energy systems. The optimization sub-objectives are minimizing the overall system operating cost and maximizing the renewable energy absorption rate. The scheduling strategy is solved by a preset optimization algorithm.

[0250] The day-ahead energy supply strategy serves as the input benchmark for the combined heat pump and tri-generation system. Together with the corresponding power constraints, it constitutes the boundary conditions for thermoelectric coupling scheduling. The day-ahead photovoltaic output forecast data and day-ahead electrical load forecast data provided by the edge prediction model are used to characterize the energy supply and demand relationship for the next day. The operating status of the energy storage system is used to limit the charging and discharging range. After multi-objective optimization, the three factors form an executable time-sharing scheduling result.

[0251] In this embodiment, the preset multi-objective optimization function is a global coupled optimization objective of the multi-energy system, such as operational economy objective, system stability objective and energy comfort objective. The preset multi-objective optimization function is reused throughout the service side and user side to ensure the overall coordinated optimality of the multi-energy system, rather than the local optimality of a single device.

[0252] Specifically, the operational economics objective is to minimize the total daily operating cost of the household multi-energy system, including grid power purchase costs, energy storage charging and discharging losses, and heat pump operating energy costs, while maximizing the benefits of local photovoltaic consumption and adapting to the economic operating needs of users. The system stability objective is to minimize fluctuations in grid-connected power, photovoltaic power curtailment, and frequent fluctuations in energy storage power to ensure a balance between power supply and demand in the multi-energy system and improve grid adaptability and equipment operational stability. The energy comfort objective is to ensure that the supply of building heating and cooling loads and domestic hot water loads meets user needs, control the deviation of heat pump heating / cooling power within a reasonable range, and avoid energy shortages and large temperature fluctuations.

[0253] In this application, the thermoelectric coupling optimization process in steps S303 and S307 both adopt the same multi-objective optimization objective system and thermoelectric coupling solution algorithm, and uniformly take the optimal system operation economy, optimal operation stability, and optimal user energy comfort as the global optimization direction.

[0254] The difference between the two lies in the different input prediction data and equipment constraint boundaries: Step S303 takes the heat load prediction data as the core input, coupled with the electric side reference constraints, and prioritizes solving the day-ahead energy supply strategy of the heat pump that meets the user's heat demand, locking the rigid operating boundary of the load side; Step S307, based on the determined heat pump operating constraints, combined with the day-ahead photovoltaic output prediction data, the day-ahead electric load prediction data and the equipment power constraints, optimizes the day-ahead absorption and allocation strategy of the power supply side photovoltaic system and the day-ahead charging and discharging timing strategy of the energy storage system, under the premise of ensuring that the heat end energy demand remains unchanged.

[0255] Based on the above analysis, it can be seen that this hierarchical same-objective optimization architecture not only ensures the global collaborative optimality of multi-energy systems, but also achieves differentiated and precise scheduling of thermal and electrical equipment through step-by-step constraint solving, reducing the difficulty of model solving and improving the feasibility and stability of scheduling strategies.

[0256] Understandably, after adopting the above implementation method, the available power range of the heat pump tri-generation system can correspond one-to-one with the scheduling strategy, and the timing of photovoltaic consumption, load supply and energy storage charging and discharging can be determined collaboratively within the same optimization framework, so that the day-ahead scheduling results are consistent with the actual operating status, and the scheduling accuracy and execution stability of the multi-energy system are improved.

[0257] S308: The edge energy management unit responds to the day-ahead energy dispatch strategy generated by the multi-energy system, acquires multi-source heterogeneous data collected by the edge energy management unit, and acquires the current time and instantaneous operation data of the multi-energy system.

[0258] S309: The edge energy management unit determines the target preset period corresponding to the current time, as well as the day-ahead load forecast data within the target preset period, and inputs the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system.

[0259] Steps S308-S309 are similar to steps S101-S102, and will not be described again here.

[0260] S310: The edge energy management unit compares the real-time load forecast data with the day-ahead load forecast data to obtain the scheduling deviation value of the corresponding multi-energy system.

[0261] S311: The edge power management unit determines the correction strategy that matches the scheduling deviation value as the target correction strategy.

[0262] S312: The edge energy management unit corrects the daytime time-sharing data corresponding to the target preset period according to the target correction strategy to obtain the intraday time-sharing data.

[0263] S313: When the instantaneous operating data does not meet the abnormal operating conditions, the edge energy management unit controls the multi-energy system to operate according to the intraday time-sharing scheduling data within the target preset period.

[0264] The scheduling deviation value is used to characterize the degree of deviation between the energy scheduling plan obtained from the real-time forecast results and the energy scheduling plan obtained from the day-ahead forecast results.

[0265] In this embodiment of the application, when the cloud prediction model and the edge prediction model perform corresponding load data prediction, the dimensions of the output load prediction data are the same; for example, the output load prediction data all include the time-series prediction results of the building's heating and cooling load, domestic hot water load, household appliance load, photovoltaic output, and heat pump energy consumption load. The corresponding data dimensions, time-series granularity, and parameter definitions are completely matched, thereby ensuring that the data scope of the cloud global planning and the edge local control are consistent and can be seamlessly connected and iterated.

[0266] Because the actual application scenarios of cloud-based and edge-based prediction models differ, the scope and purpose of the prediction data output by the two models differ in actual scheduling scenarios. Specifically, when the energy coordination unit of cloud server 201 performs day-ahead scheduling, it only specifically calls data in the heat load prediction dimension to prioritize the solution of the day-ahead energy supply strategy for heat pumps. Predictions for other dimensions such as electricity load and photovoltaic output are only used as baseline constraints to assist in modeling. However, when the edge energy management unit performs intraday real-time rolling scheduling and deviation correction, it fully calls all-dimensional load and output prediction data output by the edge prediction model. Combined with real-time operating conditions, it completes the coordinated optimization and control of all equipment, including photovoltaic consumption, energy storage charging and discharging, and heat pump operation. This allows it to adapt to the hierarchical scheduling logic of day-ahead hierarchical baseline planning and intraday precise dynamic correction, effectively improving the adaptability and stability of multi-energy system coordinated control.

[0267] Understandably, real-time load forecast data is the load forecast result for the target preset period, inferred and output by the edge forecast model based on the measured data of the current period; day-ahead load forecast data is the baseline load forecast result for the same target preset period, obtained by the cloud energy coordination unit on the day-ahead; the difference between the two sets of forecast data is compared, and the day-ahead time-sharing scheduling data corresponding to the target preset period is corrected and adjusted according to the forecast deviation; among which, the corresponding adjustment objects include the day-ahead consumption allocation strategy of photovoltaic system and the day-ahead charging and discharging timing strategy of energy storage system.

[0268] In practice, after determining the current target preset period, the edge energy management unit extracts the day-ahead load forecast data corresponding to that period and aligns it with the real-time load forecast data output by the edge forecast model. It then calculates the load change in each time period according to the same time granularity to obtain the scheduling deviation value.

[0269] When the scheduling deviation is within a small range, the edge energy management unit can locally update the heat pump energy supply, energy storage charging and discharging power, or photovoltaic consumption allocation for the corresponding time period in the day-ahead time-of-use scheduling data. When the scheduling deviation exceeds the preset threshold, the real-time energy scheduling data can be recalculated by combining system configuration data, intraday heat load forecast data, intraday photovoltaic output forecast data, intraday electricity load forecast data, and equipment boundary constraints, and the real-time energy scheduling data can be used to replace or update the day-ahead time-of-use scheduling data.

[0270] Understandably, the above correction process can be completed locally in the edge energy management unit, and the correction results are output as intraday time-sharing scheduling data in time series form.

[0271] By comparing real-time load forecast data with day-ahead load forecast data and selecting a target correction strategy accordingly, day-ahead time-of-use scheduling data can be dynamically updated according to the actual operating status of the target preset cycle. This makes the intraday time-of-use scheduling data more closely match the current load changes and keep in line with the real-time operating requirements of multi-energy systems.

[0272] Meanwhile, compared to control methods that rely solely on daily plans, this correction mechanism can improve the timeliness and matching of scheduling results, and reduce the impact of accumulated prediction deviations on system stability during execution.

[0273] In one possible implementation, the correction strategy includes a first correction strategy; when the scheduling deviation is less than a preset deviation threshold, the first correction strategy is determined as the target correction strategy; based on real-time load forecast data, the system load status corresponding to the target preset period is determined; based on the scheduling deviation and the system load status, the compensation priority and adjustable margin of different energy systems are determined; based on the compensation priority and adjustable margin of different energy systems, the day-ahead time-of-day scheduling data corresponding to the target preset period is corrected to obtain intraday time-of-day scheduling data.

[0274] Understandably, the first correction strategy is a local correction strategy on the edge side. When this strategy is invoked, it uses the system load status obtained from real-time load forecast data and the obtained scheduling deviation value as input to determine the compensation priority and actual compensation capacity of each energy system, thereby completing the correction of the day-ahead time-of-day scheduling data of the target preset period.

[0275] The adjustable margin varies for different energy systems; for example, the adjustable margin of an energy storage system refers to the remaining discharge margin and charge margin; the adjustable margin of a photovoltaic system refers to the margin for increased power generation and absorption; and the adjustable margin of a combined heat pump and tri-generation system refers to the power margin that can be reduced.

[0276] In this embodiment of the application, a multi-energy collaborative compensation mechanism with clear priorities and limited adjustment directions is established during the deviation correction process of intraday time-sharing scheduling. Different adjustment devices are called in a differentiated manner according to the deviation type of real-time supply and demand in the multi-energy system.

[0277] Among them, the energy storage system is the first priority regulation unit, with bidirectional regulation capability. It can charge and absorb excess power under positive deviation conditions of power surplus, and discharge to supplement the gap under negative deviation conditions of power deficit, thus achieving power buffering and smoothing under all operating conditions. The photovoltaic system adopts the regulation logic of "only increasing and not decreasing", only participating in the increase deviation compensation under negative power deficit conditions. It makes up for the power gap by improving the local absorption capacity, and strictly prohibits the reduction and curtailment of photovoltaic output. The heat pump tri-generation system adopts the regulation logic of "only decreasing and not increasing", only participating in the decrease deviation correction under positive power surplus conditions. It can dissipate the system's surplus electricity by appropriately reducing the operating power. When there is a power deficit and insufficient heat supply, it does not forcibly supplement heat by increasing the heat pump power, but relies on energy storage discharge to ensure supply and demand balance and energy comfort.

[0278] In other words, when there is a surplus of positive power, energy storage is the primary method for absorption and storage, while heat pumps are used to reduce load, and photovoltaics maintain their maximum output. When there is a shortage of negative power, energy storage is the primary method for energy replenishment, while photovoltaics are used to absorb the excess power, and heat pumps maintain their basic heat load. By coordinating the unidirectional and dedicated adjustment characteristics of each device, adjustment conflicts are avoided, and the deviation of the multi-energy system is corrected smoothly, safely, and economically.

[0279] In practice, after the edge energy management unit obtains the real-time load forecast data, it compares it with the day-ahead load forecast data to obtain the scheduling deviation value, and then compares the scheduling deviation value with the preset deviation threshold.

[0280] When the scheduling deviation is less than the preset deviation threshold, the first correction strategy is adopted. Subsequently, the load level within the target preset period is mapped according to the real-time load forecast data to form the system load status. Then, combining the magnitude of the scheduling deviation and the system load status, compensation priorities are configured for different energy systems such as photovoltaic systems, energy storage systems, combined heat and power systems, and smart home appliance loads. Adjustable margins are determined according to their respective power boundaries, charge and discharge boundaries, or start-stop boundaries. Then, under the constraints of priority and margin, local corrections are performed on the day-ahead time-of-use scheduling data to output the intraday time-of-use scheduling data.

[0281] Understandably, after selecting applicable correction strategies based on preset deviation thresholds, real-time load forecast data is used to update the system load status, thereby coordinating the compensation sequence and adjustment magnitude of different energy systems to ensure that the correction results are consistent with the current operating status. Since the first correction strategy is executed only when the scheduling deviation value is small, and the correction process is constrained by compensation priority and adjustable margin, it is possible to perform refined updates to day-ahead time-of-day scheduling data.

[0282] By adopting the above method, intraday time-sharing scheduling data can be locally corrected when the scheduling deviation is small. The correction range matches the actual adjustable range of each energy system, thereby improving the consistency between the day-ahead plan and intraday operation, and making the scheduling results of multiple energy systems more consistent with real-time load changes.

[0283] In one possible implementation, the correction strategy further includes a second correction strategy; the real-time load forecast data includes intraday heat load forecast data, intraday photovoltaic power output forecast data, and intraday electrical load forecast data; if the scheduling deviation value is not less than a preset deviation threshold, the second correction strategy is determined as the target correction strategy; the equipment boundary constraints and system configuration data of the multi-energy system are determined; based on the system configuration data, intraday heat load forecast data, intraday photovoltaic power output forecast data, intraday electrical load forecast data, and equipment boundary constraints, the optimal solution for thermoelectric coupling scheduling is performed with the preset multi-objective function as the optimization direction to obtain the real-time energy scheduling data corresponding to the target preset period; based on the real-time energy scheduling data, the day-ahead time-of-day scheduling data corresponding to the target preset period is updated to obtain the intraday time-of-day scheduling data.

[0284] Among them, the equipment boundary constraints are applied to the modification process of the scheduling strategy within the preset period, and serve as constraints for solving the scheduling strategy. They are used to limit the operating range of the photovoltaic system, energy storage system, and combined heat and power system at the scheduling plan level. The equipment boundary constraints include at least one of the following: energy storage system charging and discharging power constraints, energy storage state of charge scheduling constraints, heat pump power constraints, indoor temperature constraints, and water temperature constraints.

[0285] Understandably, system configuration data includes external inputs and system configuration data that participate in optimization calculations, including at least one of the following: electricity price, room temperature setpoint, domestic water tank temperature setpoint, heat pump unit start / stop status, unit operating mode, heat pump temperature setpoint, and inverter start / stop status.

[0286] In practice, after receiving the real-time load forecast results, the edge energy management unit subdivides them into intraday heat load forecast data, intraday photovoltaic power output forecast data, and intraday electricity load forecast data, and calculates the scheduling deviation value by combining the intraday time-of-day scheduling data within the current target preset period.

[0287] When the scheduling deviation is not less than the preset deviation threshold, the second correction strategy is invoked. The current electricity price, room temperature setpoint, domestic water tank temperature setpoint, heat pump unit start / stop status, unit operating mode, heat pump temperature setpoint, and inverter start / stop status are read and input into the thermoelectric coupling scheduling model along with the energy storage system's charging and discharging power constraints, energy storage state of charge scheduling constraints, heat pump power constraints, indoor temperature constraints, and water temperature constraints. A multi-objective function is used to jointly optimize the electricity purchase cost, equipment operation balance, and energy supply matching degree, and output the real-time energy scheduling data corresponding to the target preset period.

[0288] Based on the real-time energy dispatch data obtained from the solution, the daytime time-sharing dispatch data corresponding to the preset period of the target is iteratively updated, replacing the optimal power control parameters of each device in the current preset period. At the same time, the overall timing framework, safety constraints and global optimization logic of the daytime dispatch are retained. Finally, intraday time-sharing dispatch data that takes into account both the global optimum throughout the day and the adaptability to short-term operating conditions and can be directly sent to devices for execution is generated, realizing the rolling optimization and update of the multi-energy system dispatch strategy.

[0289] Understandably, the optimization solution for thermoelectric coupling scheduling can be completed locally by the edge energy management unit, and the real-time energy scheduling data reflects the feasible optimal scheduling result under the current external inputs, equipment status, and boundary conditions. Since the daily heat load, photovoltaic output, and electrical load are included in the calculation simultaneously, and the update process is limited by equipment boundary constraints, the intraday time-of-use scheduling data obtained can remain consistent with the current operating conditions.

[0290] When this method is adopted, if the scheduling deviation is large, the second correction strategy can be switched to and the solution can be re-solved to generate real-time energy scheduling data that is more in line with the current state. This makes the intraday time-sharing scheduling data more consistent with the actual load changes, equipment operating status and constraints, thereby improving the accuracy and consistency of scheduling updates.

[0291] In one possible implementation, the abnormal operating conditions include a preset abnormal threshold and a preset allowable range. After correcting the daytime time-sharing scheduling data corresponding to the target preset period based on real-time load forecast data and daytime load forecast data to obtain intraday time-sharing scheduling data, the instantaneous power deviation of the multi-energy system is determined based on instantaneous operating data. If the instantaneous power deviation does not exceed the preset abnormal threshold and the instantaneous operating data does not exceed the preset allowable range, it is determined that the instantaneous operating data does not meet the abnormal operating conditions. If the instantaneous power deviation exceeds the preset abnormal threshold and / or the instantaneous operating data exceeds the preset allowable range, it is determined that the instantaneous operating data meets the abnormal operating conditions.

[0292] The instantaneous power deviation is used to reflect the degree of deviation between power supply and demand in the multi-energy system at the current instantaneous moment; the instantaneous operating data includes at least one of the following: photovoltaic instantaneous power, energy storage instantaneous charging and discharging power, heat pump instantaneous power, load instantaneous power, grid connection point frequency detection data, and grid connection point voltage detection data.

[0293] In practice, after receiving intraday time-sharing scheduling data, the edge energy management unit continuously receives instantaneous operating data from the photovoltaic system, energy storage system, combined heat and power system, load meter, and grid connection detection unit, and maps the data to the same moment for verification.

[0294] If the difference between the combined instantaneous power of photovoltaics and the instantaneous charging and discharging power of energy storage and the instantaneous power of the load exceeds the preset abnormal threshold, the calculated instantaneous power deviation is characterized as abnormal; if the grid connection point frequency or voltage exceeds the preset allowable range, the corresponding instantaneous operating data is directly determined to meet the abnormal operating conditions.

[0295] The frequency detection data can be collected by the grid connection point frequency sensor, and the voltage detection data can be collected by the grid connection point voltage sampling module. In practical applications, other models of the above sensors can also be selected, and this application does not limit them.

[0296] When the instantaneous operating data is determined not to meet the abnormal operating conditions, the edge energy management unit continues to output control commands based on the intraday time-sharing scheduling data, so that the multi-energy system can operate stably according to the corrected scheduling scheme. When the abnormal operating conditions are determined to be met, the system switches to abnormal control and compensation processing, and redistributes the energy storage charging and discharging power, heat pump operating power, or grid-connected exchange power. At this time, by jointly judging the instantaneous power deviation and the out-of-bounds situation of the instantaneous operating data, the system can identify abnormalities in a timely manner when there is power imbalance or abnormal equipment status, thereby improving the real-time performance and stability of intraday scheduling execution and reducing the probability of the abnormal state continuing to expand.

[0297] For example, the instantaneous power deviation can be the power difference between the instantaneous total power generation and the instantaneous total power consumption at the multi-energy grid connection point, which is used to reflect the degree of imbalance between power supply and demand in the multi-energy system.

[0298] Specifically, this instantaneous power deviation can be ,in, This refers to the instantaneous output power of the photovoltaic system. The instantaneous charging and discharging power of energy storage follows the rule of positive for discharging and negative for charging; This refers to the instantaneous power of the total system load, which includes the heat pump operating load and the household's regular electricity load. The instantaneous power deviation is used to characterize the real-time supply and demand matching status of the system.

[0299] When there are sudden changes in lighting, sudden changes in load, or power grid disturbances... The increase will be sharp, indicating an instantaneous power imbalance in the system.

[0300] In one possible implementation, if it is determined that the instantaneous operating data meets the abnormal operating conditions, the abnormal disturbance type and degree of the multi-energy system are determined based on the instantaneous operating data; the target compensation strategy of the multi-energy system is determined according to the abnormal disturbance type and degree; the intraday time-sharing scheduling data is corrected according to the target compensation strategy to obtain the corrected time-sharing scheduling data, and the multi-energy system is controlled to operate according to the corrected time-sharing scheduling data within the target preset period.

[0301] The target compensation strategy includes at least one of the following: energy storage compensation control strategy, heat pump power regulation control strategy, and load reduction control strategy.

[0302] In practical implementation, the edge energy management unit inputs the abnormal disturbance type and abnormal disturbance degree into the abnormal compensation judgment module, and outputs the target compensation strategy in combination with the preset mapping relationship.

[0303] Specifically, for situations where there is power imbalance and the energy storage system is available, the control module generates energy storage charge and discharge correction amounts and reconstructs the energy storage power commands in the intraday time-of-use scheduling data; for situations where energy storage compensation is insufficient or the combined heat and power system is limited, the control module corrects the heat pump power commands according to the allowable power range of the combined heat and power system; for situations with a high degree of abnormality and insufficient compensation as mentioned above, the control module reduces interruptible loads, forms corrected time-of-use scheduling data, and sends it to the energy storage system, the combined heat and power system, and the load controller for execution.

[0304] In one possible implementation, when the abnormal disturbance type is a power imbalance disturbance, the priority execution framework is determined to be a three-level progressive control strategy with energy storage as the priority, heat pump as the auxiliary, and load reduction as the last resort. The energy storage compensation duration, heat pump adjustment duration, or grid load reduction execution duration are matched level by level according to the degree of abnormal disturbance.

[0305] When the abnormal disturbance type is a fault disturbance of the energy storage body, the energy storage compensation control is eliminated, and the priority execution framework is determined to be a two-level control strategy of heat pump priority and load reduction as a fallback. The duration of continuous adjustment of heat pump is determined according to the degree of abnormal disturbance, or the emergency load reduction execution duration is directly matched.

[0306] When the abnormal disturbance type is a heat pump fault disturbance, the heat pump power regulation control is eliminated, and the priority execution framework is determined to be a two-level control strategy with energy storage priority and load reduction as a fallback. The duration of continuous energy storage compensation is determined according to the degree of abnormal disturbance, and the grid load reduction duration is matched when the disturbance exceeds the limit.

[0307] When the abnormal disturbance type is a severe disturbance on the grid side, the load reduction control strategy is directly activated, and the corresponding gradient of demand response or emergency load reduction duration is matched according to the severity of the abnormal disturbance.

[0308] In practice, the edge power management unit calculates the instantaneous power deviation based on instantaneous operating data and compares it with preset thresholds and adjustable boundaries of the equipment to determine the type and degree of disturbance.

[0309] When a power imbalance disturbance is identified, the energy storage system is first controlled to maintain its charging and discharging power within a limited time period according to the energy storage compensation duration. Then, if the energy storage compensation is insufficient, the power of the combined heat and power system (CHP) is adjusted according to the heat pump regulation duration. If the deviation still exceeds the tolerance, the grid load reduction execution duration is output to reduce non-critical loads. When a fault disturbance in the energy storage itself is identified, the control logic no longer sends compensation commands to the energy storage side. Instead, it generates a continuous heat pump regulation duration based on the available power range of the CHP system. If the heat pump load increase or decrease still cannot meet the requirements... If the balance requirements are met, the system switches to emergency load shedding execution duration. When a heat pump fault disturbance is identified, the system disables heat pump power regulation control and generates continuous energy storage compensation duration based solely on the energy storage state of charge, charging and discharging power constraints, and remaining available capacity. If the energy storage has reached its limit, the system enters grid load shedding execution duration control. When a severe disturbance on the grid side is identified, the system directly generates load shedding control quantity and matches the corresponding continuous execution duration between demand response and emergency load shedding according to the severity of the disturbance, so that the user-side power can quickly return to a safe range.

[0310] Understandably, through the above control methods, the system can automatically switch control objects and control levels for different abnormal disturbances, so that the energy storage system, the combined heat pump system and the load reduction can take over control in an appropriate order under fault or over-limit scenarios. This allows the abnormal compensation path to match the disturbance characteristics, and enables the corrected time-sharing scheduling data to be restored to the safe operating range in a shorter time.

[0311] In one possible implementation, the target preset period can be the remaining scheduling period corresponding to the current time, and the corrected time-sharing scheduling data can be directly used as the basis for the operation of this period.

[0312] Understandably, this method first identifies the source of the anomaly after abnormal operating conditions are triggered, then selects a compensation path based on the anomaly intensity, and writes the compensation result into the intraday time-sharing scheduling data, enabling each device to operate collaboratively according to the new scheduling sequence. Thus, partial reconfiguration control under abnormal scenarios can be completed without relying on real-time intervention from a cloud server, while maintaining the operational continuity and scheduling consistency of the multi-energy system within the target preset period.

[0313] By adopting this approach, multi-energy systems can employ matched compensation control for different abnormal disturbances, reducing the continued spread of power deviations and enabling energy storage systems, combined heat pump systems, and load control to work together to correct the situation, thereby improving control stability and scheduling accuracy under abnormal scenarios.

[0314] The control method for a multi-energy system provided in this embodiment acquires historical multi-source heterogeneous data sent by the edge energy management unit through an energy coordination unit, and inputs the historical multi-source heterogeneous data into a cloud prediction model to obtain the day-ahead prediction data of the multi-energy system. Based on the heat load prediction data, and with the optimization direction of a preset multi-objective function as the optimization direction, thermoelectric coupling scheduling optimization is performed to obtain day-ahead energy supply strategies corresponding to multiple preset periods, and the day-ahead energy supply strategies are transmitted to the edge energy management unit. The edge energy management unit determines the power constraints corresponding to the day-ahead energy supply strategies, and then determines the historical multi-source heterogeneous data collected by the edge energy management unit, and inputs the historical multi-source heterogeneous data into the edge prediction model to obtain day-ahead photovoltaic output prediction data and day-ahead electrical load prediction data. The operating status of the energy storage system is determined, and based on the operating status, day-ahead photovoltaic output prediction data, day-ahead electrical load prediction data, and power constraints, and with the optimization direction of the preset multi-objective function as the optimization direction, thermoelectric coupling is performed. The scheduling optimization process yields day-ahead consumption allocation strategies and day-ahead charging / discharging timing strategies corresponding to multiple preset periods. Responding to the day-ahead energy scheduling strategies generated by the multi-energy system, it acquires multi-source heterogeneous data collected by the edge energy management unit and obtains the current moment and instantaneous operating data of the multi-energy system. It determines the target preset period corresponding to the current moment, as well as the day-ahead load forecast data within the target preset period, and inputs the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system. It compares the real-time load forecast data with the day-ahead load forecast data to obtain the scheduling deviation value of the corresponding multi-energy system, and determines the correction strategy matching the scheduling deviation value as the target correction strategy. It corrects the day-ahead time-of-day scheduling data corresponding to the target preset period according to the target correction strategy to obtain intraday time-of-day scheduling data. If the instantaneous operating data does not meet the abnormal operating conditions, it controls the multi-energy system to operate according to the intraday time-of-day scheduling data within the target preset period. This method addresses the problems of existing home energy management systems, such as heavy reliance on centralized cloud-based scheduling, insufficient coordination between daily planning and intraday operation, and easy accumulation of prediction errors, which lead to poor real-time performance, stability, and economy in scheduling. At the same time, it reduces the dependence of multi-energy system collaborative control on continuous cloud communication and command issuance, and enhances the local autonomous control capability and response capability to dynamic operating conditions of home multi-energy systems under multi-timescale collaboration, thereby improving the real-time performance, accuracy, and stability of energy scheduling under dynamic operating conditions.

[0315] Figure 4 This is a schematic diagram of the control device for the multi-energy system provided in this application. Figure 4 As shown, this application provides a control device for a multi-energy system, applied to the edge energy management unit of a multi-energy system. The control device 400 for the multi-energy system includes:

[0316] The acquisition module 401 is used to acquire multi-source heterogeneous data collected by the edge energy management unit in response to the day-ahead energy scheduling strategy generated by the multi-energy system, and to acquire the current time and instantaneous operation data of the multi-energy system; the day-ahead energy scheduling strategy includes day-ahead time-sharing scheduling data corresponding to multiple preset periods.

[0317] The processing module 402 is used to determine the target preset period corresponding to the current time, as well as the day-ahead load forecast data within the target preset period, and input the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system.

[0318] The processing module 402 is also used to correct the daytime time-sharing scheduling data corresponding to the target preset period based on the real-time load forecast data and the daytime load forecast data to obtain the intraday time-sharing scheduling data.

[0319] The processing module 402 is also used to control the multi-energy system to operate according to the intraday time-sharing scheduling data within the target preset period when the instantaneous operating data does not meet the abnormal operating conditions.

[0320] In one possible implementation, the processing module 402 is further configured to determine the power constraint conditions corresponding to the day-ahead energy supply strategy in response to the day-ahead energy supply strategy sent by the energy coordination unit; the power constraint conditions are used to characterize the available power range of the heat pump tri-generation system.

[0321] The processing module 402 is also used to determine the historical multi-source heterogeneous data collected by the edge energy management unit, and input the historical multi-source heterogeneous data into the edge prediction model to obtain the day-ahead photovoltaic power output prediction data and the day-ahead electrical load prediction data.

[0322] The processing module 402 is also used to determine the operating status of the energy storage system, and based on the operating status, day-ahead photovoltaic output prediction data, day-ahead electrical load prediction data and power constraints, it performs thermoelectric coupling scheduling optimization with the optimization direction of the preset multi-objective function as the optimization direction, and obtains day-ahead consumption allocation strategy and day-ahead charging and discharging timing strategy corresponding to multiple preset periods; the thermoelectric coupling scheduling optimization is implemented by at least one of the weighted summation multi-objective optimization algorithm, particle swarm optimization algorithm and genetic algorithm.

[0323] In one possible implementation, the processing module 402 is further configured to compare the real-time load forecast data with the day-ahead load forecast data to obtain the scheduling deviation value of the corresponding multi-energy system.

[0324] The processing module 402 is also used to determine the correction strategy that matches the scheduling deviation value as the target correction strategy.

[0325] The processing module 402 is also used to correct the day-ahead time-sharing scheduling data corresponding to the target preset period according to the target correction strategy, so as to obtain the day-ahead time-sharing scheduling data.

[0326] In one possible implementation, the processing module 402 is further configured to determine the first correction strategy as the target correction strategy when the scheduling deviation value is less than a preset deviation threshold.

[0327] The processing module 402 is also used to determine the system load status corresponding to the target preset period based on real-time load forecast data.

[0328] The processing module 402 is also used to determine the compensation priority and adjustable margin of different energy systems based on scheduling deviation and system load status.

[0329] The processing module 402 is also used to correct the daytime time-sharing data corresponding to the target preset cycle based on the compensation priority and adjustable margin of different energy systems, so as to obtain the intraday time-sharing data.

[0330] In one possible implementation, the processing module 402 is further configured to determine the second correction strategy as the target correction strategy if the scheduling deviation value is not less than a preset deviation threshold.

[0331] The processing module 402 is also used to determine the equipment boundary constraints and system configuration data of the multi-energy system.

[0332] The processing module 402 is also used to perform thermal-electric coupling scheduling optimization based on system configuration data, intraday heat load forecast data, intraday photovoltaic power output forecast data, intraday electricity load forecast data and equipment boundary constraints, with the optimization direction of the preset multi-objective function as the optimization direction, to obtain real-time energy scheduling data corresponding to the preset target period.

[0333] The processing module 402 is also used to update the daytime time-sharing data corresponding to the target preset cycle based on real-time energy dispatch data to obtain intraday time-sharing data.

[0334] In one possible implementation, the processing module 402 is further configured to determine the instantaneous power deviation of the multi-energy system based on instantaneous operating data; the instantaneous power deviation is used to reflect the degree of deviation between power supply and demand of the multi-energy system at the current instantaneous moment.

[0335] The processing module 402 is also used to determine that the instantaneous operating data does not meet the abnormal operating conditions when the instantaneous power deviation does not exceed the preset abnormal threshold and the instantaneous operating data does not exceed the preset allowable range.

[0336] The processing module 402 is also used to determine that the instantaneous operating data meets the abnormal operating conditions when the instantaneous power deviation exceeds the preset abnormal threshold and / or the instantaneous operating data exceeds the preset allowable range.

[0337] The instantaneous operating data includes at least one of the following: photovoltaic instantaneous power, energy storage instantaneous charging and discharging power, heat pump instantaneous power, load instantaneous power, grid connection point frequency detection data, and grid connection point voltage detection data.

[0338] In one possible implementation, the processing module 402 is further configured to determine the type and degree of abnormal disturbance of the multi-energy system based on instantaneous operating data.

[0339] The processing module 402 is also used to determine the target compensation strategy for the multi-energy system based on the type and degree of abnormal disturbance.

[0340] The processing module 402 is also used to correct the intraday time-sharing scheduling data according to the target compensation strategy, obtain the corrected time-sharing scheduling data, and control the multi-energy system to operate according to the corrected time-sharing scheduling data within the target preset period.

[0341] The target compensation strategy includes at least one of the following: energy storage compensation control strategy, heat pump power regulation control strategy, and load reduction control strategy.

[0342] Figure 5 This is a schematic diagram of the control device for the multi-energy system provided in this application. Figure 5 As shown, this application provides a control device for a multi-energy system, applied to an energy coordination unit of a multi-energy system. The control device 500 for the multi-energy system includes:

[0343] The acquisition module 501 is used to acquire historical multi-source heterogeneous data sent by the edge energy management unit.

[0344] The processing module 502 is used to input historical multi-source heterogeneous data into the cloud prediction model to obtain day-ahead prediction data of multi-energy systems; the day-ahead prediction data includes heat load prediction data.

[0345] The processing module 502 is also used to perform thermoelectric coupling scheduling optimization based on heat load prediction data, with the optimization direction of preset multi-objective function as the optimization direction, to obtain daytime energy supply strategies corresponding to multiple preset periods.

[0346] The processing module 502 is also used to send the day-ahead energy supply strategy to the edge energy management unit.

[0347] In one possible implementation, the processing module 502 is also used to extract photovoltaic power output data, building thermal environment parameters, domestic hot water usage data, household appliance load data and heat pump performance data from historical multi-source heterogeneous data.

[0348] The processing module 502 is also used to determine the data types of photovoltaic power output data, building thermal environment parameters, domestic hot water usage data, household appliance load data and heat pump performance data respectively, and for any data type, match the target modeling algorithm corresponding to the data type.

[0349] The processing module 502 is also used to perform differentiated training on historical multi-source heterogeneous data by using target modeling algorithms corresponding to different data types, and obtain multiple prediction branches.

[0350] The processing module 502 is also used to integrate multiple prediction branches to obtain a cloud prediction model;

[0351] Among them, the photovoltaic power generation prediction branch was trained using a long short-term memory network model; the building thermal environment prediction branch was trained using an RC building thermal environment model; the domestic hot water load prediction branch was trained using an ARIMA model; the household appliance load prediction branch was trained using a random forest model; and the heat pump performance prediction branch was trained using an SVR model.

[0352] In one possible implementation, the processing module 502 is further configured to perform knowledge distillation on the cloud prediction model to obtain a lightweight edge prediction model.

[0353] The processing module 502 is also used to send the edge prediction model to the edge energy management unit.

[0354] Figure 6 A schematic diagram of the structure of the electronic device provided in this application. Figure 6 As shown, the electronic device 600 provided in this embodiment includes at least one processor 601 and a memory 602. Optionally, the device 600 further includes a communication component 603. The processor 601, memory 602, and communication component 603 are connected via a bus 604.

[0355] In a specific implementation, at least one processor 601 executes computer execution instructions stored in memory 602, causing at least one processor 601 to perform the above-described method.

[0356] The specific implementation process of processor 601 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0357] 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.

[0358] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0359] 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.

[0360] This application also provides a multi-energy system, including an edge energy management unit and an energy coordination unit;

[0361] The edge energy management unit is deployed on the user side of the multi-energy system; the energy coordination unit is deployed on the service side of the multi-energy system; the energy coordination unit communicates with the edge energy management unit.

[0362] The edge energy management unit is used to receive lightweight prediction models, use lightweight prediction models to perform load prediction of local multi-energy systems, and combine thermoelectric coupling scheduling optimization solutions to output time-sharing scheduling data of multi-energy systems, thus completing photovoltaic-storage-thermal coupling optimization control.

[0363] The energy coordination unit is used to train the prediction model based on global multi-energy data, generate a lightweight prediction model, and distribute the lightweight model to the edge energy management unit; it is also used to generate the day-ahead energy supply strategy in the day-ahead energy dispatch strategy.

[0364] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described control method for a multi-energy system.

[0365] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described control method for a multi-energy system.

[0366] 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.

[0367] 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.

[0368] 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.

[0369] 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.

[0370] 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.

[0371] 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.

[0372] 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.

[0373] 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.

Claims

1. A control method for a multi-energy system, characterized in that, The method is applied to the multi-energy system, which includes an edge energy management unit deployed on the user side of the multi-energy system; the method includes: The edge power management unit performs the following steps: In response to the day-ahead energy scheduling strategy generated by the multi-energy system, the system acquires multi-source heterogeneous data collected by the edge energy management unit and acquires the current time and instantaneous operation data of the multi-energy system; the day-ahead energy scheduling strategy includes day-ahead time-sharing scheduling data corresponding to multiple preset periods; The target preset period corresponding to the current time is determined, as well as the day-ahead load forecast data within the target preset period. The multi-source heterogeneous data is then input into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system. The edge prediction model is an inference model that has been lightweighted by the server and then sent to the edge energy management unit. Based on the real-time load forecast data and the day-ahead load forecast data, the day-ahead time-sharing scheduling data corresponding to the target preset period is corrected to obtain the intraday time-sharing scheduling data. If the instantaneous operating data does not meet the abnormal operating conditions, the multi-energy system is controlled to operate according to the intraday time-sharing scheduling data within the target preset period; The multi-energy system further includes a combined heat and power (CHP) system, a photovoltaic system, and an energy storage system; the day-ahead energy dispatch strategy includes a day-ahead energy supply strategy for the CHP system, a day-ahead energy consumption and allocation strategy for the photovoltaic system, and a day-ahead charge and discharge timing strategy for the energy storage system; prior to responding to the day-ahead energy dispatch strategy of the multi-energy system, the method further includes: In response to the day-ahead energy supply strategy sent by the energy coordination unit, the power constraint conditions corresponding to the day-ahead energy supply strategy are determined; the power constraint conditions are used to characterize the available power range of the heat pump tri-generation system; The historical multi-source heterogeneous data collected by the edge energy management unit is determined, and the historical multi-source heterogeneous data is input into the edge prediction model to obtain day-ahead photovoltaic power output prediction data and day-ahead electricity load prediction data. The operating status of the energy storage system is determined, and based on the operating status, the day-ahead photovoltaic output forecast data, the day-ahead electrical load forecast data, and the power constraints, a thermoelectric coupling scheduling optimization solution is performed with the optimization direction of a preset multi-objective function, to obtain the day-ahead consumption allocation strategy and the day-ahead charge-discharge timing strategy corresponding to multiple preset periods; the thermoelectric coupling scheduling optimization solution is implemented using at least one of the following: weighted summation multi-objective optimization algorithm, particle swarm optimization algorithm, and genetic algorithm.

2. The method according to claim 1, characterized in that, The step of correcting the daytime time-sharing scheduling data corresponding to the target preset period based on the real-time load forecast data and the daytime load forecast data to obtain intraday time-sharing scheduling data includes: The real-time load forecast data and the day-ahead load forecast data are compared to obtain the scheduling deviation value corresponding to the multi-energy system. The correction strategy that matches the scheduling deviation value is determined as the target correction strategy; The daily time-sharing scheduling data corresponding to the target preset period is corrected according to the target correction strategy to obtain the intraday time-sharing scheduling data.

3. The method according to claim 2, characterized in that, The correction strategy includes a first correction strategy; determining the correction strategy that matches the scheduling deviation value as the target correction strategy includes: If the scheduling deviation value is less than a preset deviation threshold, the first correction strategy is determined as the target correction strategy; The step of correcting the daytime time-sharing scheduling data corresponding to the target preset period according to the target correction strategy to obtain the intraday time-sharing scheduling data includes: Based on the real-time load forecast data, the system load status corresponding to the target preset period is determined; Based on the scheduling deviation and the system load status, the compensation priority and adjustable margin of different energy systems are determined; Based on the compensation priority and adjustable margin of different energy systems, the daytime time-sharing data corresponding to the target preset period is corrected to obtain the intraday time-sharing data.

4. The method according to claim 2, characterized in that, The correction strategy further includes a second correction strategy; the step of determining the correction strategy that matches the scheduling deviation value as the target correction strategy further includes: If the scheduling deviation value is not less than a preset deviation threshold, the second correction strategy is determined as the target correction strategy; The real-time load forecast data includes intraday heat load forecast data, intraday photovoltaic power output forecast data, and intraday electricity load forecast data; the step of correcting the day-ahead time-of-use scheduling data corresponding to the target preset period according to the target correction strategy to obtain the intraday time-of-use scheduling data further includes: Determine the equipment boundary constraints and system configuration data of the multi-energy system; Based on the system configuration data, the intraday heat load forecast data, the intraday photovoltaic output forecast data, the intraday electricity load forecast data, and the equipment boundary constraints, the optimal solution for the thermoelectric coupling scheduling is performed with the optimization direction of the preset multi-objective function, so as to obtain the real-time energy scheduling data corresponding to the preset target period. Based on the real-time energy dispatch data, the daytime time-sharing dispatch data corresponding to the target preset period is updated to obtain the intraday time-sharing dispatch data.

5. The method according to claim 1, characterized in that, The abnormal operating conditions include a preset abnormal threshold and a preset allowable range; after correcting the daytime time-sharing scheduling data corresponding to the target preset period based on the real-time load forecast data and the daytime load forecast data to obtain the intraday time-sharing scheduling data, the method further includes: Based on the instantaneous operating data, the instantaneous power deviation of the multi-energy system is determined; the instantaneous power deviation is used to reflect the degree of deviation between power supply and demand of the multi-energy system at the current instantaneous moment. If the instantaneous power deviation does not exceed the preset abnormal threshold and the instantaneous operating data does not exceed the preset allowable range, it is determined that the instantaneous operating data does not meet the abnormal operating conditions. If the instantaneous power deviation exceeds the preset abnormal threshold, and / or the instantaneous operating data exceeds the preset allowable range, the instantaneous operating data is determined to meet the abnormal operating conditions. The instantaneous operating data includes at least one of the following: photovoltaic instantaneous power, energy storage instantaneous charge and discharge power, heat pump instantaneous power, load instantaneous power, grid connection point frequency detection data, and grid connection point voltage detection data.

6. The method according to claim 5, characterized in that, After determining that the instantaneous operating data meets the abnormal operating conditions, the method further includes: Based on the instantaneous operating data, determine the type and degree of abnormal disturbance of the multi-energy system; Based on the type and degree of the abnormal disturbance, the target compensation strategy for the multi-energy system is determined; According to the target compensation strategy, the intraday time-sharing scheduling data is corrected to obtain corrected time-sharing scheduling data, and the multi-energy system is controlled to operate according to the corrected time-sharing scheduling data within the target preset period; The target compensation strategy includes at least one of the following: energy storage compensation control strategy, heat pump power regulation control strategy, and load reduction control strategy.

7. The method according to claim 1, characterized in that, The multi-energy system includes an energy coordination unit deployed on the service side of the multi-energy system; the method includes: The energy coordination unit performs the following steps: Acquire historical multi-source heterogeneous data sent by the edge energy management unit; The historical multi-source heterogeneous data is input into the cloud-based prediction model to obtain the day-ahead prediction data of the multi-energy system; the day-ahead prediction data includes heat load prediction data. Based on the heat load forecast data, with the optimization direction of the preset multi-objective function as the optimization direction, the thermoelectric coupling scheduling optimization is performed to obtain the daytime energy supply strategy corresponding to multiple preset cycles. The daytime energy supply strategy is sent to the edge energy management unit.

8. The method according to claim 7, characterized in that, The cloud-based prediction model includes branches for photovoltaic power generation prediction, building thermal environment prediction, domestic hot water load prediction, household appliance load prediction, and heat pump performance prediction; the method also includes: Photovoltaic output data, building thermal environment parameters, domestic hot water usage data, household appliance load data, and heat pump performance data are extracted from the aforementioned historical multi-source heterogeneous data. The data types of the photovoltaic output data, the building thermal environment parameters, the domestic hot water usage data, the household appliance load data, and the heat pump performance data are determined respectively, and for any of the data types, a target modeling algorithm corresponding to the data type is matched; The target modeling algorithms corresponding to different data types are used to perform differentiated training on the historical multi-source heterogeneous data to obtain multiple prediction branches; The multiple prediction branches are integrated to obtain the cloud prediction model; Specifically, the photovoltaic power generation prediction branch is obtained by training a long short-term memory network model; the building thermal environment prediction branch is obtained by training an RC building thermal environment model; the domestic hot water load prediction branch is obtained by training an ARIMA model; the household appliance load prediction branch is obtained by training a random forest model; and the heat pump performance prediction branch is obtained by training an SVR model.

9. The method according to claim 8, characterized in that, After integrating the multiple prediction branches to obtain the cloud prediction model, the method further includes: The cloud prediction model is subjected to knowledge distillation to obtain a lightweight edge prediction model; The edge prediction model is sent to the edge energy management unit.

10. A multi-energy system, characterized in that, This includes a combined heat pump and power system, a photovoltaic system, an energy storage system, an edge energy management unit, and an energy synergy unit; The edge energy management unit is deployed on the user side of the multi-energy system; the energy coordination unit is deployed on the service side of the multi-energy system; the energy coordination unit is communicatively connected to the edge energy management unit. The energy coordination unit is used to acquire historical multi-source heterogeneous data sent by the edge energy management unit; input the historical multi-source heterogeneous data into the cloud prediction model to obtain the day-ahead prediction data of the multi-energy system; the day-ahead prediction data includes heat load prediction data. Based on the heat load forecast data, with the optimization direction of the preset multi-objective function as the optimization direction, the thermoelectric coupling scheduling optimization is performed to obtain the day-ahead energy supply strategy corresponding to multiple preset periods; the day-ahead energy supply strategy is sent to the edge energy management unit; The edge energy management unit is used to respond to the day-ahead energy supply strategy sent by the energy coordination unit, determine the power constraints corresponding to the day-ahead energy supply strategy; input the historical multi-source heterogeneous data collected by the edge energy management unit into the edge prediction model to obtain day-ahead photovoltaic power output prediction data and day-ahead electrical load prediction data; determine the operating status of the energy storage system, and based on the operating status, the day-ahead photovoltaic power output prediction data, the day-ahead electrical load prediction data and the power constraints, perform thermoelectric coupling scheduling optimization with the optimization direction of the preset multi-objective function as the optimization direction, to obtain the day-ahead consumption allocation strategy and the day-ahead charge and discharge timing strategy corresponding to multiple preset periods; The day-ahead energy dispatch strategy includes the day-ahead energy supply strategy of the combined heat pump and combined heat and power system, the day-ahead energy consumption and distribution strategy of the photovoltaic system, and the day-ahead charge and discharge timing strategy of the energy storage system; The edge energy management unit is also used to respond to the day-ahead energy scheduling strategy generated by the multi-energy system, acquire the multi-source heterogeneous data collected by the edge energy management unit, and acquire the current time and instantaneous operation data of the multi-energy system; the day-ahead energy scheduling strategy includes day-ahead time-sharing scheduling data corresponding to multiple preset periods; Determine the target preset period corresponding to the current time, and the day-ahead load forecast data within the target preset period, and input the multi-source heterogeneous data into the edge prediction model for inference to obtain the real-time load forecast data of the multi-energy system; Based on the real-time load forecast data and the day-ahead load forecast data, the day-ahead time-sharing scheduling data corresponding to the target preset period is corrected to obtain the intraday time-sharing scheduling data. If the instantaneous operating data does not meet the abnormal operating conditions, the multi-energy system is controlled to operate according to the intraday time-sharing scheduling data within the target preset period.

Images