Intelligent energy consumption distribution method for multi-source energy system

Abstract

The invention discloses an intelligent energy consumption distribution method for a multi-source energy system. The method comprises the steps of energy supply data fusion, dual-channel joint prediction, energy response lag correction, multi-objective optimization and intelligent energy consumption distribution. The invention relates to the technical field of data processing of power management and resource scheduling, and the method comprises the steps: carrying out the unified collection and archiving of operation control data and environment information of power grid energy, distributed photovoltaic energy and an energy storage system, and constructing an energy supply fusion data set; respectively extracting energy availability and user load characteristics by adopting a double-domain attention decoupling fusion joint supply and demand prediction method, and realizing double-channel joint modeling; modeling, estimating and correcting response lags of different energy sources in combination with a dynamic response curve reconstruction method; and further constructing a layered target penalty function containing real-time response error, economic cost, carbon emission and response penalty terms, and embedding the layered target penalty function into a particle swarm optimization algorithm to obtain an energy distribution strategy under multi-target optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of energy system regulation and intelligent optimization, and specifically to an intelligent energy consumption allocation method for a multi-source energy system. Background Art

[0002] Intelligent energy allocation for multi-source energy systems is a dynamic energy scheduling technology based on optimization algorithms. It aims to achieve an optimal balance of economy, reliability, and sustainability while meeting user needs by coordinating the supply ratios of multiple energy sources, including renewable energy, energy storage systems, and traditional power grids, in real time. This method uses load demand as a constraint and combines energy costs, response delay penalties (such as a secondary penalty for energy sources that respond late), and physical limitations of equipment (such as energy storage SOC and grid power caps) to dynamically optimize the allocation of various energy sources. This reduces energy costs, improves response speed, enhances system stability, and promotes the efficient use of clean energy. This approach is suitable for scenarios such as microgrids, smart buildings, and industrial energy management, and is a core means of promoting the intelligent and low-carbon development of energy systems.

[0003] However, existing intelligent energy consumption allocation methods for multi-source energy systems generally adopt control methods guided by static priorities, fixed rules, or simple linear rules, and lack systematic modeling of the heterogeneous characteristics of multi-source energy (such as response speed, availability fluctuations, and scheduling delays). As a result, under dynamic load changes and complex weather disturbances, the system is prone to frequent energy switching, imbalanced supply and demand matching, and uncontrolled response lags. In existing dual-channel joint prediction methods, traditional dual-channel prediction methods usually directly splice energy availability and user load as two input features and input them into a unified network model, failing to effectively identify the differences in information structure and independence of driving logic between the two. This leads to technical problems such as severe feature coupling interference, weak key factor identification ability, and high inaccuracy rate in extreme scenarios in the prediction model. In existing energy response lag correction methods, existing energy scheduling systems usually only use static offset methods or empirical lag correction means to compensate for response delays, failing to fully model the dynamic response curves of equipment under different load conditions, resulting in large lag estimation deviations and failure of early or delayed strategy execution. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent energy consumption allocation method for a multi-source energy system, which includes the following steps:

[0005] Step S1: Energy supply data fusion;

[0006] Step S2: dual-channel joint prediction;

[0007] Step S3: Energy response hysteresis correction;

[0008] Step S4: multi-objective optimization;

[0009] Step S5: Intelligent energy consumption allocation.

[0010] Furthermore, in step S1, the energy supply data fusion is used to collect and integrate multiple energy and environmental information data, specifically to uniformly collect, classify and structure grid energy, distributed photovoltaic energy, energy storage system energy and related environmental variables, and to collect structured operation data and control records for each energy source to obtain an energy supply fusion data set;

[0011] The energy supply fusion data set specifically includes operation control data and external environment data.

[0012] Furthermore, in step S2, the dual-channel joint prediction is used for dual-channel modeling of energy availability and user load. Specifically, based on the energy supply fusion dataset, a dual-domain attention decoupling fusion joint supply and demand prediction method is adopted to perform dual-channel joint prediction to obtain joint energy supply and demand change prediction data, including the following steps:

[0013] Step S21: Dual-channel data construction, specifically, based on the energy supply fusion dataset, extracting external energy availability data as first-channel raw sequence data input, and extracting user load record data as second-channel raw sequence data input, performing normalization processing on the first-channel raw sequence data input and the second-channel raw sequence data input, and introducing a time embedding vector based on a sine and cosine function, thereby constructing dual-channel input data;

[0014] Step S22: constructing a dual-channel feature extraction network, specifically, constructing a dual-channel feature extraction network based on the dual-channel input data, and performing dual-channel feature extraction using the dual-channel feature extraction network to obtain energy availability feature data and user historical load feature data;

[0015] The dual-channel feature extraction network includes an energy availability channel and a user load channel;

[0016] Step S23: Decoupled attention mechanism modeling, specifically, constructing a self-attention module for each of the energy availability feature data and the user historical load feature data to obtain decoupled attention feature data; the decoupled attention feature data specifically includes energy availability decoupled attention features and user historical load decoupled attention features;

[0017] Step S24: Dual-domain fusion residual correction, specifically, concatenating the energy availability decoupled attention feature and the user history load decoupled attention feature and inputting the concatenated features into a multi-layer perceptron module to obtain a fused dual-domain feature, and performing a dual-domain fusion residual correction on the fused dual-domain feature by introducing a cross-channel residual correction function to obtain corrected fusion feature data;

[0018] Step S25: dual-channel joint prediction, specifically, decoding and predicting based on the modified fusion feature data to obtain joint supply and demand prediction data, and obtaining joint energy supply and demand change prediction data by subdividing the joint supply and demand prediction data;

[0019] The combined energy supply and demand change prediction data specifically includes the grid energy availability prediction value, the distributed photovoltaic energy availability prediction value, the energy storage system energy availability prediction value, the user load demand prediction value and the supply and demand change prediction difference.

[0020] Furthermore, in step S3, the energy response lag correction is used to model and correct the response time and efficiency of different energy sources. Specifically, based on the energy supply fusion data set and the joint energy supply and demand change prediction data, a multi-source energy lag correction method combined with dynamic response curve reconstruction is used to perform energy response lag correction to obtain energy correction prediction data, including the following steps:

[0021] Step S31: Response curve modeling, specifically, based on the combined energy supply and demand change prediction data, by collecting the historical control instruction sequence of each energy source and the actual output power at the corresponding time, performing response curve model fitting to obtain energy response curve data; the response curve model fitting specifically refers to fitting the grid energy and energy storage system energy using an improved first-order delayed linear function; and fitting the distributed photovoltaic energy using a saturation improved nonlinear exponential function;

[0022] Step S32: Response time difference estimation, specifically, constructing a hysteresis time window search model and calculating the maximum correlation coefficient between the control signal and the response value of each energy source by sliding the hysteresis step, and calculating the maximum correlation hysteresis value parameter for the grid energy and the energy storage system energy. For the distributed photovoltaic energy, the maximum correlation hysteresis value parameter is estimated using the hysteresis correlation peak positioning algorithm to obtain the optimal response delay estimation parameter;

[0023] Step S33: Lag alignment correction, specifically, performing time alignment processing based on the energy response curve data and the optimal response delay estimation parameter to obtain aligned energy supply prediction data; the time alignment processing is performed by calculating the grid energy and energy storage system energy by inverse linear response model; and the distributed photovoltaic energy is calculated by numerical interpolation approximation solution;

[0024] Step S34: Dynamic correction and fusion, specifically, constructing an aligned comprehensive energy supply prediction vector including grid energy, distributed photovoltaic energy, and energy storage system energy based on the aligned energy supply prediction data, to obtain multi-source energy supply prediction matrix data;

[0025] Step S35: Energy response lag correction, specifically performing energy response lag correction through the response curve modeling, the response time difference estimation, the lag alignment correction and the dynamic correction fusion, outputting the multi-source energy supply prediction matrix data as the energy supply prediction data after energy response lag correction, and obtaining energy correction prediction data.

[0026] Furthermore, in step S4, the multi-objective optimization is used to generate a multi-objective optimal energy allocation strategy. Specifically, based on the energy correction prediction data, an energy allocation particle swarm optimization algorithm combined with a hierarchical objective penalty function is adopted. By constructing an improved hierarchical objective penalty function and embedding it into the fitness evaluation link of the particle swarm optimization, and defining optimization variables, multi-objective optimization is performed to obtain energy allocation strategy reference data;

[0027] The improved hierarchical objective penalty function is constructed by adaptively weighting the real-time response error, economic cost, and total carbon emissions as optimization targets and introducing an energy response lag penalty term;

[0028] The optimization variables specifically include the total load demand distribution, energy supply upper limit and equipment physical constraints;

[0029] The fitness evaluation link embedded in the particle swarm optimization specifically targets a set of energy allocation schemes represented by each particle, calculates the evaluation indicators of each energy allocation scheme in terms of real-time response error, economic cost, and total carbon emissions, and combines the response lag degree of each energy source in the energy correction prediction data with the improved hierarchical objective penalty function to accurately score the fitness of the particles and obtain fitness parameters. By adopting the standard particle swarm optimization algorithm and based on the fitness parameters, the particles are guided to dynamically adjust their position and speed in the search space, gradually approaching the optimal energy consumption allocation strategy under multi-objective trade-offs, and obtaining energy allocation strategy reference data.

[0030] Furthermore, in step S5, the intelligent energy consumption distribution is used to realize intelligent energy regulation and coordination, specifically, according to the energy distribution strategy reference data, the optimal energy supply distribution amount of each type of energy is analyzed, and combined with the operating status parameters of each type of energy, a control data frame is formatted and generated to obtain an intelligent energy consumption distribution control instruction;

[0031] The intelligent energy consumption allocation control instruction specifically includes a timestamp, an energy type, a control power value, a control allowable range value, an instruction type, a delay lag compensation parameter and a control priority level.

[0032] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0033] (1) In view of the fact that the existing intelligent energy consumption allocation methods for multi-source energy systems generally adopt static priority, fixed rules or simple linear rule-guided control methods, and lack systematic modeling of the heterogeneous characteristics of multi-source energy (such as response speed, availability fluctuation, and scheduling delay), the system is prone to frequent energy switching, imbalance in supply and demand matching, and delayed response under dynamic load changes and complex weather disturbances. This solution creatively adopts a comprehensive intelligent method of dual-channel joint prediction, lag correction and multi-objective optimization, and realizes dynamic adjustment of energy response matching strategies based on accurate prediction of future energy supply and demand trends, and balances real-time performance, economy and carbon emissions, thus realizing a multi-source collaborative, dynamic matching and intelligently controllable energy consumption allocation mechanism.

[0034] (2) In the existing dual-channel joint prediction methods, there is a problem that traditional dual-channel prediction methods usually directly take energy availability and user load as two input features and splice them into a unified network model, but fail to effectively identify the differences in information structure and the independence of driving logic between the two, resulting in technical problems such as severe feature coupling interference, weak key factor identification ability, and high inaccuracy rate in extreme scenarios in the prediction model. This solution creatively adopts a joint supply and demand prediction method based on dual-domain attention decoupling and fusion. By constructing independent time series modeling channels and self-attention modules, the key driving factors of energy and load are extracted respectively, and a residual compensation mechanism is introduced for feature fusion, thus achieving a multi-channel collaborative prediction effect with decoupling of prediction features, enhanced fusion, and improved supply and demand fitting accuracy.

[0035] (3) In view of the existing energy response lag correction methods, there is a problem in the existing energy dispatching system. Usually only static offset method or empirical lag correction means are used to compensate for response delay, which fails to fully model the dynamic response curve of the equipment under different load conditions, resulting in large lag estimation deviation and failure of strategy advance or delay execution, which is particularly significant in the photovoltaic fluctuation or energy storage critical SOC range. This solution creatively adopts a multi-source energy lag correction method combined with dynamic response curve reconstruction to correct energy response lag. It establishes multi-type response behavior models for energy sources such as power grid, photovoltaic and energy storage, and combines historical energy supply behavior with control signals to reverse the lag step size, dynamically correct the prediction results, and achieve the goals of accurate response alignment, enhanced timeliness and improved lag adaptive scheduling capabilities;

[0036] (4) In view of the fact that the existing multi-objective optimization and control methods often rely on traditional weighted methods or Pareto solution methods for energy allocation, and fail to combine the response behavior and control delay characteristics of different energy types to construct an optimization target structure with real-time adaptability, resulting in the optimization results being reasonable in theory but poor in actual feasibility, especially the lack of strategic penalty constraints for response-delayed equipment, this scheme creatively adopts an energy allocation particle swarm optimization algorithm combined with a hierarchical objective penalty function. By constructing an improved hierarchical objective penalty function and embedding it into the fitness evaluation link of the particle swarm optimization, and defining optimization variables, multi-objective optimization is performed, thus realizing a multi-source energy allocation optimization mechanism with multi-objective indicator fusion, dynamic penalty for response delay, and enhanced strategy feasibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of a flow chart of a method for improving the performance of intelligent agricultural machinery provided by the present invention;

[0038] Figure 2 Schematic diagram of the process of dual-channel joint prediction in step S2;

[0039] Figure 3 Schematic diagram of the dual-channel structure for dual-channel joint prediction in step S2;

[0040] Figure 4 Schematic diagram of the process of energy response lag correction in step S3.

[0041] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] In the description of the present invention, it should be understood that terms such as "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0044] Example 1, see Figure 1 The present invention provides a method for improving the performance of intelligent agricultural machinery, which includes the following steps:

[0045] Step S1: Energy supply data fusion;

[0046] Step S2: dual-channel joint prediction;

[0047] Step S3: Energy response hysteresis correction;

[0048] Step S4: multi-objective optimization;

[0049] Step S5: Intelligent energy consumption allocation.

[0050] By performing the above operations, it is found that in the existing intelligent energy consumption allocation methods for multi-source energy systems, the intelligent energy consumption allocation methods in the current multi-source energy systems generally adopt a control method guided by static priority, fixed rules or simple linear rules, and lack systematic modeling of the heterogeneous characteristics between multi-source energy sources (such as response speed, availability fluctuation, and scheduling delay). As a result, under dynamic load changes and complex weather disturbances, the system is prone to technical problems such as frequent energy switching, imbalance in supply and demand matching, and delayed response and out of control. This solution creatively adopts a comprehensive intelligent method of dual-channel joint prediction, lag correction and multi-objective optimization, and realizes dynamic adjustment of energy response matching strategies based on accurate prediction of future energy supply and demand trends, and balances real-time performance, economy and carbon emissions, thereby realizing a multi-source collaborative, dynamic matching, and intelligently controllable energy consumption allocation mechanism.

[0051] Example 2, see Figure 1 and Figure 2 In step S1, the energy supply data fusion is used to collect and integrate multiple energy and environmental information data, specifically to uniformly collect, classify and structure grid energy, distributed photovoltaic energy, energy storage system energy and related environmental variables, and to collect structured operation data and control records for each energy source to obtain an energy supply fusion data set;

[0052] The energy supply fusion data set specifically includes operation control data and external environment data;

[0053] The operation control data specifically includes current and historical energy supply power data, energy supply upper and lower limit data, start and stop control instruction data, electricity price data, energy storage charge state data and control mode status data;

[0054] The external environment data specifically includes light intensity, sunshine duration, wind speed, temperature and humidity, weather type, cloud cover index and air quality index.

[0055] Example 3, see Figure 1 、 Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S2, the dual-channel joint prediction is used for dual-channel modeling of energy availability and user load. Specifically, based on the energy supply fusion dataset, a dual-domain attention decoupling fusion joint supply and demand prediction method is adopted to perform dual-channel joint prediction to obtain joint energy supply and demand change prediction data, including the following steps:

[0056] Step S21: Dual-channel data construction, specifically, based on the energy supply fusion dataset, extracting external energy availability data as first-channel raw sequence data input, and extracting user load record data as second-channel raw sequence data input, performing normalization processing on the first-channel raw sequence data input and the second-channel raw sequence data input, and introducing a time embedding vector based on a sine and cosine function, thereby constructing dual-channel input data;

[0057] Step S22: constructing a dual-channel feature extraction network, specifically, constructing a dual-channel feature extraction network based on the dual-channel input data, and performing dual-channel feature extraction using the dual-channel feature extraction network to obtain energy availability feature data and user historical load feature data;

[0058] The dual-channel feature extraction network includes an energy availability channel and a user load channel;

[0059] The energy availability channel performs local feature enhancement through a one-dimensional convolutional network and inputs the local features into a bidirectional long short-term memory network to obtain energy availability feature data. The calculation formula is:

[0060] ;

[0061] Where H s is the energy availability feature data, BiLSTM is a bidirectional long short-term memory network, Conv1D is a one-dimensional convolutional network, X s It is the first channel original sequence data input in the dual channel input data;

[0062] The user load channel is enhanced by identifying periodic fluctuation features through a gated recurrent unit network and introducing a periodic attention module to obtain user historical load feature data. The calculation formula is:

[0063] ;

[0064] Where H d is the user's historical load feature data, PACGRU is a gated recurrent unit network that introduces a periodic attention module, and X d It is the second channel original sequence data input in the dual channel input data;

[0065] Step S23: Decoupled attention mechanism modeling, specifically, constructing a self-attention module for each of the energy availability feature data and the user historical load feature data to obtain decoupled attention feature data; the decoupled attention feature data specifically includes energy availability decoupled attention features and user historical load decoupled attention features;

[0066] Step S24: Dual-domain fusion residual correction, specifically, concatenating the energy availability decoupled attention feature and the user history load decoupled attention feature and inputting the concatenated features into a multi-layer perceptron module to obtain a fused dual-domain feature, and performing a dual-domain fusion residual correction on the fused dual-domain feature by introducing a cross-channel residual correction function to obtain corrected fusion feature data;

[0067] The calculation formula for the fusion of dual-domain features is:

[0068] ;

[0069] Where, It is the fusion of dual domain features, MLP is the multi-layer perceptron module, Z s is the energy availability decoupled attention feature, Z d is the user history load decoupled attention feature, || is the feature concatenation operator;

[0070] The calculation formula of the cross-channel residual correction function is:

[0071] ;

[0072] Where Z fused is to correct the fusion feature data, It is the fusion of dual domain features. is the residual adjustment factor, which is automatically learned through training;

[0073] Step S25: dual-channel joint prediction, specifically, decoding and predicting based on the modified fusion feature data to obtain joint supply and demand prediction data, and obtaining joint energy supply and demand change prediction data by subdividing the joint supply and demand prediction data;

[0074] The combined energy supply and demand change prediction data specifically includes the grid energy availability prediction value, the distributed photovoltaic energy availability prediction value, the energy storage system energy availability prediction value, the user load demand prediction value and the supply and demand change prediction difference.

[0075] By performing the above operations, in the existing dual-channel joint prediction methods, there is a problem that traditional dual-channel prediction methods usually directly take energy availability and user load as two input features and splice them into a unified network model, failing to effectively identify the differences in information structure and independence of driving logic between the two, resulting in severe feature coupling interference in the prediction model, weak key factor identification ability, and high inaccuracy rate in extreme scenarios. Technical problems, this solution creatively adopts a joint supply and demand prediction method of dual-domain attention decoupling and fusion. By constructing independent time series modeling channels and self-attention modules, the key driving factors of energy and load are extracted respectively, and a residual compensation mechanism is introduced for feature fusion, achieving a multi-channel collaborative prediction effect with decoupling of prediction features, enhanced fusion and improved supply and demand fitting accuracy.

[0076] Example 4, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S3, the energy response lag correction is used to model and correct the response time and efficiency of different energy sources. Specifically, based on the energy supply fusion data set and the joint energy supply and demand change prediction data, a multi-source energy lag correction method combined with dynamic response curve reconstruction is used to perform energy response lag correction to obtain energy correction prediction data, including the following steps:

[0077] Step S31: Response curve modeling, specifically, based on the combined energy supply and demand change prediction data, by collecting the historical control instruction sequence of each energy source and the actual output power at the corresponding time, performing response curve model fitting to obtain energy response curve data; the response curve model fitting specifically refers to fitting the grid energy and energy storage system energy using an improved first-order delayed linear function; and fitting the distributed photovoltaic energy using a saturation improved nonlinear exponential function;

[0078] The improved first-order delayed linear function specifically introduces a dynamic response gain adjustment parameter and a control signal fluctuation suppression factor to improve the anti-interference capability. The calculation formula is:

[0079] ;

[0080] Where P(t) is the output power of the improved first-order delayed linear function, K d is the dynamic response gain parameter, which is calculated by automatic fitting. t is the current time index, and t0 is the time index of the control signal. is the lag time parameter, which is used to indicate the time offset corresponding to the energy. is the system response time constant, is the fluctuation suppression factor, the default value is set to 0.1, and the maximum value is not higher than 0.3. u(t) is the original input signal of the control signal. It is the actual input signal that takes effect on the system after the backtracking lag time of the control signal at the current time t;

[0081] The saturation-improved nonlinear exponential function is specifically improved by constructing a model with saturation gating and power cap fitting mechanisms to improve the nonlinear change modeling capability. The calculation formula is:

[0082] ;

[0083] Where, P PV (t) is a nonlinear exponential function of saturation improvement, P max is the rated maximum power, The overall is the power cap fitting term, is the response steepness coefficient, I sun (t) is the light intensity parameter, is the lag time parameter, I th is the response threshold, is the saturation gating term;

[0084] Preferably, the calculation formula of the saturation gating term is:

[0085] ;

[0086] Where, t sunrise is the sunrise time, t sunset It’s sunset time;

[0087] Step S32: Response time difference estimation, specifically, constructing a hysteresis time window search model and calculating the maximum correlation coefficient between the control signal and the response value of each energy source by sliding the hysteresis step, and calculating the maximum correlation hysteresis value parameter for the grid energy and the energy storage system energy. For the distributed photovoltaic energy, the maximum correlation hysteresis value parameter is estimated using the hysteresis correlation peak positioning algorithm to obtain the optimal response delay estimation parameter;

[0088] Step S33: Lag alignment correction, specifically, performing time alignment processing based on the energy response curve data and the optimal response delay estimation parameter to obtain aligned energy supply prediction data; the time alignment processing is performed by calculating the grid energy and energy storage system energy by inverse linear response model; and the distributed photovoltaic energy is calculated by numerical interpolation approximation solution;

[0089] The linear response model specifically refers to the data model corresponding to the improved first-order delayed linear function;

[0090] The numerical interpolation approximation solution specifically refers to dynamic time interpolation based on the saturation-improved nonlinear exponential function;

[0091] Step S34: Dynamic correction and fusion, specifically, constructing an aligned comprehensive energy supply prediction vector including grid energy, distributed photovoltaic energy, and energy storage system energy based on the aligned energy supply prediction data, to obtain multi-source energy supply prediction matrix data;

[0092] The calculation formula of the multi-source energy supply prediction matrix is:

[0093] ;

[0094] Where, is the multi-source energy supply prediction matrix, is the aligned energy supply forecast data of the grid energy, where t is the time index, is the time offset of the grid energy, It is the aligned energy supply forecast data of distributed photovoltaic energy. is the time offset of distributed photovoltaic energy, It is the energy supply forecast data of the energy storage system. is the time offset of the energy storage system energy;

[0095] Step S35: Energy response lag correction, specifically performing energy response lag correction through the response curve modeling, the response time difference estimation, the lag alignment correction and the dynamic correction fusion, outputting the multi-source energy supply prediction matrix data as the energy supply prediction data after energy response lag correction, and obtaining energy correction prediction data.

[0096] By performing the above operations, in order to address the technical problems in existing energy response lag correction methods, the existing energy scheduling system usually only adopts static offset method or empirical lag correction means to compensate for response delay, fails to fully model the dynamic response curve of the equipment under different load conditions, resulting in large lag estimation deviation and failure of strategy advance or delayed execution, which is particularly significant in the photovoltaic fluctuation or critical SOC range of energy storage. This solution creatively adopts a multi-source energy lag correction method combined with dynamic response curve reconstruction to correct energy response lag, establishes multi-type response behavior models for energy sources such as power grid, photovoltaic and energy storage, and combines historical energy supply behavior with control signal to reverse the lag step, dynamically corrects the prediction results, and achieves the goals of precise response alignment, enhanced timeliness and improved lag adaptive scheduling capabilities.

[0097] Example 4, see Figure 1This embodiment is based on the above embodiment. In step S4, the multi-objective optimization is used to generate a multi-objective optimal energy allocation strategy. Specifically, based on the energy correction prediction data, an energy allocation particle swarm optimization algorithm combined with a hierarchical objective penalty function is adopted. By constructing an improved hierarchical objective penalty function and embedding it into the fitness evaluation link of the particle swarm optimization, and defining optimization variables, multi-objective optimization is performed to obtain energy allocation strategy reference data.

[0098] The improved hierarchical objective penalty function is constructed by adaptively weighting the real-time response error, economic cost, and total carbon emissions as optimization targets, and introducing an energy response lag penalty term. The calculation formula is:

[0099] ;

[0100] Where, F total is the improved hierarchical objective penalty function, is the real-time response error weight, F1 is the real-time response error target, is the economic cost weight, F2 is the economic cost target, is the total carbon emission weight, F3 is the total carbon emission target, K is the total number of energy types, k is the energy index, is the hysteresis sensitivity weight of the k-th energy source, P delay (k) is the energy lag penalty term, and the specific calculation formula is: , where n k is the penalty coefficient of the kth energy, T d is the difference between the actual response time and the required time, T e It is the maximum allowed response time threshold;

[0101] The optimization variables specifically include the total load demand distribution, energy supply upper limit and equipment physical constraints;

[0102] The fitness evaluation link embedded in the particle swarm optimization specifically targets a set of energy allocation schemes represented by each particle, calculates the evaluation indicators of each energy allocation scheme in terms of real-time response error, economic cost, and total carbon emissions, and combines the response lag degree of each energy source in the energy correction prediction data with the improved hierarchical objective penalty function to accurately score the fitness of the particles and obtain fitness parameters. By adopting the standard particle swarm optimization algorithm and based on the fitness parameters, the particles are guided to dynamically adjust their position and speed in the search space, gradually approaching the optimal energy consumption allocation strategy under multi-objective trade-offs, and obtaining energy allocation strategy reference data.

[0103] By performing the above operations, in view of the fact that in the existing multi-objective optimization and control methods, current multi-objective optimization methods often rely on traditional weighted methods or Pareto solution methods for energy allocation, and fail to combine the response behavior and control delay characteristics of different energy types to construct an optimization target structure with real-time adaptability, resulting in the optimization results being theoretically reasonable but poor in actual executability, especially the lack of strategy penalty constraints for response-delayed equipment. This scheme creatively adopts an energy allocation particle swarm optimization algorithm combined with a hierarchical objective penalty function. By constructing an improved hierarchical objective penalty function and embedding it into the fitness evaluation link of the particle swarm optimization, and defining optimization variables for multi-objective optimization, a multi-source energy allocation optimization mechanism with multi-objective indicator fusion, dynamic penalty for response delay, and enhanced strategy executability is realized.

[0104] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the intelligent energy consumption distribution is used to realize intelligent energy regulation and coordination. Specifically, the optimal energy supply distribution amount of each type of energy is analyzed based on the energy distribution strategy reference data, and the control data frame is formatted and generated in combination with the operating status parameters of each type of energy to obtain the intelligent energy consumption distribution control instruction.

[0105] The intelligent energy consumption allocation control instruction specifically includes a timestamp, an energy type, a control power value, a control allowable range value, an instruction type, a delay lag compensation parameter and a control priority level.

[0106] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0107] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0108] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An intelligent energy consumption allocation method for a multi-source energy system, characterized by: The method comprises the following steps: Step S1: energy supply data fusion to obtain an energy supply fusion data set; Step S2: dual-channel joint prediction, based on the energy supply fusion dataset, adopting the dual-domain attention decoupling fusion joint supply and demand prediction method to perform dual-channel joint prediction and obtain joint energy supply and demand change prediction data; Step S3: Energy response lag correction, based on the energy supply fusion data set and the joint energy supply and demand change prediction data, adopting a multi-source energy lag correction method combined with dynamic response curve reconstruction to perform energy response lag correction to obtain energy correction prediction data; Step S4: Multi-objective optimization: Based on the energy correction prediction data, an energy allocation particle swarm optimization algorithm combined with a hierarchical objective penalty function is used. An improved hierarchical objective penalty function is constructed and embedded into the fitness evaluation link of the particle swarm optimization, and optimization variables are defined to perform multi-objective optimization to obtain energy allocation strategy reference data. The improved hierarchical objective penalty function is specifically constructed by adaptively weighting the real-time response error, economic cost, and total carbon emissions as optimization objectives, and introducing an energy response lag penalty term. The optimization variables specifically include the total load demand allocation, the energy supply upper limit, and the equipment physical constraints. Step S5: Intelligent energy consumption allocation, formatting and generating a control data frame to obtain an intelligent energy consumption allocation control instruction.

2. The intelligent energy consumption allocation method for a multi-source energy system according to claim 1, characterized in that: In step S1, the energy supply data fusion is used to collect and integrate multiple energy and environmental information data, specifically to uniformly collect, classify and structure grid energy, distributed photovoltaic energy, energy storage system energy and related environmental variables, and to collect structured operation data and control records for each energy source to obtain an energy supply fusion data set; The energy supply fusion data set specifically includes operation control data and external environment data.

3. The intelligent energy consumption allocation method for a multi-source energy system according to claim 2, characterized in that: In step S2, the dual-channel joint prediction is used for dual-channel modeling of energy availability and user load. Specifically, based on the energy supply fusion dataset, a dual-domain attention decoupling fusion joint supply and demand prediction method is used to perform dual-channel joint prediction to obtain joint energy supply and demand change prediction data, including the following steps: Step S21: Dual-channel data construction, specifically, based on the energy supply fusion dataset, extracting external energy availability data as first-channel raw sequence data input, and extracting user load record data as second-channel raw sequence data input, performing normalization processing on the first-channel raw sequence data input and the second-channel raw sequence data input, and introducing a time embedding vector based on a sine and cosine function, thereby constructing dual-channel input data; Step S22: constructing a dual-channel feature extraction network, specifically, constructing a dual-channel feature extraction network based on the dual-channel input data, and performing dual-channel feature extraction using the dual-channel feature extraction network to obtain energy availability feature data and user historical load feature data; The dual-channel feature extraction network includes an energy availability channel and a user load channel; Step S23: Decoupled attention mechanism modeling, specifically, constructing a self-attention module for each of the energy availability feature data and the user historical load feature data to obtain decoupled attention feature data; the decoupled attention feature data specifically includes energy availability decoupled attention features and user historical load decoupled attention features; Step S24: Dual-domain fusion residual correction, specifically, concatenating the energy availability decoupled attention feature and the user history load decoupled attention feature and inputting the concatenated features into a multi-layer perceptron module to obtain a fused dual-domain feature, and performing a dual-domain fusion residual correction on the fused dual-domain feature by introducing a cross-channel residual correction function to obtain corrected fusion feature data; Step S25: dual-channel joint prediction, specifically decoding and predicting based on the modified fusion feature data to obtain joint supply and demand prediction data, and obtaining joint energy supply and demand change prediction data by subdividing the joint supply and demand prediction data.

4. The intelligent energy consumption allocation method for a multi-source energy system according to claim 3, characterized in that: In step S2, the combined energy supply and demand change prediction data specifically includes the grid energy available prediction value, the distributed photovoltaic energy available prediction value, the energy storage system energy available prediction value, the user load demand prediction value and the supply and demand change prediction difference.

5. The intelligent energy consumption allocation method for a multi-source energy system according to claim 4, characterized in that: In step S3, the energy response lag correction is used to model and correct the response time and efficiency of different energy sources. Specifically, based on the energy supply fusion data set and the joint energy supply and demand change prediction data, a multi-source energy lag correction method combined with dynamic response curve reconstruction is used to perform energy response lag correction to obtain energy correction prediction data, including the following steps: Step S31: Response curve modeling, specifically, based on the combined energy supply and demand change prediction data, by collecting the historical control instruction sequence of each energy source and the actual output power at the corresponding time, performing response curve model fitting to obtain energy response curve data; the response curve model fitting specifically refers to fitting the grid energy and energy storage system energy using an improved first-order delayed linear function; and fitting the distributed photovoltaic energy using a saturation improved nonlinear exponential function; The improved first-order delayed linear function specifically introduces a dynamic response gain adjustment parameter and a control signal fluctuation suppression factor to improve the anti-interference capability; The saturation-improved nonlinear exponential function specifically improves the nonlinear change modeling capability by constructing a model with saturation gating and power cap fitting mechanisms; Step S32: Response time difference estimation, specifically, constructing a hysteresis time window search model and calculating the maximum correlation coefficient between the control signal and the response value of each energy source by sliding the hysteresis step, and calculating the maximum correlation hysteresis value parameter for the grid energy and the energy storage system energy. For the distributed photovoltaic energy, the maximum correlation hysteresis value parameter is estimated using the hysteresis correlation peak positioning algorithm to obtain the optimal response delay estimation parameter; Step S33: Lag alignment correction, specifically, performing time alignment processing based on the energy response curve data and the optimal response delay estimation parameter to obtain aligned energy supply prediction data; the time alignment processing is performed by calculating the grid energy and energy storage system energy by inverse linear response model; and the distributed photovoltaic energy is calculated by numerical interpolation approximation solution; The linear response model specifically refers to the data model corresponding to the improved first-order delayed linear function; The numerical interpolation approximation solution specifically refers to dynamic time interpolation based on the saturation-improved nonlinear exponential function; Step S34: Dynamic correction and fusion, specifically, constructing an aligned comprehensive energy supply prediction vector including grid energy, distributed photovoltaic energy, and energy storage system energy based on the aligned energy supply prediction data, to obtain multi-source energy supply prediction matrix data; Step S35: Energy response lag correction, specifically performing energy response lag correction through the response curve modeling, the response time difference estimation, the lag alignment correction and the dynamic correction fusion, outputting the multi-source energy supply prediction matrix data as the energy supply prediction data after energy response lag correction, and obtaining energy correction prediction data.

6. The intelligent energy consumption allocation method for a multi-source energy system according to claim 5, characterized in that: In step S4, the multi-objective optimization is used to generate a multi-objective optimal energy allocation strategy. Specifically, based on the energy correction prediction data, an energy allocation particle swarm optimization algorithm combined with a hierarchical objective penalty function is adopted. By constructing an improved hierarchical objective penalty function and embedding it into the fitness evaluation link of the particle swarm optimization, and defining optimization variables, multi-objective optimization is performed to obtain energy allocation strategy reference data.

7. The intelligent energy consumption allocation method for a multi-source energy system according to claim 6, characterized in that: The fitness evaluation link embedded in the particle swarm optimization specifically targets a set of energy allocation schemes represented by each particle, calculates the evaluation indicators of each energy allocation scheme in terms of real-time response error, economic cost, and total carbon emissions, and combines the response lag degree of each energy source in the energy correction prediction data with the improved hierarchical objective penalty function to accurately score the fitness of the particles and obtain fitness parameters. By adopting the standard particle swarm optimization algorithm and based on the fitness parameters, the particles are guided to dynamically adjust their position and speed in the search space, gradually approaching the optimal energy consumption allocation strategy under multi-objective trade-offs, and obtaining energy allocation strategy reference data.

8. The intelligent energy consumption allocation method for a multi-source energy system according to claim 6, characterized in that: In step S5, the intelligent energy consumption allocation is used to realize intelligent energy regulation and coordination. Specifically, the optimal energy supply allocation of each type of energy is analyzed based on the energy allocation strategy reference data, and the control data frame is formatted and generated in combination with the operating status parameters of each type of energy to obtain the intelligent energy consumption allocation control instruction; The intelligent energy consumption allocation control instruction specifically includes a timestamp, an energy type, a control power value, a control allowable range value, an instruction type, a delay lag compensation parameter and a control priority level.

Images