A Smart Optimization Method for Energy Management Based on Population Algorithm

Abstract

This invention discloses an intelligent optimization method for energy management based on a swarm algorithm, comprising the following steps: S1, collecting multi-source time-series data and adjustable equipment parameters from the energy management system to generate a prediction sequence, constituting a scheduling space; S2, calculating a risk index sequence to form a risk map; S3, constructing a mapping structure between scheduling schemes and gap fields; S4, constructing a coupled coding structure between population individuals and gap fields to form an initial population cluster; S5, forming an updated population cluster using an improved HHO algorithm; S6, performing selection, replication, and elimination operations to form a new generation population cluster; S7, performing convergence determination, updating the risk map and prediction sequence, and completing the optimization closed loop. This invention enables risk perception, dynamic adjustment, and global optimization of complex energy systems across multiple time scales, improving the economy, energy efficiency, and operational safety of the energy management process.

Description

Technical Field

[0001] This invention relates to the field of energy management optimization technology, and in particular to an intelligent optimization method for energy management based on swarm algorithms. Background Technology

[0002] With the rapid increase in the proportion of new energy sources and the widespread integration of distributed energy in various scenarios, energy management systems face greater uncertainty and volatility at the operation and scheduling level. Operation optimization places higher demands on the ability to coordinate power generation equipment, energy storage units, and adjustable loads. Existing energy management optimization methods mostly rely on static scheduling models or heuristic search strategies based on fixed preset rules, typically using forecast data at a single time scale as the basis for decision-making. However, renewable energy output is highly stochastic, and load demand exhibits superimposed fluctuations across multiple time scales, making single-scale forecast-driven scheduling strategies difficult to accurately reflect future operational risks. In practical applications, scheduling systems often fail to identify gap risks during critical periods in a timely manner, leading to insufficient reserve capacity, premature or delayed energy storage scheduling, and inefficient adjustable load response.

[0003] In swarm intelligence algorithm scheduling frameworks, traditional algorithms are typically based on general random search or iterative mechanisms based on a single fitness rule, lacking deep coupling with the risk structure of energy systems. Represented by common particle swarm optimization (PSO), genetic algorithms, and unmodified eagle swarm optimization (ASSO), their search behavior is insufficiently coupled with the multidimensional constraints of energy systems, making it difficult for these algorithms to adapt to the cumulative effects of uncertainty and time-segmented risk changes in the scenario. Furthermore, the jump update and convergence mechanisms of existing algorithms cannot be dynamically adjusted according to the strength of risk, easily getting trapped in local optima in high-risk areas or exhibiting excessive jumps in low-risk areas, leading to unstable scheduling schemes and failing to meet the requirements of safety, economy, and energy efficiency synergistic optimization for complex energy systems. Therefore, there is an urgent need for an energy management optimization method that can integrate multi-timescale prediction information, express the relationship between the scheduling space and the risk map, and possess adaptive transition capabilities to improve the accuracy, stability, and global optimization capabilities of scheduling schemes in complex time-varying environments.

[0004] Therefore, how to provide an intelligent optimization method for energy management based on swarm algorithms is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an intelligent optimization method for energy management based on a swarm algorithm. This invention generates a prediction sequence by preprocessing multi-source time-series data, constructing a scheduling space covering the scheduling cycle. A risk map is then formed based on the prediction sequence. A gap field is constructed between the scheduling space and the risk map, creating a mapping structure that records field strength information and field strength changes. An improved HHO algorithm is used to generate basic update quantities, segmented gradient energy difference processing, and segmented transition update quantities for the initial population within the scheduling space. Furthermore, selection, replication, and elimination operations are performed based on the field strength values ​​corresponding to different time scales in the gap field to form a new generation of populations. When convergence conditions are met, a scheduling scheme is output, and the prediction sequence and risk map are updated based on validation data, forming an optimization closed loop. This invention improves the quality of scheduling scheme search and enhances the response capability to risk changes through the synergistic application of the swarm algorithm and the gap field, possessing advantages such as strong full-cycle coordination, high optimization accuracy, and good operational stability.

[0006] An intelligent optimization method for energy management based on a swarm optimization algorithm according to an embodiment of the present invention includes the following steps:

[0007] S1. Collect multi-source time series data and adjustable equipment parameters from the energy management system; perform preprocessing on the multi-source time series data to generate a prediction sequence; construct a scheduling space based on the adjustable equipment parameters;

[0008] S2. Calculate the risk indicator sequence based on the predicted sequence and form a risk map in chronological order;

[0009] S3. Construct a gap field based on the scheduling space and risk map. By mapping the scheduling schemes in the scheduling space to the risk map to form field strength values, extract the field gradient direction and field gradient magnitude, and construct the mapping structure between the scheduling schemes and the gap field.

[0010] S4. Based on the mapping structure, an initial population is generated in the scheduling space, and a coupled coding structure between the population individuals and the gap field is constructed to form the initial population.

[0011] S5. Based on the initial population, a basic update amount is generated for the population individuals using an improved HHO algorithm; a gradient energy difference is generated based on the field strength value and field gradient direction of the gap field. The improved HHO algorithm introduces a piecewise transition mechanism to process the basic update amount through the gradient energy difference, forming an updated population.

[0012] S6. Based on the field strength values ​​corresponding to the updated population and gap field at different time scales, perform selection, replication and elimination operations to form a new generation of population;

[0013] S7. Based on the new generation of population and gap field, perform convergence judgment. If the convergence condition is met, output the scheduling scheme, execute and verify the scheduling scheme and collect verification data, update the risk map and prediction sequence, and complete the optimization closed loop.

[0014] Optionally, the preprocessing of the multi-source time series data to generate the prediction sequence specifically includes:

[0015] The multi-source time series data includes load series, renewable power output series, energy storage charge series, time-of-use electricity price series, and meteorological series;

[0016] Preprocessing is performed on the multi-source time prediction series to generate load prediction series, renewable energy output prediction series, energy storage load prediction series, time-of-use electricity price prediction series, and meteorological prediction series; prediction series are generated through the time series prediction process; the prediction series include hourly prediction series, daily prediction series, and weekly prediction series.

[0017] Optionally, the configuration of the scheduling space based on adjustable device parameters specifically includes:

[0018] The adjustable equipment parameters include power generation equipment parameters, energy storage equipment parameters, and adjustable load parameters;

[0019] The parameters of the power generation equipment include the rated output, minimum output, maximum ramp rate, minimum start-up time and minimum shutdown time of the power generation equipment. Based on the parameters of the power generation equipment, the upper and lower bounds and step intervals of the power generation output variable are set during the dispatch period.

[0020] Energy storage equipment parameters include energy storage capacity, maximum charging power, maximum discharging power, upper and lower limits of state of charge, and charging and discharging efficiency parameters. Based on the energy storage equipment parameters, the value range of energy storage charging variables and energy storage discharging variables and state of charge constraints are set during the scheduling period.

[0021] Adjustable load parameters include adjustable load power range, allowable peak shifting period, continuous adjustment duration limit and recovery power limit. Based on the adjustable load parameters, the value range and adjustment granularity of the load adjustment variables are set within the scheduling period.

[0022] By combining power generation output variables, energy storage charging variables, energy storage discharging variables, and load adjustment variables within the scheduling period, a set of scheduling variables is formed, constituting a scheduling space covering the scheduling cycle.

[0023] Optionally, S2 specifically includes:

[0024] The power difference sequence is calculated based on the load forecast sequence and the renewable energy output forecast sequence, and the instantaneous gap index for the corresponding time period is generated; the cumulative gap index is generated based on the cumulative result of the instantaneous gap index within the scheduling cycle; the fluctuation index is generated based on the change amplitude of the power difference sequence between adjacent time periods; and the uncertainty index is generated based on the deviation relationship between the forecast sequences.

[0025] The instantaneous gap indicator, cumulative gap indicator, volatility indicator, and uncertainty indicator are arranged in chronological order to form a risk indicator sequence. The risk indicator sequence is then organized according to the scheduling period and time scale to generate a risk map.

[0026] Optionally, S3 specifically includes:

[0027] A gap field is constructed based on the scheduling space and risk map. According to the set of scheduling variables corresponding to each scheduling scheme in the scheduling space within the scheduling cycle, the power generation output variable, energy storage charging variable, energy storage discharging variable, load adjustment variable, instantaneous gap index, cumulative gap index, fluctuation index, and uncertainty index are combined and processed in the order of the scheduling cycle to form a risk superposition sequence corresponding to the scheduling scheme.

[0028] The risk superposition sequence is aggregated according to the scheduling cycle to form the field strength value corresponding to the scheduling scheme.

[0029] The field strength values ​​of all scheduling schemes are arranged in the order of scheduling space to form a field strength distribution structure. Differential processing is performed on the differences between the power generation output variable dimension, energy storage charging variable dimension, energy storage discharging variable dimension and load adjustment variable dimension in the field strength distribution structure to form a field strength change structure.

[0030] The field strength distribution structure, field strength change structure and scheduling scheme are linked in the order of scheduling cycle to form a gap field. The gap field records the field strength information and field strength change information.

[0031] Based on the set of scheduling variables corresponding to the scheduling scheme within the scheduling period, the field strength value change corresponding to the scheduling scheme is extracted from the gap field, and the field gradient direction and field gradient magnitude corresponding to the scheduling scheme are determined based on the field strength value change.

[0032] The scheduling scheme, field strength value, field gradient direction and field gradient magnitude are associated according to the scheduling scheme index to form a mapping structure between the scheduling scheme and the gap field.

[0033] Optionally, S4 specifically includes:

[0034] The scheduling schemes in the scheduling space are segmented according to the field strength values ​​recorded in the mapping structure, and a scheduling scheme is selected in each field strength value interval to form a set of candidate scheduling schemes.

[0035] For each scheduling scheme in the candidate scheduling scheme set, the scheduling variable set is read within the scheduling period, and the corresponding field strength value, field gradient direction and field gradient amplitude are read from the gap field. The power generation output variable, energy storage charging variable, energy storage discharging variable, load adjustment variable, field strength value, field gradient direction and field gradient amplitude are spliced ​​in time order to form the population individual encoding vector.

[0036] Assign individual identifiers and time index identifiers to the population individual coding vectors, and combine the individual identifiers, time index identifiers, and population individual coding vectors to form population individual data units;

[0037] All individual data units of the population are arranged according to their individual identifiers to form the initial population cluster;

[0038] In the initial population, scheduling variable indexes are established for the set of scheduling variables according to the scheduling cycle order. Field parameter indexes are established for the field strength value, field gradient direction, and field gradient magnitude. The scheduling variable indexes and field parameter indexes are associated within the individual data units of the population to form a coupled coding structure between the individual population and the gap field.

[0039] Optionally, the step of generating basic update quantities for individuals in the population based on the initial population using an improved HHO algorithm specifically involves:

[0040] In the initial population, the location data of each individual population is read, and the eagle population location vector is constructed in the scheduling space. The power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable are recorded in the order of scheduling cycle to form the location vector.

[0041] The average position vector is calculated based on the position vectors of all individuals in the population within the scheduling space, and the average position vector is used as the population center position data.

[0042] Read the location vector of the scheduling scheme with the minimum field strength value in the field strength distribution structure within the scheduling space, and record the location vector of the scheduling scheme as the target location vector;

[0043] Perform a difference operation on the position vector and the target position vector to form a difference vector. Perform vector multiplication on the difference vector and the random distribution factor to form the basic jump component.

[0044] Within the variable dimension of the scheduling space, a linear combination process is performed on the position vector and the group center position data according to a preset scaling factor to form a collaborative contraction component;

[0045] Perform a vector superposition process on the basic jump component and the cooperative contraction component to form the basic update quantity;

[0046] Within the scheduling cycle, the basic update quantity forms update components for the variable dimensions corresponding to the power generation output variable, energy storage charging variable, energy storage discharging variable, and load adjustment variable. The update components are compared with the upper and lower bounds of the scheduling space. Components exceeding the boundary are truncated to the boundary of the scheduling space to form the basic update quantity after boundary constraints.

[0047] Optionally, the gradient energy difference is generated based on the field strength value and field gradient direction of the gap field; the gradient energy difference is compared with a first preset difference threshold and a second preset difference threshold, and divided into high difference segment, medium difference segment and low difference segment; in the high difference segment, the basic update amount is subjected to jump amplification processing along the negative gradient direction; in the medium difference segment, the basic update amount is subjected to direction shift processing; in the low difference segment, the basic update amount is subjected to jump suppression processing and enters encirclement update, forming a segmented jump update amount; the position data of the population individuals and historical field strength records are updated based on the segmented jump update amount to form an updated population group, specifically:

[0048] In the initial population, the location data and historical field strength records of each individual population are read, and the corresponding field strength value and field gradient direction are read from the gap field.

[0049] Construct a pre-updated position along the field gradient direction, query the field strength value in the gap field for the pre-updated position, and define the difference between the field strength value of the pre-updated position and the corresponding field strength value of the position data as the gradient energy difference;

[0050] The gradient energy difference is compared with the first preset difference threshold and the second preset difference threshold. If the gradient energy difference is greater than the second preset difference threshold, it is recorded as a high difference segment. If the gradient energy difference is between the first preset difference threshold and the second preset difference threshold, it is recorded as a medium difference segment. If the gradient energy difference is less than the first preset difference threshold, it is recorded as a low difference segment.

[0051] In the opposite direction of the field gradient direction along the elevation difference segment, an amplification factor is superimposed on the basic update amount to generate an amplified jump component. The amplified jump component is then vector-superimposed with the basic update amount to form the segmented transition update amount of the elevation difference segment.

[0052] In the intermediate difference segment, an orthogonal offset direction is constructed based on the field gradient direction, and the offset component is superimposed on the basic update amount to form the segmented transition update amount of the intermediate difference segment.

[0053] In the low-difference segment, a shrinkage coefficient is superimposed on the basic update amount to generate a jump suppression component. The jump suppression component is then vector-superimposed on the basic update amount, and the target scheduling scheme position data is introduced according to the encirclement update strategy to form the low-difference segment segmented jump update amount.

[0054] The segmented transition update amounts for the high-difference segment, the medium-difference segment, and the low-difference segment are written into the population individuals according to the difference segment identifier corresponding to the gradient energy difference, forming the segmented transition update amounts;

[0055] The position data of individuals in the population within the scheduling space are updated based on the segmented transition update amount. The new field strength value is queried in the gap field for the updated position, and the new field strength value is appended to the historical field strength record to form an updated population.

[0056] Optionally, S6 specifically includes:

[0057] Economic, energy efficiency, and safety indicators are calculated based on the set of scheduling variables of individuals in the updated population during the scheduling cycle.

[0058] Read the field strength values ​​corresponding to different time scales from the gap field, and generate risk weights based on the field strength values;

[0059] The economic, energy efficiency, and safety indicators are weighted and integrated according to risk weights to form a fitness value.

[0060] The updated population is sorted, replicated, and eliminated according to its fitness value to form a new generation of population.

[0061] Optionally, S7 specifically includes:

[0062] The fitness values ​​of individuals in the new generation population are read, sorted, and fitness change and iteration round data are generated. The fitness change is compared with a preset convergence threshold, and the iteration rounds are compared with a preset minimum iteration round and a preset maximum iteration round. The convergence condition is determined based on the comparison results.

[0063] When the convergence condition is met, the individual with the highest fitness value is selected from the new generation population. A scheduling scheme is formed based on the set of scheduling variables recorded by the individual population. Multi-source time series data is collected during the scheduling period to record economic data and operational data, forming validation data.

[0064] The multi-source time series data is updated based on the validation data, a revised prediction series is generated, the risk indicator series and risk map are updated, and the validation data is written into the historical operation data set to form an optimization closed loop.

[0065] If the convergence condition is not met, the new generation population is used as the updated population, and the update process based on the improved HHO algorithm, as well as the selection, replication and elimination operations based on fitness values, are continued in combination with the gap field.

[0066] The beneficial effects of this invention are:

[0067] This invention preprocesses multi-source time series data within a scheduling period to generate a prediction sequence, forming a scheduling space covering the scheduling period. It then combines the prediction sequence with risk indicator sequences to form a risk map. A gap field is constructed between the scheduling space and the risk map, and field strength information and its changes are recorded. The field strength values ​​and gradient directions in the gap field guide the generation of basic update quantities and segmented transition update quantities based on the initial population, effectively improving the consistency of search direction and convergence stability of the scheduling scheme under different risk levels. Furthermore, this invention calculates fitness values ​​within the updated population based on economic, energy efficiency, and safety indicators. This invention combines risk weights to perform sorting, replication, and elimination operations, enabling dynamic screening and optimization of scheduling schemes at different time scales and corresponding risk levels. This addresses the problems of insufficient response to risk fluctuations, weak full-cycle coordination capabilities, and susceptibility to local optima in traditional scheduling methods. When convergence conditions are met, the invention outputs a scheduling scheme based on the optimal population individuals and updates the prediction sequence and risk map through verification data, forming an optimization closed loop. This achieves accuracy, stability, and continuous optimization capabilities of the scheduling scheme in actual operation, resulting in strong risk response capabilities, high scheduling accuracy, fast convergence speed, and good full-cycle optimization effects. Attached Figure Description

[0068] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0069] Figure 1 This is a flowchart of an intelligent optimization method for energy management based on a swarm algorithm proposed in this invention;

[0070] Figure 2 This is a schematic diagram of the structure of the improved HHO algorithm proposed in this invention;

[0071] Figure 3 This is a data flow diagram of an intelligent optimization method for energy management based on a swarm algorithm proposed in this invention. Detailed Implementation

[0072] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0073] refer to Figure 1-3 A smart optimization method for energy management based on swarm optimization includes the following steps:

[0074] S1. Collect multi-source time series data and adjustable equipment parameters from the energy management system; perform preprocessing on the multi-source time series data to generate a prediction sequence; construct a scheduling space based on the adjustable equipment parameters;

[0075] S2. Calculate the risk indicator sequence based on the predicted sequence and form a risk map in chronological order;

[0076] S3. Construct a gap field based on the scheduling space and risk map. By mapping the scheduling schemes in the scheduling space to the risk map to form field strength values, extract the field gradient direction and field gradient magnitude, and construct the mapping structure between the scheduling schemes and the gap field.

[0077] S4. Based on the mapping structure, an initial population is generated in the scheduling space, and a coupled coding structure between the population individuals and the gap field is constructed to form the initial population.

[0078] S5. Based on the initial population, a basic update amount is generated for each individual in the population using an improved HHO algorithm. A gradient energy difference is generated based on the field strength value and field gradient direction of the gap field. The improved HHO algorithm introduces a segmented transition mechanism to process the basic update amount through the gradient energy difference, forming an updated population. The segmented transition mechanism includes comparing the gradient energy difference with a first preset difference threshold and a second preset difference threshold, dividing it into a high difference segment, a medium difference segment, and a low difference segment. In the high difference segment, a jump amplification process is performed on the basic update amount along the negative gradient direction. In the medium difference segment, a direction offset process is performed on the basic update amount. In the low difference segment, a jump suppression process is performed on the basic update amount, and an encirclement update is entered, forming a segmented transition update amount. The position data and historical field strength records of each individual in the population are updated based on the segmented transition update amount, forming an updated population.

[0079] S6. Based on the field strength values ​​corresponding to the updated population and gap field at different time scales, perform selection, replication and elimination operations to form a new generation of population;

[0080] S7. Based on the new generation of population and gap field, perform convergence judgment. If the convergence condition is met, output the scheduling scheme, execute and verify the scheduling scheme and collect verification data, update the risk map and prediction sequence, and complete the optimization closed loop.

[0081] In this embodiment, the preprocessing of multi-source time series data to generate a predicted sequence specifically includes:

[0082] The multi-source time series data includes load series, renewable power output series, energy storage charge series, time-of-use electricity price series, and meteorological series;

[0083] The load sequence is processed by missing segment completion, noise segment smoothing, and time alignment to generate a load prediction sequence.

[0084] The renewable energy output sequence is subjected to anomaly peak removal, trend segmentation, and scale normalization processes to generate a renewable energy output prediction sequence.

[0085] The energy storage charge sequence is subjected to boundary truncation, discrete segment interpolation and sampling frequency unification processes to generate an energy storage charge prediction sequence.

[0086] The time-of-use electricity price series is processed by price level extraction, price jump smoothing and period expansion to generate a time-of-use electricity price prediction series.

[0087] The meteorological sequence is subjected to invalid value replacement, relevant factor screening and time delay completion processes to generate a meteorological forecast sequence.

[0088] A multi-source input set is constructed based on the preprocessed load sequence, renewable power output sequence, energy storage charge sequence, time-of-use electricity price sequence and meteorological sequence. A time series forecasting process is performed on the multi-source input set to generate a forecast sequence.

[0089] The forecast series includes hourly, daily, and weekly forecast series. The hourly forecast series consists of load forecasts, renewable energy output forecasts, energy storage load forecasts, time-of-use pricing forecasts, and weather forecasts arranged on an hourly scale. The daily forecast series consists of load forecasts, renewable energy output forecasts, energy storage load forecasts, time-of-use pricing forecasts, and weather forecasts arranged on a daily scale. The weekly forecast series consists of load forecasts, renewable energy output forecasts, energy storage load forecasts, time-of-use pricing forecasts, and weather forecasts arranged on a weekly scale.

[0090] In this embodiment, the step of constructing a scheduling space based on adjustable device parameters specifically refers to:

[0091] The adjustable equipment parameters include power generation equipment parameters, energy storage equipment parameters, and adjustable load parameters;

[0092] The parameters of the power generation equipment include the rated output, minimum output, maximum ramp rate, minimum start-up time and minimum shutdown time of the power generation equipment. Based on the parameters of the power generation equipment, the upper and lower bounds and step intervals of the power generation output variable are set during the dispatch period.

[0093] Energy storage equipment parameters include energy storage capacity, maximum charging power, maximum discharging power, upper and lower limits of state of charge, and charging and discharging efficiency parameters. Based on the energy storage equipment parameters, the value range of energy storage charging variables and energy storage discharging variables and state of charge constraints are set during the scheduling period.

[0094] Adjustable load parameters include adjustable load power range, allowable peak shifting period, continuous adjustment duration limit and recovery power limit. Based on the adjustable load parameters, the value range and adjustment granularity of the load adjustment variables are set within the scheduling period.

[0095] By combining power generation output variables, energy storage charging variables, energy storage discharging variables, and load adjustment variables in each scheduling period to form a set of scheduling variables, a scheduling space covering the scheduling cycle is constructed. The scheduling space is composed of a set of multidimensional variables covering the scheduling cycle. In each scheduling period, corresponding variable dimensions are established according to power generation output variables, energy storage charging variables, energy storage discharging variables, and load adjustment variables. The variable dimensions form discrete value points according to the adjustable equipment parameters. The variable dimensions of all scheduling periods are arranged in chronological order to form a segmented multidimensional grid structure. The combination of each segmented multidimensional grid structure constitutes the overall structure of the scheduling space.

[0096] In this embodiment, S2 specifically refers to:

[0097] Calculate the power difference sequence based on the load forecast sequence and the renewable energy output forecast sequence, and generate the instantaneous gap index for the corresponding time period;

[0098] The cumulative gap index is generated based on the cumulative result of the instantaneous gap index within the scheduling cycle.

[0099] A volatility index is generated based on the magnitude of change in the power difference sequence between adjacent time periods;

[0100] An uncertainty index is generated based on the deviation relationship between hourly, daily, and weekly forecast sequences.

[0101] The instantaneous gap indicator, cumulative gap indicator, volatility indicator, and uncertainty indicator are arranged in chronological order to form a risk indicator sequence. The risk indicator sequence is then organized according to the scheduling period and time scale to generate a risk map. The risk map consists of a multi-time-scale risk matrix covering the scheduling cycle. Corresponding risk matrices are established according to hourly, daily, and weekly scales. Each risk matrix consists of indicator column vectors formed by arranging the instantaneous gap indicator, cumulative gap indicator, volatility indicator, and uncertainty indicator in chronological order. All indicator column vectors are spliced ​​together according to the time dimension to form a multi-column risk structure. The risk structures of each time scale are combined according to the scale order to form the overall structure of the risk map.

[0102] In this embodiment, S3 specifically refers to:

[0103] Based on the power generation output variables, energy storage charging variables, energy storage discharging variables, and load adjustment variables corresponding to each scheduling scheme in the scheduling space within the scheduling cycle, instantaneous gap indicators, cumulative gap indicators, fluctuation indicators, and uncertainty indicators arranged in the same time period of the risk map are selected. The power generation output variables, energy storage charging variables, energy storage discharging variables, load adjustment variables, instantaneous gap indicators, cumulative gap indicators, fluctuation indicators, and uncertainty indicators are combined in the order of time periods to form the risk superposition sequence corresponding to the scheduling scheme.

[0104] The risk superposition sequence is aggregated over time according to the scheduling cycle to form the field strength value corresponding to the scheduling scheme.

[0105] The field strength values ​​of all scheduling schemes are arranged into a field strength distribution structure according to the order of scheduling space.

[0106] In the field strength distribution structure, a differential process is performed according to the differences between scheduling schemes in the dimensions of power generation output, energy storage charging, energy storage discharging and load adjustment variables to form the field strength change structure.

[0107] The field strength distribution structure, field strength change structure and scheduling scheme in the scheduling space are linked in the order of scheduling cycle to form a gap field. The gap field records the field strength information and field strength change information.

[0108] In the scheduling space, a set of scheduling schemes is selected. The power generation output variables, energy storage charging variables, energy storage discharging variables and load adjustment variables of the scheduling schemes in each scheduling period are combined with the instantaneous gap index, cumulative gap index, fluctuation index and uncertainty index of the risk map in the same period. The risk superposition value of the scheduling scheme in each scheduling period is formed by combining them according to the preset weight coefficient.

[0109] The process of aggregating the risk superposition values ​​according to the scheduling cycle is used to form the field strength value corresponding to the scheduling scheme.

[0110] A set of local disturbance scheduling schemes is constructed around the scheduling scheme. The set of local disturbance scheduling schemes is offset according to the disturbance step size in the dimensions of power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable, forming a set of post-disturbance scheduling schemes. Field strength values ​​are generated for the set of post-disturbance scheduling schemes.

[0111] The field gradient direction is generated based on the difference in field strength between the scheduling scheme and the set of scheduling schemes after the disturbance, and the field gradient magnitude is generated based on the absolute value of the difference in field strength.

[0112] The scheduling scheme, field strength value, field gradient direction and field gradient magnitude are associated according to the scheduling scheme index to form a mapping structure between the scheduling scheme and the gap field. The gap field records the field strength value and field gradient parameters in the scheduling space.

[0113] Scheduling scheme = {

[0114] ( : Power generation output variable 1, energy storage charging variable 1, energy storage discharging variable 1, load regulation variable 1).

[0115] ( : Power generation output variable 2, energy storage charging variable 2, energy storage discharging variable 2, load regulation variable 2).

[0116] ...

[0117] ( (Power generation output variable n, energy storage charging variable n, energy storage discharging variable n, load regulation variable n)

[0118] };in to This indicates the scheduling periods arranged in order of scheduling cycle.

[0119] In this embodiment, S4 specifically refers to:

[0120] The scheduling schemes in the scheduling space are segmented according to the field strength values ​​recorded in the mapping structure, and a scheduling scheme is selected in each field strength value interval to form a set of candidate scheduling schemes.

[0121] For each scheduling scheme in the candidate scheduling scheme set, the power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable are read within the scheduling period. The corresponding field strength value, field gradient direction and field gradient amplitude are read from the gap field. The power generation output variable, energy storage charging variable, energy storage discharging variable, load adjustment variable, field strength value, field gradient direction and field gradient amplitude are spliced ​​in time order to form the population individual encoding vector.

[0122] Assign individual identifiers and time index identifiers to the population individual coding vectors, and combine the individual identifiers, time index identifiers, and population individual coding vectors to form population individual data units;

[0123] All individual data units of the population are arranged according to their individual identifiers to form the initial population cluster;

[0124] In the initial population, scheduling variable indices are established for power generation output variables, energy storage charging variables, energy storage discharging variables, and load regulation variables in the order of scheduling cycle. Field parameter indices are established for field strength values, field gradient directions, and field gradient amplitudes. The scheduling variable indices and field parameter indices are associated within the individual data units of the population to form a coupled coding structure between the individual population and the gap field.

[0125] In this embodiment, the step of generating basic update quantities for individuals in the population based on the initial population using an improved HHO algorithm specifically involves:

[0126] In the initial population, the location data of each individual population is read, and the eagle population location vector is constructed in the scheduling space. The power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable are recorded in the order of scheduling cycle to form the location vector.

[0127] The average position vector is calculated based on the position vectors of all individuals in the population within the scheduling space, and the average position vector is used as the population center position data.

[0128] Read the location vector of the scheduling scheme with the minimum field strength value in the field strength distribution structure within the scheduling space, and record the location vector of the scheduling scheme as the target location vector;

[0129] Perform a difference operation on the position vector and the target position vector to form a difference vector. Perform vector multiplication on the difference vector and the random distribution factor to form the basic jump component.

[0130] Within the variable dimension of the scheduling space, a linear combination process is performed on the position vector and the group center position data according to a preset scaling factor to form a collaborative contraction component;

[0131] Perform a vector superposition process on the basic jump component and the cooperative contraction component to form the basic update quantity;

[0132] Within the scheduling cycle, the basic update quantity forms update components for the variable dimensions corresponding to the power generation output variable, energy storage charging variable, energy storage discharging variable, and load adjustment variable. The update components are compared with the upper and lower bounds of the scheduling space. Components exceeding the boundary are truncated to the boundary of the scheduling space to form the basic update quantity after boundary constraints.

[0133] In this embodiment, the gradient energy difference is generated by the field strength value based on the gap field and the field gradient direction; the gradient energy difference is compared with a first preset difference threshold and a second preset difference threshold, and divided into a high difference segment, a medium difference segment, and a low difference segment; in the high difference segment, a jump amplification process is performed on the basic update amount along the negative gradient direction; in the medium difference segment, a direction shift process is performed on the basic update amount; and in the low difference segment, a jump suppression process is performed on the basic update amount, and encirclement update is entered, forming a segmented jump update amount. The position data and historical field strength records of the population individuals are updated based on the segmented jump update amount to form an updated population group, specifically:

[0134] In the initial population, the location data and historical field strength records of each individual population are read, and the corresponding field strength value and field gradient direction are read from the gap field.

[0135] Construct a pre-updated position along the field gradient direction, query the field strength value in the gap field for the pre-updated position, and define the difference between the field strength value of the pre-updated position and the corresponding field strength value of the position data as the gradient energy difference;

[0136] The gradient energy difference is compared with the first preset difference threshold and the second preset difference threshold. If the gradient energy difference is greater than the second preset difference threshold, it is recorded as a high difference segment. If the gradient energy difference is between the first preset difference threshold and the second preset difference threshold, it is recorded as a medium difference segment. If the gradient energy difference is less than the first preset difference threshold, it is recorded as a low difference segment.

[0137] In the opposite direction of the field gradient direction along the elevation difference segment, an amplification factor is superimposed on the basic update amount to generate an amplified jump component. The amplified jump component is then vector-superimposed with the basic update amount to form the segmented transition update amount of the elevation difference segment.

[0138] In the intermediate difference segment, an orthogonal offset direction is constructed based on the field gradient direction, and the offset component is superimposed on the basic update amount to form the segmented transition update amount of the intermediate difference segment.

[0139] In the low-difference segment, a contraction coefficient is superimposed on the basic update amount to generate a jump suppression component. The jump suppression component is then vector-superimposed with the basic update amount, and the target scheduling scheme position data is introduced according to the encirclement update strategy to form the segmented jump update amount in the low-difference segment. The encirclement update strategy is as follows: within the scheduling space, an encirclement direction is constructed based on the difference vector between the position vector of the individual population and the target position vector. The difference vector is then multiplied with the contraction factor to generate a contraction component. The contraction component is then vector-superimposed with the position vector according to the scheduling cycle order to form an encirclement position vector. The encirclement position vector is then truncated within the scheduling space, cutting off position components that exceed the range of values ​​in the scheduling space to the boundary position of the scheduling space. Finally, the encirclement position vector is written into the individual population to form the encirclement update result.

[0140] The segmented transition update amounts for the high-difference segment, the medium-difference segment, and the low-difference segment are written into the population individuals according to the difference segment identifier corresponding to the gradient energy difference, forming the segmented transition update amounts;

[0141] The position data of individuals in the population within the scheduling space are updated based on the segmented transition update amount. The new field strength value is queried in the gap field for the updated position, and the new field strength value is appended to the historical field strength record to form an updated population.

[0142] In this embodiment, the improved HHO algorithm addresses the multidimensional coupling characteristics, non-uniform field strength distribution, and gradient sensitivity of the energy management and scheduling problem by structurally improving upon the standard HHO algorithm. The standard HHO algorithm, when generating jump and contraction components in a complex scheduling space, only constructs its update behavior based on the difference between the individual and target positions, failing to distinguish between gradient changes in the scheduling space, field strength variation characteristics of the risk map, and the sensitivity of different scheduling variables to risk. This easily leads to insufficient jumps in high-gradient regions and excessive jumps in low-gradient regions, resulting in reduced convergence speed and local oscillations in the search process. To address these issues, the improved HHO algorithm introduces gradient energy difference, incorporating the field strength value and gradient direction in the gap field into the update process, and calculating the gradient... The energy difference is used to divide the algorithm into high-difference, medium-difference, and low-difference segments based on the gradient energy difference, providing a basis for segmented adjustment of jump and contraction components. In the high-difference segment, a jump amplification in the negative gradient direction is constructed to enhance the individual's ability to escape quickly in high-risk areas. In the medium-difference segment, a directional offset is constructed to enable the individual to explore smoothly in areas where risk changes are not significant. In the low-difference segment, jumps are suppressed and encirclement updates are added, enabling the individual to maintain a convergence trend in low-risk areas and stably approach the target scheduling scheme. This improves the HHO algorithm to have stronger exploration capabilities in high-risk areas, adaptive adjustment capabilities in medium-risk areas, and fast convergence capabilities in low-risk areas, thereby achieving a higher feasibility search rate, a more stable convergence path, and better scheduling scheme quality in the global scheduling optimization process.

[0143] The improved HHO algorithm performs the update process around the gap field in the scheduling space. Each point in the scheduling space corresponds to a field strength value and a local gradient direction. The field strength value represents the risk level of the scheduling scheme, and the gradient direction represents the risk change trend. By applying a threshold to the gradient energy difference, an update strategy is dynamically assigned to each individual in the population. This allows individuals to exhibit divergent jumping in high-risk areas, offsetting search in medium-risk areas, and contracting approximation in low-risk areas, ultimately forming a convergent path around the low-risk areas. This structure makes the update trajectory exhibit a gradual convergence from the outside in, significantly different from the random jumping path in the standard HHO.

[0144] In this embodiment, S6 specifically refers to:

[0145] In updating the population, based on the power generation output, energy storage charging, energy storage discharging and load adjustment variables of the population individuals within the scheduling cycle, the power generation cost, grid power purchase cost, energy storage charging and discharging cost and wind and solar curtailment penalty cost are calculated. The power generation cost, grid power purchase cost, energy storage charging and discharging cost and wind and solar curtailment penalty cost are summed to form the economic indicators of the population individuals.

[0146] Based on the ratio between the renewable output of an individual population in the scheduling cycle that meets the load and the total electricity consumption, the energy storage cycle efficiency and the load peak shifting depth, energy utilization parameters are calculated, and combined processing is performed on the energy utilization parameters to form the energy efficiency index of the individual population.

[0147] Based on the reserve capacity margin, the number of times the energy storage state of charge exceeds the limit, and the number of times the power exceeds the limit for each individual in the population during the scheduling cycle, the operational risk frequency parameter and the safety margin parameter are calculated. The operational risk frequency parameter and the safety margin parameter are combined and processed to form the safety index of each individual in the population.

[0148] Read the hourly, daily, and weekly field strength values ​​within the scheduling period from the gap field. Calculate the time scale weight based on the magnitude relationship between the hourly, daily, and weekly field strength values. Calculate the time period weight based on the distribution of field strength values ​​in different time periods within the scheduling period. Combine the time scale weight and the time period weight to form the risk weight.

[0149] The economic, energy efficiency, and safety indicators are weighted and integrated according to risk weights to form the fitness values ​​of individual populations.

[0150] The individuals in the updated population are sorted according to their fitness values. Individuals with higher fitness values ​​are selected from the sorted results and retained. The retained individuals are then copied according to a set replication ratio to form duplicate individuals. Individuals with lower fitness values ​​are eliminated to create vacant positions. The duplicate individuals are then filled into the vacant positions to form a new generation of population.

[0151] In this embodiment, S7 specifically refers to:

[0152] In the new generation of population, the fitness value of each individual is read, and the fitness values ​​are sorted to determine the optimal fitness value and the average fitness value of the individual.

[0153] In consecutive iterations, the fitness values ​​of the best individual in the population and the average fitness value are recorded. The difference in fitness values ​​between two adjacent iterations is defined as the fitness change. The fitness change is compared with a preset convergence threshold, the iteration number is recorded, and the iteration number is compared with a preset maximum iteration number.

[0154] If the change in fitness is less than the preset convergence threshold and the number of iterations is greater than or equal to the preset minimum number of iterations, or the number of iterations is equal to the preset maximum number of iterations, then the convergence condition is met. The individual with the highest fitness value is selected from the new generation population as the target population individual. The power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable recorded by the target population individual in the scheduling cycle are combined in time order to form a scheduling scheme.

[0155] The dispatch scheme is applied to the operation of the energy management system. During the dispatch cycle, actual load data, actual renewable power output data, actual energy storage charge data, actual time-of-use electricity price data, and actual meteorological data are collected. The generation cost, grid purchase cost, energy storage charging and discharging cost, wind and solar curtailment, reserve capacity margin, number of times energy storage charge status exceeds the limit, and number of times power exceeds the limit are recorded during the dispatch cycle to form verification data.

[0156] The multi-source time series data is updated based on the actual load data, actual renewable energy output data, actual energy storage charge data, actual time-of-use electricity price data and actual meteorological data in the verification data. A preprocessing process is performed based on the updated multi-source time series data to generate a corrected prediction sequence. Based on the corrected prediction sequence, the power difference sequence, instantaneous gap index, cumulative gap index, volatility index and uncertainty index are calculated, and the risk index sequence and risk map are updated.

[0157] The generation cost, grid purchase cost, energy storage charging and discharging cost, wind and solar curtailment, reserve capacity margin, number of times energy storage charge status exceeded limits and power exceeded limits from the verification data are written into the historical operation data set to form an optimization closed loop;

[0158] The convergence criteria include the fitness change criterion and the iteration round criterion. The fitness change criterion means that the difference in fitness values ​​between consecutive iteration rounds is less than a preset convergence threshold and the iteration round is greater than or equal to a preset minimum iteration round. The iteration round criterion means that the iteration round reaches a preset maximum iteration round. If either the fitness change criterion or the iteration round criterion is met, it is considered that the convergence criteria are met.

[0159] If the convergence condition is not met, the new generation population will be used as the updated population, and the update process based on the improved HHO algorithm, along with the selection, replication, and elimination operations based on fitness values, will be re-executed in conjunction with the gap field.

[0160] Example 1:

[0161] To verify the feasibility of this invention in practice, it was applied to a comprehensive energy management center in a coastal city as an example. This center manages a comprehensive energy park integrating a photovoltaic power station, wind farm, natural gas turbine units, lithium battery energy storage station, and several adjustable industrial loads. The installed capacity is approximately 30 MW of photovoltaic power, 20 MW of wind power, and 25 MW of gas turbine units (three units total), with an energy storage system capacity of 10 MWh and an adjustable load capacity of approximately 8 MW. The park's original dispatching model used day-ahead load forecasting with a fixed safety margin, combined with empirical rules to adjust power generation plans and energy storage strategies. This approach struggled to balance economic efficiency, energy efficiency, and operational safety under conditions of drastic wind and solar power fluctuations and multi-timescale changes in electricity prices. Especially during the low-irradiance, high-load periods of winter and early spring, wind and solar power curtailment was frequent, and problems such as insufficient reserve capacity and exceeding the state of charge limit of energy storage occurred regularly.

[0162] During the scheduling and operation phase, the method of this invention was compared with the park's original traditional rule-based scheduling method over a three-month testing period. During the testing period, the traditional rule-based scheduling method still generated scheduling schemes according to the original load forecasting and empirical safety margin strategy; the method of this invention, within a daily rolling window, generates basic update quantities for individual populations based on the initial population cluster and the gap field using an improved HHO algorithm. Within the scheduling space, it simultaneously references the population center position and the minimum field strength position to construct jump components and cooperative contraction components, forming the basic update quantities after boundary constraints. Combining the field strength values ​​and field gradient directions in the gap field, it calculates the gradient energy difference, dividing the search process into high-difference, medium-difference, and low-difference segments. Jump amplification processing is performed in the high-difference segment, direction offset processing in the medium-difference segment, and jump suppression and encirclement update processing in the low-difference segment, forming segmented jump update quantities. This iteratively updates the position data and historical field strength records of individual populations, resulting in an updated population cluster. Subsequently, based on economic, energy efficiency, and safety indicators, and combined with risk weights generated from field strength values ​​at different time scales, the fitness values ​​of the updated population are calculated, and selection, replication, and elimination operations are performed to form a new generation of populations until convergence conditions are met. The scheduling scheme is then output and written into the operation plan of the energy management system.

[0163] Throughout the three-month testing period, operational data including generation cost, grid purchase cost, energy storage charging and discharging cost, wind and solar power curtailment, renewable energy utilization rate, load shifting depth, reserve capacity margin, number of times energy storage charge status exceeded limits, and number of times power exceeded limits were recorded to form verification data. At the same time, the verification data was written into the historical operational data set, and multi-source time series data, prediction sequences and risk maps were continuously updated to form an optimization closed loop that can be rolled out over a long period of time.

[0164] To more intuitively demonstrate the advantages of the method of this invention in terms of economy and energy efficiency, the operating data over three months were summarized to construct a comparison table of economic and energy efficiency indicators.

[0165] Table 1. Comparison of economic and energy efficiency indicators based on swarm optimization algorithm.

[0166]

[0167] As can be observed from the data in Table 1, the method of this invention exhibits significant advantages in both economic efficiency and energy efficiency during the three-month testing period. Firstly, regarding power generation costs, the costs of this invention in the first to third months were RMB 2.952 million, RMB 2.834 million, and RMB 2.891 million, respectively, showing a stable decrease compared to the traditional rule-based dispatching method's RMB 3.125 million, RMB 2.987 million, and RMB 3.054 million, with an average reduction of approximately 5%–7%. The cost of purchasing electricity from the grid also shows a significant downward trend. The method of this invention reduces the cost of purchased electricity by 10%–12%, directly reflecting a more optimized dispatching scheme in terms of renewable energy utilization and energy storage compensation. Although the cost of energy storage charging and discharging increases slightly, the increase is within a controllable range and does not cause a deterioration in economic efficiency.

[0168] The reduction in wind and solar power curtailment is particularly significant. For example, in the first month, it decreased from 486 MWh to 321 MWh, a reduction of over 30%, directly leading to an increase in renewable energy utilization rate from 81.4% to 89.2%. This verifies that the method of this invention effectively reduces renewable energy waste by prioritizing low-field-strength areas in the dispatch space search. The load shifting depth increased to approximately 31%–34%, indicating that energy storage and adjustable loads have better coordinated regulation performance under the guidance of the risk map. The comprehensive energy efficiency index improved from 0.71–0.74 in traditional dispatch to 0.83–0.86, showing a significant overall improvement, demonstrating that this invention has a consistent optimization capability in energy efficiency improvement.

[0169] Table 2 Comparison of Safety Operation and Risk Indicators Based on Gap Field Guidance

[0170]

[0171] Table 2 shows a comparison of different scheduling methods in terms of operational safety and risk indicators. The method of this invention outperforms traditional rule-based scheduling methods in terms of reserve capacity margin, with an average improvement of over 3 percentage points over three months, for example, reaching 15.1% in the third month, further enhancing the system's ability to cope with sudden power fluctuations. The number of energy storage state-of-charge violations and power violations are significantly reduced; for example, the number of violations in the first month decreased from 14 and 11 to 6 and 4 respectively, a reduction of over 50%. This indicates that the method of this invention effectively constrains energy storage scheduling behavior by controlling the field strength and gradient direction in the gap field, making energy storage operation more stable and reliable.

[0172] The mean values ​​of both the instantaneous gap index and the volatility index decreased by approximately 40% under the method of this invention. For example, in the first month, the instantaneous gap index decreased from 6.3MW to 3.8MW, and the volatility index decreased from 4.7MW to 2.9MW. This indicates that the scheduling scheme based on the field strength distribution structure can better mitigate the deviation between load and output, and improve the stability of system operation. The uncertainty index also decreased from approximately 0.64 to 0.47, reflecting the enhanced role of multi-timescale prediction sequences in the risk profile, making the scheduling scheme more robust to prediction biases.

[0173] The overall operational safety index significantly improved from 0.72–0.74 to 0.86–0.88, demonstrating a stable improvement in overall performance. This proves that by constructing a scheduling space, risk map, and gap field, and combining the segmented transition update amount generated by the improved HHO algorithm, the present invention can significantly improve system security and risk resistance while ensuring economy.

[0174] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A population algorithm-based intelligent optimization method for energy management, characterized in that, Includes the following steps: S1. Collect multi-source time series data and adjustable equipment parameters from the energy management system; perform preprocessing on the multi-source time series data to generate a prediction sequence; construct a scheduling space based on the adjustable equipment parameters; S2. Calculate the risk indicator sequence based on the predicted sequence and form a risk map in chronological order; S3. Construct a gap field based on the scheduling space and risk map. By mapping the scheduling schemes in the scheduling space to the risk map to form field strength values, extract the field gradient direction and field gradient magnitude, and construct the mapping structure between the scheduling schemes and the gap field. Specifically, S3 is: A gap field is constructed based on the scheduling space and risk map. According to the set of scheduling variables corresponding to each scheduling scheme in the scheduling space within the scheduling cycle, the power generation output variable, energy storage charging variable, energy storage discharging variable, load adjustment variable, instantaneous gap index, cumulative gap index, fluctuation index, and uncertainty index are combined and processed in the order of the scheduling cycle to form a risk superposition sequence corresponding to the scheduling scheme. The risk superposition sequence is aggregated according to the scheduling cycle to form the field strength value corresponding to the scheduling scheme. The field strength values ​​of all scheduling schemes are arranged in the order of scheduling space to form a field strength distribution structure. Differential processing is performed on the differences between the power generation output variable dimension, energy storage charging variable dimension, energy storage discharging variable dimension and load adjustment variable dimension in the field strength distribution structure to form a field strength change structure. The field strength distribution structure, field strength change structure and scheduling scheme are linked in the order of scheduling cycle to form a gap field. The gap field records the field strength information and field strength change information. Based on the set of scheduling variables corresponding to the scheduling scheme within the scheduling period, the field strength value change corresponding to the scheduling scheme is extracted from the gap field, and the field gradient direction and field gradient magnitude corresponding to the scheduling scheme are determined based on the field strength value change. The scheduling scheme, field strength value, field gradient direction and field gradient magnitude are associated according to the scheduling scheme index to form a mapping structure between the scheduling scheme and the gap field; S4. Based on the mapping structure, an initial population is generated in the scheduling space, and a coupled coding structure between the population individuals and the gap field is constructed to form the initial population. S5. Based on the initial population, the improved HHO algorithm is used to generate basic update values ​​for individual populations. The improved HHO algorithm generates a gradient energy difference based on the field strength value and field gradient direction of the gap field. It introduces a piecewise transition mechanism to process the basic update amount through the gradient energy difference, forming an update species cluster. The method of generating basic update quantities for individuals in the population based on the initial population using an improved HHO algorithm is as follows: In the initial population, the location data of each individual population is read, and the eagle population location vector is constructed in the scheduling space. The power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable are recorded in the order of scheduling cycle to form the location vector. The average position vector is calculated based on the position vectors of all individuals in the population within the scheduling space, and the average position vector is used as the population center position data. Read the location vector of the scheduling scheme with the minimum field strength value in the field strength distribution structure within the scheduling space, and record the location vector of the scheduling scheme as the target location vector; Perform a difference operation on the position vector and the target position vector to form a difference vector. Perform vector multiplication on the difference vector and the random distribution factor to form the basic jump component. Within the variable dimension of the scheduling space, a linear combination process is performed on the position vector and the group center position data according to a preset scaling factor to form a collaborative contraction component; Perform a vector superposition process on the basic jump component and the cooperative contraction component to form the basic update quantity; The basic update quantity forms update components for the variable dimensions corresponding to the power generation output variable, energy storage charging variable, energy storage discharging variable and load adjustment variable within the scheduling cycle. The update components are compared with the upper and lower bounds of the scheduling space. Components that exceed the boundary are truncated to the boundary of the scheduling space to form the basic update quantity after boundary constraints. The field strength value based on the gap field and the field gradient direction generate a gradient energy difference. The improved HHO algorithm introduces a piecewise transition mechanism to process the basic update amount through the gradient energy difference, forming an update seed cluster, specifically: In the initial population, the location data and historical field strength records of each individual population are read, and the corresponding field strength value and field gradient direction are read from the gap field. Construct a pre-updated position along the field gradient direction, query the field strength value in the gap field for the pre-updated position, and define the difference between the field strength value of the pre-updated position and the corresponding field strength value of the position data as the gradient energy difference; The gradient energy difference is compared with the first preset difference threshold and the second preset difference threshold. If the gradient energy difference is greater than the second preset difference threshold, it is recorded as a high difference segment. If the gradient energy difference is between the first preset difference threshold and the second preset difference threshold, it is recorded as a medium difference segment. If the gradient energy difference is less than the first preset difference threshold, it is recorded as a low difference segment. In the opposite direction of the field gradient direction along the elevation difference segment, an amplification factor is superimposed on the basic update amount to generate an amplified jump component. The amplified jump component is then vector-superimposed with the basic update amount to form the segmented transition update amount of the elevation difference segment. In the intermediate difference segment, an orthogonal offset direction is constructed based on the field gradient direction, and the offset component is superimposed on the basic update amount to form the segmented transition update amount of the intermediate difference segment. In the low-difference segment, a shrinkage coefficient is superimposed on the basic update amount to generate a jump suppression component. The jump suppression component is then vector-superimposed on the basic update amount, and the target scheduling scheme position data is introduced according to the encirclement update strategy to form the low-difference segment segmented jump update amount. The segmented transition update amounts for the high-difference segment, the medium-difference segment, and the low-difference segment are written into the population individuals according to the difference segment identifier corresponding to the gradient energy difference, forming the segmented transition update amounts; The position data of individuals in the population within the scheduling space are updated according to the segmented transition update amount. The new field strength value is queried in the gap field for the updated position, and the new field strength value is appended to the historical field strength record to form an updated population. S6. Based on the field strength values ​​corresponding to the updated population and gap field at different time scales, perform selection, replication and elimination operations to form a new generation of population; S7. Based on the new generation of population and gap field, perform convergence judgment. If the convergence condition is met, output the scheduling scheme, execute and verify the scheduling scheme and collect verification data, update the risk map and prediction sequence, and complete the optimization closed loop.

2. The population algorithm-based intelligent optimization method for energy management according to claim 1, wherein, The preprocessing of multi-source time series data to generate a predicted sequence specifically involves: The multi-source time series data includes load series, renewable power output series, energy storage charge series, time-of-use electricity price series, and meteorological series; Preprocessing is performed on the multi-source time prediction series to generate load prediction series, renewable energy output prediction series, energy storage load prediction series, time-of-use electricity price prediction series, and meteorological prediction series; prediction series are generated through the time series prediction process; the prediction series include hourly prediction series, daily prediction series, and weekly prediction series.

3. The method of claim 1, wherein the method further comprises: The scheduling space based on adjustable device parameters is specifically as follows: The adjustable equipment parameters include power generation equipment parameters, energy storage equipment parameters, and adjustable load parameters; The parameters of the power generation equipment include the rated output, minimum output, maximum ramp rate, minimum start-up time and minimum shutdown time of the power generation equipment. Based on the parameters of the power generation equipment, the upper and lower bounds and step intervals of the power generation output variable are set during the dispatch period. Energy storage equipment parameters include energy storage capacity, maximum charging power, maximum discharging power, upper and lower limits of state of charge, and charging and discharging efficiency parameters. Based on the energy storage equipment parameters, the value range of energy storage charging variables and energy storage discharging variables and state of charge constraints are set during the scheduling period. Adjustable load parameters include adjustable load power range, allowable peak shifting period, continuous adjustment duration limit and recovery power limit. Based on the adjustable load parameters, the value range and adjustment granularity of the load adjustment variables are set within the scheduling period. By combining power generation output variables, energy storage charging variables, energy storage discharging variables, and load adjustment variables within the scheduling period, a set of scheduling variables is formed, constituting a scheduling space covering the scheduling cycle.

4. The intelligent optimization method for energy management based on swarm optimization algorithm according to claim 1, characterized in that, Specifically, S2 is: The power difference sequence is calculated based on the load forecast sequence and the renewable energy output forecast sequence, and the instantaneous gap index for the corresponding time period is generated; the cumulative gap index is generated based on the cumulative result of the instantaneous gap index within the scheduling cycle; the fluctuation index is generated based on the change amplitude of the power difference sequence between adjacent time periods; and the uncertainty index is generated based on the deviation relationship between the forecast sequences. The instantaneous gap indicator, cumulative gap indicator, volatility indicator, and uncertainty indicator are arranged in chronological order to form a risk indicator sequence. The risk indicator sequence is then organized according to the scheduling period and time scale to generate a risk map.

5. The intelligent optimization method for energy management based on swarm optimization as described in claim 1, characterized in that, Specifically, S4 is: The scheduling schemes in the scheduling space are segmented according to the field strength values ​​recorded in the mapping structure, and a scheduling scheme is selected in each field strength value interval to form a set of candidate scheduling schemes. For each scheduling scheme in the candidate scheduling scheme set, the scheduling variable set is read within the scheduling period, and the corresponding field strength value, field gradient direction and field gradient amplitude are read from the gap field. The power generation output variable, energy storage charging variable, energy storage discharging variable, load adjustment variable, field strength value, field gradient direction and field gradient amplitude are spliced ​​in time order to form the population individual encoding vector. Assign individual identifiers and time index identifiers to the population individual coding vectors, and combine the individual identifiers, time index identifiers, and population individual coding vectors to form population individual data units; All individual data units of the population are arranged according to their individual identifiers to form the initial population cluster; In the initial population, scheduling variable indexes are established for the set of scheduling variables according to the scheduling cycle order. Field parameter indexes are established for the field strength value, field gradient direction, and field gradient magnitude. The scheduling variable indexes and field parameter indexes are associated within the individual data units of the population to form a coupled coding structure between the individual population and the gap field.

6. The intelligent optimization method for energy management based on swarm optimization algorithm according to claim 1, characterized in that, Specifically, S6 is: Economic, energy efficiency, and safety indicators are calculated based on the set of scheduling variables of individuals in the updated population during the scheduling cycle. Read the field strength values ​​corresponding to different time scales from the gap field, and generate risk weights based on the field strength values; The economic, energy efficiency, and safety indicators are weighted and integrated according to risk weights to form a fitness value. The updated population is sorted, replicated, and eliminated according to its fitness value to form a new generation of population.

7. The intelligent optimization method for energy management based on swarm optimization algorithm according to claim 1, characterized in that, Specifically, S7 is: The fitness values ​​of individuals in the new generation population are read, sorted, and fitness change and iteration round data are generated. The fitness change is compared with a preset convergence threshold, and the iteration rounds are compared with a preset minimum iteration round and a preset maximum iteration round. The convergence condition is determined based on the comparison results. When the convergence condition is met, the individual with the highest fitness value is selected from the new generation population. A scheduling scheme is formed based on the set of scheduling variables recorded by the individual population. Multi-source time series data is collected during the scheduling period to record economic data and operational data, forming validation data. The multi-source time series data is updated based on the validation data, a revised prediction series is generated, the risk indicator series and risk map are updated, and the validation data is written into the historical operation data set to form an optimization closed loop. If the convergence condition is not met, the new generation population is used as the updated population, and the update process based on the improved HHO algorithm, as well as the selection, replication and elimination operations based on fitness values, are continued in combination with the gap field.

Images