A method for energy system planning based on multi-objective genetic algorithm

Abstract

The application provides an energy system planning method based on a multi-objective genetic algorithm, obtains decision variables and combines the multi-objective genetic algorithm to perform genetic iteration; a non-dominated sorting method based on sequential search is used to obtain multiple non-dominated levels sorted according to non-dominated priority, and under the condition that genetic iteration meets a preset iteration stop condition, an optimal solution set corresponding to a current population is output; a planning scheme individual in the optimal solution set is subjected to multi-index evaluation, and a planning scheme is output. The application realizes deep coupling of planning scheme design and actual operation conditions, effectively improves the engineering landing adaptability of the planning scheme; further, by combining the non-dominated sorting method based on sequential search, the iteration calculation efficiency of multi-objective optimization is guaranteed while the iteration calculation precision is improved; the optimal solution set is subjected to multi-index evaluation and the final planning scheme is output based on the evaluation result, objective quantitative screening of the planning scheme under multi-objective conflict is realized, and multi-element optimization objectives are effectively balanced.

Description

Technical Field

[0001] This application relates to the field of energy planning technology, specifically to an energy system planning method based on a multi-objective genetic algorithm. Background Technology

[0002] Integrated energy systems have become an important development direction for achieving low carbon emissions and improving energy efficiency in the energy sector. As a prerequisite for project design and engineering implementation, the rationality of energy system planning directly determines the operational effectiveness of the system throughout its entire life cycle. In complex system scenarios involving multiple energy couplings, energy system planning must simultaneously consider multiple mutually constraining optimization objectives.

[0003] In existing energy system planning methods, there is a significant disconnect between the planning-level scheme design and actual operating conditions. The final output planning scheme fails to achieve the preset optimization effect in actual engineering operation, and the scheme's engineering implementation adaptability is insufficient, making it difficult to meet the diverse constraints and needs of actual operating scenarios.

[0004] Furthermore, in the process of multi-objective optimization, especially in the application scenario of large-scale population and multiple optimization objectives, the existing technology has low computational efficiency in individual selection during the iterative optimization process, limited convergence speed of the iteration, and difficulty in ensuring the convergence and distribution balance of the output candidate solution set, thus failing to provide a comprehensive and reliable basis for the selection of the final solution.

[0005] Existing planning methods, after obtaining a set of alternative solutions for multi-objective optimization, often rely on a single indicator or experience-based judgment for the final selection of the solution. This makes it difficult to conduct a comprehensive and objective quantitative evaluation of the solutions under multiple objectives, effectively balance multiple conflicting optimization objectives, and adapt to the diverse decision-making needs of actual engineering projects.

[0006] In summary, existing energy system planning methods suffer from technical problems such as a disconnect between planning and design and actual operating conditions, insufficient optimization iteration efficiency and solution quality, and a lack of rigorous evaluation methods for solution selection. Summary of the Invention

[0007] This application addresses the problems existing in the prior art by providing an energy system planning scheme that balances engineering adaptability, optimized efficiency, and multi-objective comprehensive benefits.

[0008] To achieve the above objectives, the technical solution adopted in this application is as follows: This application provides an energy system planning method based on a multi-objective genetic algorithm, which includes: Determine the decision variables for energy system planning, and generate an initial parent population based on the decision variables; Genetic iteration is performed based on the initial parent population and the multi-objective genetic algorithm; During the genetic iteration process, for each planning scheme individual in the current parent and offspring populations, a simulation is performed hourly throughout the year to calculate the multinomial objective function value corresponding to each planning scheme individual. The objective function value is then used as the fitness of the corresponding planning scheme individual for the genetic iteration. During the genetic iteration process, the current parent population and the offspring population are merged to obtain a merged population. A non-dominated sorting method based on sequential search is used to divide all planning scheme individuals in the merged population into non-dominated levels, resulting in multiple non-dominated levels sorted by non-dominated priority. According to the order of non-dominated priority from high to low, the planning scheme individuals in each of the non-dominated levels are added to the next generation parent population in sequence. If the genetic iteration satisfies the preset iteration stopping condition, output the Pareto optimal solution set corresponding to the current population; The individual planning schemes in the Pareto optimal solution set are evaluated using multiple indicators, and the planning scheme is output based on the results of the multi-indicator evaluation.

[0009] Optionally, the decision variables include the installed capacity of various energy devices in the energy system; In the initial parent population, each individual planning scheme includes energy equipment installed capacity parameters corresponding to the decision variables.

[0010] Optionally, during the process of sequentially adding the planning scheme individuals from each of the non-dominant levels to the next generation parent population, the cumulative number of individuals in the next generation parent population is counted in real time. When the cumulative number of individuals exceeds a preset population size threshold after all planning scheme individuals in the target non-dominated level are added, the target non-dominated level is determined as the critical frontier. The crowding distance is calculated for all planning scheme individuals within the critical frontier, and a corresponding number of planning scheme individuals are selected from the critical frontier based on the crowding distance to be added to the next generation parent population. The number of individuals in the next generation parent population is equal to the preset population size threshold.

[0011] Optionally, after each round of genetic iteration completes the construction of the next generation of parent population, it is determined whether the genetic iteration meets the preset iteration stopping condition; If the preset iteration stopping condition is not met, the next generation parent population is used as the current parent population for a new round of genetic iteration, and the genetic iteration operation is performed. If the preset iteration stopping condition is met, the population corresponding to the current iteration round is taken as the optimized population, and the Pareto optimal solution set corresponding to the optimized population is output.

[0012] Optionally, the non-dominated sorting method based on sequential search includes: For all planning scheme individuals in the merged population, pre-sorting is performed according to a preset objective function sorting strategy to obtain a pre-sorted sequence of individuals arranged in a fixed order; According to the order of the pre-sorted individual sequence, the dominance relationship of each planning scheme individual is judged in turn, and each planning scheme individual is assigned to the lowest non-dominance level that meets the admission conditions.

[0013] Optionally, the preset objective function sorting strategy includes: According to the pre-set priority of the objective function, the objective function values ​​corresponding to the individual planning schemes are arranged in ascending or descending order in turn. If the objective function values ​​of the previous priority are the same, sort them according to the objective function values ​​of the next priority, until the pre-sorting process of all individual planning schemes is completed.

[0014] Optionally, before performing the hourly simulation throughout the year, the basic input data for the simulation can be obtained in advance; The simulation input data includes hourly meteorological data for the target planning area throughout the year, hourly multi-category energy load demand data for the whole year, and technical and cost parameters of various energy equipment.

[0015] Optionally, the values ​​of the multiple objective functions include the annual total cost, net carbon emissions, and efficiency of the corresponding planning scheme. The multiple objective functions include minimizing annual total cost and net carbon emissions, and maximizing efficiency.

[0016] Optionally, the initial parent population is generated using the Tent chaotic mapping method, including: Generate a chaotic sequence with traversal characteristics within a preset numerical range; The chaotic sequence is transformed into the preset value constraint range corresponding to the decision variable through linear mapping to obtain multiple sets of individual planning schemes that meet the constraint conditions; The initial parent population is generated based on the multiple groups of individuals in the planning scheme.

[0017] Optionally, the multi-index evaluation adopts the TOPSIS evaluation method; In the multi-index evaluation, the objective function value corresponding to each planning scheme in the Pareto optimal solution set is used as the evaluation index, and a standardized evaluation matrix is ​​constructed through positive transformation and standardization. Based on the standardized evaluation matrix, the positive ideal solution and the negative ideal solution are determined. The relative proximity of each planning scheme individual to the positive ideal solution and the negative ideal solution is calculated, and all planning scheme individuals are ranked according to the relative proximity. The best-ranked planning scheme among all individual planning schemes is output as the planning scheme.

[0018] Compared with the prior art, this application has the following advantages: This application achieves deep coupling between planning scheme design and actual operating conditions by performing hourly simulations of all planning scheme individuals throughout the year in each round of genetic iteration of a multi-objective genetic algorithm and using the calculated multi-objective function values ​​as individual fitness. This effectively improves the engineering adaptability of the planning scheme. Furthermore, by combining hourly simulations throughout the year with a non-dominated sorting method based on sequential search, the application completes the division of non-dominated levels of individuals and the selection of the next generation population. This ensures the efficiency of multi-objective optimization iteration calculation while improving the accuracy of iteration calculation, and also ensures the convergence and distribution balance of the output Pareto optimal solution set. Furthermore, by combining multi-index evaluation of planning scheme individuals within the Pareto optimal solution set and outputting the final planning scheme based on the evaluation results, this application achieves objective quantitative screening of planning schemes under multi-objective conflicts, effectively balancing multiple optimization objectives and outputting energy system planning schemes that take into account both comprehensive benefits and engineering practicality. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method in a specific embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] like Figure 1 As shown, this application provides an energy system planning method based on a multi-objective genetic algorithm, including: S1. Determine the decision variables for energy system planning and generate an initial parent population based on the decision variables; The decision variables include the installed capacity of various energy devices in the energy system; in generating the initial parent population, each planning scheme individual contains the installed capacity parameter of the energy devices corresponding to the decision variables. In this embodiment, the Tent chaotic mapping method is used to generate the initial parent population. This method includes: generating a chaotic sequence with ergodic characteristics within a preset numerical interval; transforming the chaotic sequence to the preset value constraint range corresponding to the decision variables through linear mapping to obtain multiple sets of planning scheme individuals that meet the constraint conditions; and generating the initial parent population based on the multiple sets of planning scheme individuals.

[0024] In the planning and optimization of integrated energy systems, a high-quality initial population can significantly improve global search efficiency and prevent the algorithm from prematurely falling into suboptimal solution regions. Therefore, this embodiment does not use traditional completely random initialization, but instead uses the Tent chaotic mapping with ergodic properties to generate initial solutions, ensuring that the first-generation solutions are evenly distributed in the decision space, providing a better foundation for subsequent evolutionary exploration.

[0025] S1.1 Determine decision variables and prepare parameters; To generate an initial planning scheme for a comprehensive energy system that satisfies system constraints and has good distribution characteristics, the set of decision variables for the system is first determined. Technical parameters of type M energy equipment are obtained from equipment manufacturers' technical manuals or actual engineering cases. Based on these parameters, the installed capacity decision variables, i.e., the planning scheme, are expressed as follows: , where each variable represents the installed capacity of the corresponding i-th type of equipment, in kW, and its value range is constrained by the lower and upper limits of the equipment capacity.

[0026] S1.2 generates a chaotic sequence; Randomly generate an initial value in the interval (0,1). And iterative calculations are performed based on the Tent chaotic mapping relationship: ; In the formula, This represents the chaotic value corresponding to the i-th decision variable in the n-th iteration.

[0027] Initial value By iteratively substituting the above formula, an initial population of N individuals is constructed, where each individual (i.e., the planning scheme) consists of the capacity of M types of equipment. Therefore, the total number of chaos values ​​required is... .

[0028] S1.3 performs space mapping; After obtaining the chaotic variables, the generated chaotic sequence is merely an abstract number between 0 and 1. It must be converted into a device installed capacity with actual physical meaning through linear mapping, i.e., spatial mapping. The core of this is linear scaling, and its formula is: ; In the formula, and These represent the lower and upper limits of the installed capacity of the i-th type of equipment, respectively.

[0029] Using the above formula, the decision vector Each device variable in the process executes this process independently, thereby generating a complete planning scheme. This process is repeated N times, resulting in a population of N random but uniformly distributed initial planning schemes for the integrated energy system that satisfy the capacity constraints. .

[0030] S2. Genetic iteration is performed based on the initial parent population and a multi-objective genetic algorithm; During the genetic iteration process, for each planning scheme individual in the current parent and offspring populations, a simulation is performed hourly throughout the year to calculate the multinomial objective function value corresponding to each planning scheme individual. The objective function value is then used as the fitness of the corresponding planning scheme individual for genetic iteration.

[0031] Before conducting hourly simulations throughout the year, basic input data for the simulation is acquired in advance. This basic input data includes hourly meteorological data for the target planning area throughout the year, hourly multi-category energy load demand data throughout the year, and technical and cost parameters of various energy equipment. Multiple objective function values ​​include the annual total cost, net carbon emissions, and efficiency for each individual planning scheme. The multiple objective functions include minimizing the annual total cost and net carbon emissions, and maximizing the efficiency.

[0032] In this embodiment, for each planning scheme, a comprehensive energy system simulation model is constructed based on hourly operation over 8760 hours throughout the year. The specific implementation process is as follows: After generating the initial capacity configuration schemes, the advantages and disadvantages of each scheme must be evaluated. Since the operation of energy systems is dynamic, simple static calculations cannot reflect their true performance. Therefore, this method constructs a comprehensive energy system simulation model based on hourly operation over 8760 hours throughout the year for each planning scheme generated in step one, and calculates its objective function values ​​in three dimensions: economy, environmental protection, and technological advancement.

[0033] S2.1 Timing Simulation Data Preparation and Execution Strategy; To ensure the accuracy and feasibility of the integrated energy system operation simulation results, it is necessary to prepare and organize the input data required for system operation before conducting hourly operation simulations throughout the year. The input data mainly includes the following three categories.

[0034] The first category consists of hourly meteorological data throughout the year, used to characterize the impact of external natural conditions on the output characteristics of renewable energy. Meteorological data includes at least parameters such as solar irradiance, ambient air temperature, wind speed, and relative humidity. Solar irradiance and ambient temperature are used to calculate the output power of photovoltaic and solar thermal power generation units, while wind speed is used to calculate the output power of wind power generation units. This meteorological data can be based on historical meteorological records of the target area or obtained from meteorological databases.

[0035] The second category is hourly multi-energy load demand data throughout the year, used to characterize the energy demand characteristics of the integrated energy system at different time scales. Load data includes electricity load, heat load, and cooling load demand curves. Load curves can be obtained based on historical operating data statistics or calculated through load forecasting models, and are used to constrain the energy supply and demand balance of the system in each time period.

[0036] The third category is system equipment and market-related data, including the technical and economic parameters of existing and planned energy equipment in the integrated energy system, as well as external energy market price information. Technical parameters include at least the energy conversion efficiency, rated capacity, and operating constraints of various types of equipment; economic parameters include at least equipment construction costs and operation and maintenance costs; market data includes time-of-use electricity prices, natural gas prices, and carbon trading prices, used to support the calculation of system economics and carbon emission-related indicators.

[0037] During hourly operation simulations, the integrated energy system is dispatched and controlled according to a pre-set collaborative operation strategy. This strategy adopts a "heat-driven" operation principle, prioritizing the consumption of fluctuating renewable energy sources such as wind and solar power during each operating period. When renewable energy output is insufficient to meet electricity load demand, the system sequentially calls upon the energy storage system to discharge, the concentrated solar power (CSP) unit to output power, or starts the gas turbine to supplement power supply, according to a predetermined priority. While meeting electricity load demand, the system fully utilizes the waste heat resources of the cogeneration unit, the heat storage capacity of the CSP unit, or independent heat storage devices to provide the required heat energy for the thermal load. The specific collaborative operation simulation strategy follows a pre-set decision logic, which, based on whether there is a power supply gap in the system, sequentially calls upon equipment such as energy storage batteries, CSP units, CHP units, and electric boilers to achieve real-time balance of the system's electrical and thermal loads.

[0038] By implementing the above data preparation and operation scheduling strategies, the start-up and shutdown status and output level of various equipment in the integrated energy system during the 8760 operating periods throughout the year can be obtained, thus providing a reliable operational data foundation for subsequent energy flow calculation, objective function evaluation and multi-objective optimization.

[0039] S2.2 Calculate the objective function value; Based on the detailed simulation records above, three core objective function values ​​can be calculated to quantify the performance of each scheme.

[0040] Total cost The economic viability of this measure can be calculated using the annual cost method, which includes construction costs, operation and maintenance costs, and tiered carbon trading costs. The expression is as follows: ; In the formula, Annualized construction cost of the equipment; For system operation and maintenance costs; This refers to the tiered carbon trading costs.

[0041] The formula for calculating the annualized construction cost of equipment is as follows: ; ; In the formula, Let be the cost recovery factor for the i-th type of equipment; The construction cost per unit installed capacity of equipment of type i; For the installed capacity of the i-th type of equipment; The discount rate; Let be the service life of the i-th type of equipment.

[0042] The formula for calculating system operation and maintenance costs is as follows: ; In the formula, The annual fixed operation and maintenance cost per unit installed capacity of equipment of type i; Let be the power consumption of device i during time period t; Let be the amount of natural gas consumed by the i-th type of equipment during time period t; Electricity price for period t; Let be the price of natural gas during time period t.

[0043] The formula for calculating the cost of tiered carbon trading is as follows: ; In the formula, This represents the tiered carbon trading cost generated by the system in time period t.

[0044] Net carbon emissions This measure assesses the environmental friendliness of the system by calculating all carbon dioxide emissions generated during its operation, including uncaptureable CO2 emissions from processes such as gas turbines burning natural gas, and indirect carbon emissions from purchasing external energy sources. The formula is expressed as: ; In the formula, , These represent the carbon emissions generated by the energy system's energy purchases in the planning year, and the carbon emissions that are not captured, reused, or stored by carbon capture devices. The expressions for these two categories are as follows: ; ; In the formula, , These represent the amount of electricity and natural gas purchased by the energy system during time period t, respectively. , These are the carbon emission coefficients for purchasing electricity and purchasing natural gas, respectively. This represents the amount of CO2 captured by the unit during time period t. CO2 capture rate (%) of CCS unit.

[0045] Overall efficiency This measure assesses the technological advancement of the scheme, specifically the degree of energy "quality" utilization. First, based on thermodynamic formulas, the total input of fuel, electricity, and heat to the system is summed with the total output of electricity, heat, and cooling loads. Then, the ratio is calculated, expressed as: ; In the formula, , These represent the electrical energy output and input of the energy system during time period t, respectively. , These represent the heat energy output and input of the energy system during time period t, respectively; the energy quality coefficient of electrical energy. Set it to 1; Let be the mass-energy equivalence of thermal energy, calculated as follows: ; In the formula, The ambient temperature; The temperature of the heat transfer medium.

[0046] The three objective functions described above comprehensively characterize the overall performance of the planning scheme from the three dimensions of economy, environmental protection and technological advancement, respectively, and constitute a fitness evaluation system for multi-objective optimization.

[0047] S2.3 Fitness Assignment; By performing hourly simulations and calculating the objective function in step S2, the combination of objective function values ​​corresponding to each integrated energy system planning scheme can be obtained. The combination of objective function values ​​serves as the fitness representation of the planning scheme in the current evolutionary iteration, characterizing its comprehensive performance in the multi-objective optimization process. In subsequent multi-objective evolutionary optimization processes, minimizing the annual total cost objective function and the net carbon emission objective function, as well as maximizing the overall efficiency objective function, are used as optimization directions to screen, rank, and update the planning schemes. Through this fitness assignment method, planning decision variables that originally only represented installed capacity are transformed into quantitative indicators that can be compared and evaluated in a multi-objective space, clearly reflecting the differences between different planning schemes in dimensions such as economic efficiency, environmental protection, and energy efficiency.

[0048] S3. During the genetic iteration process, the current parent population and the offspring population are merged to obtain a merged population. Using a non-dominated sorting method based on sequential search, all planning scheme individuals in the merged population are divided into non-dominated levels to obtain multiple non-dominated levels sorted by non-dominated priority. According to the order of non-dominated priority from high to low, the planning scheme individuals in each non-dominated level are added to the next generation parent population in turn.

[0049] The non-dominated sorting method based on sequential search includes: pre-sorting all planning scheme individuals in the merged population according to a preset objective function sorting strategy to obtain a pre-sorted individual sequence arranged in a fixed order; and judging the dominance relationship of each planning scheme individual in turn according to the order of the pre-sorted individual sequence, and assigning each planning scheme individual to the lowest non-dominated level that meets the admission criteria.

[0050] Furthermore, the preset objective function sorting strategy includes: sorting the objective function values ​​of individual planning schemes in ascending or descending order according to the preset objective function priority; and sorting them according to the objective function values ​​of the next priority when the objective function values ​​of the previous priority are the same, until the pre-sorting of all individual planning schemes is completed.

[0051] In this embodiment, an efficient non-dominated sorting method based on ENS-SS is adopted, and the specific implementation process is as follows: In step two, the corresponding multi-objective fitness vector has been calculated for each integrated energy system planning scheme. Based on this, the fitness vectors of all planning schemes are used as input, and the non-dominated ranking method based on ENS-SS is used to rank and classify these schemes to identify which are better and more promising solutions.

[0052] S3.1 Pre-sorting process; First, the parent population and the offspring population generated by subsequent genetic operations are merged to form a temporary population of size 2N. When the initial population is generated, there are no offspring. At this time, step S5 needs to be executed first to create offspring, and then step S3 is executed.

[0053] To improve the efficiency of subsequent comparisons and sorting, the merged population is first pre-sorted: arranged in ascending order according to the value of the first objective function. If two individuals... If the values ​​are equal, they are sorted according to the second objective function, and so on, which creates favorable conditions for subsequent order comparisons and greatly reduces the number of comparisons.

[0054] S3.2 Sequential Comparison and Hierarchical Assignment; After pre-sorting the candidate populations, an ENS-SS-based sequential non-dominated sorting method is employed to compare each planning scheme and assign non-dominated levels. The pre-sorting results reduce invalid comparisons, and sequential insertion assigns each planning scheme to the lowest non-dominated level it can access. Compared to traditional non-dominated sorting methods that require pairwise comparisons of all individuals, ENS-SS significantly reduces the number of dominance relationship judgments through pre-sorting and sequential insertion strategies. While ensuring the accuracy of the sorting results, it effectively reduces computational complexity, making it particularly suitable for scenarios involving large-scale populations and long-term simulations.

[0055] First, an empty set of non-dominated levels is initialized to store the assigned non-dominated frontiers sequentially. Then, the planning schemes in the candidate population are processed one by one according to the pre-sorted order. For the first planning scheme processed, since there are currently no assigned schemes, no dominance relationship judgment is needed, and it is directly assigned to the first non-dominated frontier (…). ).

[0056] Secondly, for each planning scheme processed subsequently, they are processed sequentially from the non-dominated hierarchy number in ascending order. Begin determining the dominance relationship. Assume the current planning scheme is... Its corresponding multi-objective fitness vector is ;in, and To minimize the objective, To maximize the goal.

[0057] The rule for determining dominance is defined as follows: if there is a non-dominant frontier... There are planning schemes in it. Such that the following condition is satisfied on all minimization objectives. , And satisfies the following in maximizing the objective. If it has a strict advantage on at least one objective function, then it is determined that... Dominate .

[0058] During the judgment process, if the current planning scheme At a certain non-dominant frontier There is no planning scheme governing it. This means If the conditions for entering the non-dominated frontier are met, it should be assigned to that frontier, and the comparison process for subsequent higher-level non-dominated frontiers should be terminated.

[0059] Secondly, if the current planning scheme In the current non-dominant frontier If a program is dominated by at least one planning scheme, it indicates that it does not meet the conditions for entering that frontier and needs to continue to interact with the next level of non-dominated frontier. Make the same judgment on the dominance relationship.

[0060] The above judgment process proceeds sequentially.

[0061] Current planning scheme All existing non-dominant frontiers When all frontiers are dominated by at least one planning scheme, a new non-dominated frontier with a later hierarchical number is created. The plan was then allocated to the newly constructed frontier.

[0062] S3.3 Hierarchical processing and population selection; Through the above-described order non-dominated sorting process based on ENS-SS, the candidate population of size 2N was divided into multiple non-dominated levels arranged from best to worst. Among them, the first non-dominant frontier The planning schemes in the current population are independent of each other and constitute an approximate representation of the Pareto optimal solution set in the current iteration.

[0063] Based on this, N planning schemes need to be selected from the above 2N planning schemes to form the parent population for the next generation of evolutionary computation. The selection process follows the principle of non-dominant hierarchy priority, that is, priority is given to non-dominant hierarchical hierarchies. All planning schemes are added to the parent population; when the size of the parent population is still smaller than the preset size N, the planning schemes of different levels are added as a whole according to the order from best to worst.

[0064] When a nondominated frontier is introduced If adding all the planning schemes within the frontier would cause the parent population size to exceed the preset size N, then this non-dominant frontier is identified as a critical frontier. In this case, further screening of the planning schemes within the critical frontier is needed to determine the final set of retained schemes. This internal screening process is completed using the crowding distance calculation method in the next step.

[0065] Through the above-mentioned hierarchical processing and parental population selection mechanism, while ensuring the overall non-dominance of the parental population, a clear hierarchical boundary and judgment basis are provided for subsequent refined selection to maintain population diversity.

[0066] S4. During the process of adding planning scheme individuals from each non-dominated level to the next generation parent population, the cumulative number of individuals in the next generation parent population is counted in real time. When the cumulative number of individuals exceeds the preset population size threshold after adding all planning scheme individuals from the target non-dominated level, the target non-dominated level is determined as the critical frontier. The crowding distance is calculated for all planning scheme individuals within the critical frontier, and the corresponding number of planning scheme individuals are selected from the critical frontier based on the crowding distance to be added to the next generation parent population. The number of individuals in the next generation parent population is equal to the preset population size threshold.

[0067] In this embodiment, the specific implementation process of crowding distance calculation and critical front screening is as follows: After completing the non-dominated sorting in step S3, planning schemes are selected in descending order of non-dominated hierarchy to construct the next generation parent population. When the total number of planning schemes exceeds the preset population size N after adding a certain non-dominated front, the non-dominated front is defined as the critical front.

[0068] S4.1 Identification of the critical frontier; In the process of sequentially adding planning schemes from each non-dominated frontier to the parent population, starting from the first non-dominated frontier and proceeding hierarchically, let S be the total number of planning schemes from the first few non-dominated frontiers already introduced. As the next non-dominated frontier is introduced... All planning schemes in the plan will cause the parent population size S to exceed the preset size N, at which point the non-dominant front is determined. It is the critical frontier.

[0069] In this case, all programming schemes in the first k-1 non-dominated frontiers are fully preserved, while the critical frontier needs to be considered. Further selection yielded NS planning schemes to supplement the parent population until its size reached exactly N. This critical front identification process provided clear processing targets and quantitative constraints for subsequent refined selection operations.

[0070] S4.2 Calculation of congestion distance; To assess the critical frontier The diversity of planning schemes in the region, and further screening of these schemes within the frontier, are all considered. Calculate the crowding distance for each individual. Specifically, for each objective function... First, the frontier All individuals are sorted according to the magnitude of the objective function value. For the two individuals located at the boundary after sorting, that is, the individuals with the largest and smallest values ​​on the objective function, their crowding distance component on the objective function is assigned a large value (such as infinity) to ensure that these boundary solutions located on the Pareto front boundary, which are used to define the current front range, can be preferentially preserved.

[0071] For any individual i in the middle of the sort, its objective function The crowding component is defined as the normalized difference between two adjacent individuals on the objective function value, and the calculation formula is: ; In the formula, and It is the entire frontier In the target The maximum and minimum values ​​on the scale are used for normalization to eliminate the influence of different target dimensions; Let be the crowding distance component of individual i on the m-th objective function; , Indicating at the current forefront Under sorting, the values ​​of the next and previous individuals adjacent to individual i on the m-th objective function.

[0072] Individual crowding distance It is the sum of the crowding components on all its objective functions, and the formula is: ; In the formula, The total number of objective functions represents three objectives in this embodiment: total annual cost, net carbon emissions, and overall energy efficiency. A larger crowding distance indicates that individual i is farther from other individuals in the objective space, resulting in a sparser solution distribution in its region and greater uniqueness in maintaining the breadth and diversity of the Pareto front. During the selection process within the critical front, individuals with larger crowding distances are preferentially retained to construct a parent population that balances solution set quality and diversity.

[0073] S4.3 Filtering based on crowding distance; In obtaining the critical frontier Crowding distance between individuals Then, N is further screened from this critical frontier. S individuals join the next generation of the parent population. To ensure the diversity of the solution set while achieving accurate selection under the constraint of quantity, the selection process adopts a binary tournament selection strategy based on crowding distance.

[0074] Specifically, at the critical frontier Two candidate individuals are randomly selected, and their corresponding crowding distance values ​​are compared. The individual with the larger crowding distance is retained and added to the parent population. Individuals not selected are not included in the subsequent selection process. This random comparison and selection operation is repeated until the number of individuals selected from the critical front reaches the required N. S.

[0075] S5. After each round of genetic iteration completes the construction of the next generation of parent population, determine whether the genetic iteration meets the preset iteration stopping condition. If the preset iteration stopping condition is not met, the next generation of parent population is used as the current parent population for the new round of genetic iteration, and the genetic iteration operation is performed. Under the preset iteration stopping condition, the population corresponding to the current iteration round is taken as the optimized population, and the Pareto optimal solution set corresponding to the optimized population is output.

[0076] In this embodiment, the genetic iteration operation generates the offspring population by performing genetic operations on the parent population. The specific implementation process is as follows: After identifying N individuals from the merged population as the parent population through non-dominated sorting and crowding distance screening, genetic operations are performed on the parent population to generate a new generation of offspring. These genetic operations include simulated binary crossover and polynomial mutation, used to introduce new search perturbations while inheriting the superior traits of the parents, thus expanding the exploration range of the solution space.

[0077] S5.1 Simulated Binary Cross (SBX); First, two individuals are randomly selected from the parent population as parent individuals. and Then, based on a preset crossover probability... (The range is typically between 0.5 and 1.0), determine whether to perform a crossover operation on the parent pair. For each decision variable, generate a random number that follows a uniform distribution in the interval [0,1). ,like If the value is positive, then the following cross-calculation is performed on the variable to generate two child values; otherwise, the child value directly inherits the value of the corresponding parent value.

[0078] If crossover is performed, for each pair of decision variables, two child values ​​are generated as follows: ; In the formula, , These are the variable values ​​of the parent individual; , These are the values ​​of the generated child variables.

[0079] It is the cross-expansion factor, which is formed by another random number uniformly distributed in the interval [0,1). and the preset distribution index (The range is usually between 5 and 20) to decide.

[0080] ; Distribution index The degree of proximity between offspring and parents is controlled; the higher the value, the closer the offspring are to their parents. By simulating binary crossover, the population's distribution in the decision space can be enhanced while maintaining the mean characteristics of the parents.

[0081] S5.2 Polynomial Variation; For each offspring generated by the crossover, iterate through each of its decision variables and apply a pre-defined mutation probability. (Usually a value less than the crossover probability, such as 0.01, 0.05, or 0.1), determine whether to perform a mutation operation on this variable. For each decision variable, generate a random number that follows a uniform distribution in the interval [0,1). ,like If the variable is mutated, then the mutation calculation is performed; otherwise, the offspring directly inherits the corresponding value from the parent.

[0082] If a mutation is performed, modify the device variable value c according to the following formula: ; In the formula, and These are the lower and upper bounds of the variable.

[0083] Variational perturbation Given a random number that is uniformly distributed in the interval [0,1). and the preset variation distribution index (The range is usually between 5 and 20) The calculation shows: ; S5.3 Iterative Loop; Through the selection, crossover, and mutation operations described above, a progeny population of size N is generated. Subsequently, the parent population and the progeny population are merged to form a new candidate population of size 2N, and the process returns to the non-dominated sorting and crowding distance screening steps to enter the next round of evolutionary iteration.

[0084] The algorithm repeats the above evolutionary process until a preset termination condition is met. The termination condition can be set as reaching the maximum number of iterations (e.g., 150 generations), or the Pareto front of the population not showing significant improvement for several consecutive generations. When the iteration terminates, the final Pareto optimal solution set is output as a set of candidate optimal planning schemes for the multi-objective programming problem of the integrated energy system.

[0085] S6. Evaluate the individual planning schemes in the Pareto optimal solution set using multiple indicators, and output the planning scheme based on the results of the multi-indicator evaluation.

[0086] The multi-index evaluation adopts the TOPSIS evaluation method. In the multi-index evaluation, the objective function value corresponding to each planning scheme in the Pareto optimal solution set is used as the evaluation index. A standardized evaluation matrix is ​​constructed through positive and standardized processing. Based on the standardized evaluation matrix, the positive ideal solution and the negative ideal solution are determined. The relative proximity of each planning scheme to the positive and negative ideal solutions is calculated, and all planning schemes are ranked according to their relative proximity. The planning scheme with the best ranking among all planning schemes is output as the planning scheme.

[0087] In this embodiment, the specific implementation process of the TOPSIS comprehensive evaluation is as follows: After multiple generations of evolution, the improved NSGA-II algorithm will output a Pareto optimal solution set, which contains multiple integrated energy system planning schemes that are independent of each other in terms of target dimensions such as total annual cost, net carbon emissions, and overall efficiency.

[0088] Since each planning scheme in the Pareto optimal solution set represents a trade-off between different objectives, it is difficult to directly determine a unique engineering implementation scheme. Therefore, after obtaining the final Pareto optimal solution set, a further comprehensive evaluation method is needed to rank and select the best candidate planning scheme, thereby outputting an optimal compromise planning scheme. For this purpose, the TOPSIS comprehensive evaluation method is used in this step to perform post-processing analysis on the Pareto optimal solution set.

[0089] S6.1 Constructing the decision matrix and data preprocessing; First, all candidate planning schemes are extracted from the final Pareto optimal solution set, assuming there are K planning schemes in total. Each planning scheme corresponds to M evaluation indicators. In this method, the evaluation indicators are directly adopted from the objective function values ​​obtained in the multi-objective optimization process, and M=3, representing the total annual cost, net carbon emissions, and overall efficiency, respectively, thus forming a K×M decision matrix D.

[0090] Because the three objectives (cost, carbon emissions, and energy efficiency) have different dimensions and tendencies (lower costs and emissions are better, higher energy efficiency is better), data preprocessing is required. This includes: (1) Forwarding: Since different evaluation indicators have different physical meanings and optimization directions, the original data needs to be preprocessed to ensure the consistency and comparability of subsequent evaluations. Considering that the annual total cost and net carbon emissions are "cost-type" indicators (the smaller the value, the better), while the overall efficiency is a "benefit-type" indicator (the larger the value, the better), all indicators need to be converted into "benefit-type" indicators first.

[0091] For cost-related indicators, a linear positive transformation method is used to convert them into: ; In the formula, This represents the original value of the j-th indicator for the i-th planning scheme; Let j be the maximum value of the j-th evaluation index among all planning schemes; The value is obtained by positive transformation of the j-th indicator value of the i-th planning scheme.

[0092] (2) Standardization: After completing the forward transformation of the objective function, in order to eliminate the differences in dimensions and numerical ranges among different objective functions, the forward-transformed objective function matrix is ​​standardized. Vector normalization is used to normalize each objective function column separately, and the calculation formula is as follows: ; In the formula, The value of the j-th indicator in the i-th planning scheme is obtained after standardization. Thus, the standardized decision matrix is ​​obtained. .

[0093] S6.2 Determine the ideal solution and calculate the distance; Based on the standardized decision matrix, positive and negative ideal solutions are constructed respectively. A positive ideal solution represents a hypothetical solution that achieves optimal performance across all evaluation indicators, while a negative ideal solution represents a hypothetical solution that performs worst performance across all evaluation indicators. Their definitions are as follows: (a) Positive ideal solution This consists of the optimal value of each indicator across all options. For indicators that have been transformed into benefit-based indicators, this is the maximum value in each column. ; (b) Negative ideal solution : Consists of the worst-case value of each indicator across all possible scenarios, i.e., the minimum value in each column: ; For each planning scheme in the Pareto optimal solution set, calculate its Euclidean distance to both the positive and negative ideal solutions to measure its approximation to the ideal state in the comprehensive evaluation space. The difference between the i-th planning scheme and the positive ideal solution is... The difference between the i-th planning scheme and the negative ideal solution is The formulas are as follows: ; ; in, The smaller the value, the closer the solution i is to the ideal state where all aspects are optimal; The larger the value, the further away from the worst-case scenario in terms of all aspects.

[0094] S6.3 Calculate proximity and sort the schemes; Based on the distance calculation results above, the relative proximity of each planning scheme is calculated: ; In the formula, Let be the proximity of the i-th planning scheme to the ideal solution. Therefore, 0 ≤ ≤1. When When = 1, it means that the solution is a positive ideal solution; when When the value is 0, it means that the solution is a negative ideal solution.

[0095] The larger the value, the better the overall performance of the scheme among all alternatives, and the closer it is to the ideal state. Based on the relative proximity, the planning schemes in the Pareto optimal solution set are ranked, and the planning scheme with the largest relative proximity is selected as the final recommended scheme output for integrated energy system planning optimization. This scheme is the most balanced and closest to the ideal point among many non-dominated schemes after comprehensively considering economic costs, environmental protection, and energy efficiency.

[0096] Output the decision variable vector corresponding to this scheme. This yields a set of specific and optimal recommendations for the installed capacity configuration of various energy equipment, which can be directly used to guide the actual planning and construction of integrated energy systems.

[0097] Finally, it should be noted that the above content is only used to illustrate the technical solution of this application, and is not intended to limit the scope of protection of this application. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of this application shall not depart from the substance and scope of the technical solution of this application.

Claims

1. An energy system planning method based on a multi-objective genetic algorithm, characterized in that, include: Determine the decision variables for energy system planning, and generate an initial parent population based on the decision variables; Genetic iteration is performed based on the initial parent population and the multi-objective genetic algorithm; During the genetic iteration process, for each planning scheme individual in the current parent and offspring populations, a simulation is performed hourly throughout the year to calculate the multinomial objective function value corresponding to each planning scheme individual. The objective function value is then used as the fitness of the corresponding planning scheme individual for the genetic iteration. During the genetic iteration process, the current parent population and the offspring population are merged to obtain a merged population. A non-dominated sorting method based on sequential search is used to divide all planning scheme individuals in the merged population into non-dominated levels, resulting in multiple non-dominated levels sorted by non-dominated priority. According to the order of non-dominated priority from high to low, the planning scheme individuals in each of the non-dominated levels are added to the next generation parent population in sequence. If the genetic iteration satisfies the preset iteration stopping condition, output the Pareto optimal solution set corresponding to the current population; The individual planning schemes in the Pareto optimal solution set are evaluated using multiple indicators, and the planning scheme is output based on the results of the multi-indicator evaluation.

2. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, The decision variables include the installed capacity of various energy equipment in the energy system; In the initial parent population, each individual planning scheme includes energy equipment installed capacity parameters corresponding to the decision variables.

3. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, During the process of sequentially adding the planning scheme individuals from each of the non-dominated levels to the next generation parent population, the cumulative number of individuals in the next generation parent population is counted in real time. When the cumulative number of individuals exceeds a preset population size threshold after all planning scheme individuals in the target non-dominated level are added, the target non-dominated level is determined as the critical frontier. The crowding distance is calculated for all planning scheme individuals within the critical frontier, and a corresponding number of planning scheme individuals are selected from the critical frontier based on the crowding distance to be added to the next generation parent population. The number of individuals in the next generation parent population is equal to the preset population size threshold.

4. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, After each round of genetic iteration completes the construction of the next generation of parent population, it is determined whether the genetic iteration meets the preset iteration stopping condition; If the preset iteration stopping condition is not met, the next generation parent population is used as the current parent population for a new round of genetic iteration, and the genetic iteration operation is performed. If the preset iteration stopping condition is met, the population corresponding to the current iteration round is taken as the optimized population, and the Pareto optimal solution set corresponding to the optimized population is output.

5. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, The non-dominated sorting method based on sequential search includes: For all planning scheme individuals in the merged population, pre-sorting is performed according to a preset objective function sorting strategy to obtain a pre-sorted sequence of individuals arranged in a fixed order; According to the order of the pre-sorted individual sequence, the dominance relationship of each planning scheme individual is judged in turn, and each planning scheme individual is assigned to the lowest non-dominance level that meets the admission conditions.

6. The energy system planning method based on a multi-objective genetic algorithm according to claim 5, characterized in that, The preset objective function sorting strategy includes: According to the pre-set priority of the objective function, the objective function values ​​corresponding to the individual planning schemes are arranged in ascending or descending order in turn. If the objective function values ​​of the previous priority are the same, sort them according to the objective function values ​​of the next priority, until the pre-sorting process of all individual planning schemes is completed.

7. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, Before executing the hourly simulation throughout the year, the basic input data for the simulation is obtained in advance. The simulation input data includes hourly meteorological data for the target planning area throughout the year, hourly multi-category energy load demand data throughout the year, and technical and cost parameters of various energy equipment.

8. The energy system planning method based on a multi-objective genetic algorithm according to claim 7, characterized in that, The values ​​of the multiple objective functions include the total annual cost, net carbon emissions, and efficiency of the corresponding planning scheme. The multiple objective functions include minimizing annual total cost and net carbon emissions, and maximizing efficiency.

9. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, The initial parent population is generated using the Tent chaotic mapping method, including: Generate a chaotic sequence with traversal characteristics within a preset numerical range; The chaotic sequence is transformed into the preset value constraint range corresponding to the decision variable through linear mapping to obtain multiple sets of individual planning schemes that meet the constraint conditions; The initial parent population is generated based on the multiple groups of individuals in the planning scheme.

10. The energy system planning method based on a multi-objective genetic algorithm according to claim 1, characterized in that, The multi-index evaluation adopts the TOPSIS evaluation method; In the multi-index evaluation, the objective function value corresponding to each planning scheme in the Pareto optimal solution set is used as the evaluation index, and a standardized evaluation matrix is ​​constructed through positive transformation and standardization. Based on the standardized evaluation matrix, the positive ideal solution and the negative ideal solution are determined. The relative proximity of each planning scheme individual to the positive ideal solution and the negative ideal solution is calculated, and all planning scheme individuals are ranked according to the relative proximity. The best-ranked planning scheme among all individual planning schemes is output as the planning scheme.

Images