A multi-objective collaborative planning method and system for a micro-grid with carbon emission constraints

Abstract

This invention proposes a multi-objective collaborative planning method and system for microgrids with carbon emission constraints. The method constructs a multi-objective collaborative planning model for the microgrid, which includes a set of decision variables to be optimized, a multi-objective function, and constraints to define the feasible operating boundary of the microgrid. Based on the barrel theory optimization algorithm, the multi-objective collaborative planning model is solved. The performance of each individual "barrel" in the barrel theory optimization algorithm is normalized according to multi-objectives, and performance bottlenecks are identified. The iterative update mechanism of the barrel theory optimization algorithm is executed, driving the evolution and updating of the barrel population through dynamic optimization of the identified performance bottlenecks. After satisfying the preset algorithm termination conditions, the obtained Pareto optimal solution set is output. This Pareto optimal solution set is used as the microgrid capacity configuration planning scheme, improving the reliability of the multi-objective collaborative planning for the microgrid.

Description

Technical Field

[0001] This invention relates to the field of power system planning and operation technology, and in particular to a multi-objective collaborative planning method and system for microgrids with carbon emission constraints. Background Technology

[0002] Against the backdrop of global efforts to address climate change and promote energy structure transformation, renewable energy technologies, represented by solar photovoltaic and wind power, are developing rapidly, and their penetration rate in power systems is growing at an unprecedented rate. Microgrids, as a modern small-scale power generation and distribution system capable of effectively integrating and managing distributed power sources, energy storage systems, controllable loads, and monitoring and protection devices, have become a key technological path for promoting local consumption of renewable energy, improving grid resilience, and enhancing power supply quality. The core advantage of microgrids lies in their flexible operating modes. They can operate in parallel with the main grid to achieve bidirectional energy exchange and complementarity, or seamlessly switch to independent operation mode when the main grid fails or when located in remote, grid-free areas, ensuring continuous power supply to critical loads.

[0003] However, the full realization of a microgrid's value hinges on the scientific rigor and rationality of its initial planning and design. Among these, the capacity configuration of various distributed power sources and energy storage systems is the most crucial and complex aspect of the planning phase. An optimized capacity configuration scheme must strike the best balance among several often conflicting performance dimensions. From a traditional perspective, the primary goal of microgrid planning is economic efficiency, aiming to minimize the total lifecycle cost (LCC) while meeting basic power supply needs. With increasing societal demands for power supply reliability, quantifying and optimizing the system's power supply guarantee capabilities—such as minimizing the probability of load failure (LOLP) or annual power shortage—has become an indispensable key consideration in planning and design.

[0004] In recent years, with the introduction of carbon peaking and carbon neutrality strategic goals, the green and low-carbon transformation of energy systems has been elevated to an unprecedented level. Therefore, the environmental benefits of microgrids, especially their total life cycle emissions (LCE) throughout their entire lifecycle (equipment manufacturing, operation, and disposal), have shifted from an additional consideration to a core planning objective alongside economic efficiency and reliability. This has transformed the microgrid capacity allocation problem from a single-objective or bi-objective optimization problem into a typical and more challenging multi-objective optimization problem. Decision-makers no longer need a single optimal solution, but rather a set of non-dominated solutions that effectively balance cost, reliability, and low carbon emissions—the Pareto optimal frontier—in order to make the most appropriate final decision based on specific policy orientations, budgets, and user needs.

[0005] To solve such complex multi-objective optimization problems, various multi-objective intelligent optimization algorithms have been widely adopted in academia and industry. Among them, evolutionary algorithms, such as the Non-Dominated Sorting Genetic Algorithm (NSGA-II), the Multi-Objective Particle Swarm Optimization Algorithm (MOPSO), and the Multi-Objective Differential Evolutionary Algorithm (MODE), have been widely used due to their advantages such as not requiring gradient information and strong adaptability to the solution space shape. However, these classic general-purpose optimization algorithms still reveal some inherent technical bottlenecks when dealing with the specific problem of multi-objective collaborative planning in microgrids. First, it is often difficult to strike a balance between exploration and utilization. In pursuit of rapid convergence, the population may prematurely cluster in a local region of the Pareto front, leading to a loss of diversity and an inability to discover a more widely distributed and better-performing global optimal solution set. Conversely, if diversity is overemphasized, the algorithm may converge slowly and incur high computational costs. Second, the search operators of existing algorithms are usually universal, such as crossover, mutation, and particle flight, failing to fully utilize the inherent structural information of multi-objective conflicts in microgrid planning problems. For example, one candidate solution might have extremely low cost but very poor reliability, while another might have excellent reliability but exceed carbon emission limits. Traditional algorithms often treat all solutions equally, randomly perturbing and exchanging information, lacking an intelligent mechanism capable of identifying and specifically improving the weaknesses of each solution. Finally, microgrid operation models contain numerous complex nonlinear constraints and dynamic processes. Traditional algorithms often reduce search efficiency when dealing with these strong constraints by generating a large number of infeasible solutions.

[0006] The "barrel theory" in management and systems engineering, also known as the "weakest link effect," presents a profound systemic perspective: the overall performance or capacity of a system is not determined by its strongest part, but rather by its weakest link. This principle has broad applicability and can be cleverly transferred to the field of multi-objective optimization. For a candidate solution to a multi-objective optimization problem, its overall evaluation is also largely constrained by the worst-performing objective dimension. A solution that is nearly perfect in terms of economy and reliability, if its carbon emissions are extremely high, remains unacceptable under current stringent environmental policies, indicating a significant weakness.

[0007] Therefore, there is an urgent need for a novel optimization algorithm that can break free from the traditional design paradigm of optimization algorithms, draw inspiration from systems theory, and creatively design an algorithm capable of intelligently identifying and correcting performance shortcomings of candidate solutions. Such an algorithm would more profoundly align with the essence of multi-objective optimization problems, thereby solving the complex problem of microgrid capacity coordination planning—a problem with both theoretical challenges and engineering value—more efficiently and in a more balanced manner. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention innovatively proposes a multi-objective collaborative planning method and system for microgrids with carbon emission constraints. This effectively solves the problem of low reliability in multi-objective collaborative planning of microgrids caused by existing technologies, and effectively improves the reliability of multi-objective collaborative planning of microgrids.

[0009] The first aspect of this invention provides a multi-objective collaborative planning method for microgrids with carbon emission constraints, comprising: A multi-objective collaborative planning model for microgrids is constructed, which includes a set of decision variables to be optimized, a multi-objective function, and constraints used to define the feasible operating boundary of the microgrid. Based on the barrel theory optimization algorithm, the multi-objective collaborative planning model of microgrid is solved. The multi-objective performance of each individual barrel in the barrel theory optimization algorithm is normalized and the performance bottleneck is identified. The iterative update mechanism of the barrel theory optimization algorithm is implemented to drive the evolution and renewal of the barrel population by dynamically optimizing the shortcomings in recognition performance. After the preset algorithm termination condition is met, the Pareto optimal solution set is output, and the Pareto optimal solution set is used as the microgrid capacity configuration planning scheme.

[0010] A second aspect of the present invention provides a multi-objective cooperative planning system for microgrids with carbon emission constraints, comprising: The module constructs a multi-objective collaborative planning model for a microgrid. The multi-objective collaborative planning model for a microgrid includes a set of decision variables to be optimized, a multi-objective function, and constraints used to define the feasible operating boundary of the microgrid. The solution module solves the microgrid multi-objective collaborative planning model based on the barrel theory optimization algorithm, and performs multi-objective performance normalization on the individual barrels in the barrel theory optimization algorithm and identifies the performance bottlenecks. The execution module implements the iterative update mechanism of the barrel theory optimization algorithm, which drives the evolution and update of the barrel population by dynamically optimizing the shortcomings in recognition performance. The output module, after satisfying the preset algorithm termination conditions, outputs the obtained Pareto optimal solution set, and uses the Pareto optimal solution set as the microgrid capacity configuration planning scheme.

[0011] The technical solution adopted in this invention has the following technical effects: 1. The technical solution of this invention constructs a multi-objective collaborative planning model for microgrids. This model includes a set of decision variables to be optimized, a multi-objective function, and constraints to define the feasible operating boundary of the microgrid. Based on the barrel theory optimization algorithm, the multi-objective collaborative planning model is solved. The performance of each individual barrel in the barrel theory optimization algorithm is normalized according to multi-objectives, and performance bottlenecks are identified. The iterative update mechanism of the barrel theory optimization algorithm is executed, driving the evolution and updating of the barrel population through dynamic optimization of the identified performance bottlenecks. After satisfying the preset algorithm termination conditions, the obtained Pareto optimal solution set is output. This Pareto optimal solution set is used as the microgrid capacity configuration planning scheme, effectively solving the problem of low reliability in multi-objective collaborative planning of microgrids caused by existing technologies, and effectively improving the reliability of multi-objective collaborative planning of microgrids.

[0012] 2. The technical solution of this invention utilizes a bottleneck identification and adaptive repair mechanism to keenly identify the performance bottlenecks of each candidate solution in the multi-dimensional target space and perform targeted, asymmetric optimization. This intelligent evolutionary approach of addressing deficiencies effectively avoids the phenomenon of traditional algorithms potentially exploiting strengths and mitigating weaknesses under random perturbations, thus more easily guiding the population towards a region where performance is balanced across all target dimensions. Therefore, the solutions in the Pareto optimal solution set obtained by this invention typically possess higher overall performance, meaning they do not exhibit extreme performance deficiencies in any particular dimension, resulting in greater practicality and acceptability.

[0013] 3. The technical solution of this invention constructs an efficient exploration-exploitation collaborative framework by organically combining and dynamically switching three strategies: adaptive repair, elite guidance, and global perturbation. The elite guidance strategy ensures that the algorithm can quickly converge to the discovered high-quality region (Pareto front), demonstrating strong exploitation capabilities; the adaptive repair strategy, within a local region, deeply mines and finely adjusts the weak dimensions of the solution, representing a more purposeful exploration; and the global perturbation strategy based on Lévy flight provides the algorithm with the ability to escape local optima and discover entirely new promising regions. The synergistic effect of these three strategies enables the algorithm to navigate efficiently in a complex, multi-modal solution space. Compared to traditional algorithms, it can converge to a Pareto front that is superior in terms of dominance and distribution breadth with fewer iterations. Furthermore, through the synergistic effect of the three mechanisms, the algorithm's sensitivity to individual hyperparameters is relatively reduced, resulting in more stable and robust overall performance.

[0014] 4. The technical solution of this invention cleverly transforms the abstract mathematical optimization problem into an intuitive and easy-to-understand physical process of repairing the weakest link in the performance of the barrel. This design makes it naturally and well-suited for typical high-dimensional, strongly nonlinear, multi-constrained black-box optimization problems such as microgrid planning, where the objective function evaluation requires time-consuming simulation.

[0015] 5. During the optimization process, the technical solution of this invention can track and record the dynamic evolution history of the shortcomings of each solution. By analyzing which objective (cost, reliability, or carbon emissions) more frequently becomes the shortcoming of most solutions in the population at different optimization stages, decision-makers can gain profound insights into the inherent contradictions of this specific planning problem. For example, if cost (LCC) generally becomes the shortcoming in the later stages of optimization, it indicates that after meeting certain reliability and environmental protection requirements, further cost reduction becomes extremely difficult, becoming the main bottleneck for improving system performance. This process-oriented diagnostic information transcends the limitations of traditional algorithms that only provide the final result, and can provide planners with valuable auxiliary decision-making basis for understanding the performance trade-off bottlenecks under different capacity configurations.

[0016] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1This is a flowchart illustrating the method of Embodiment 1 in the present invention; Figure 2 This is another flowchart illustrating the method of Embodiment 1 in the present invention; Figure 3 This is a schematic diagram of a typical microgrid system structure to which the method of Embodiment 1 of the present invention is applied; Figure 4 This is a schematic diagram illustrating the concept of applying the "barrel theory" to multi-objective optimization of microgrids, used to explain the core idea of ​​the present invention in the method of Embodiment 1 of the present invention. Figure 5 This is a two-dimensional schematic diagram comparing the Pareto optimal frontier obtained by the method of the present invention with two existing technologies (NSGA-II and MOPSO) in the method of Embodiment 1 of the present invention. Figure 6 This is a schematic diagram of the convergence curve of the average value of the three optimization objectives in the population as a function of the number of iterations during the iterative optimization process in the method of Embodiment 1 of the present invention. Figure 7 This is a schematic diagram of the system structure in Embodiment 2 of the present invention. Detailed Implementation

[0019] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure of the invention, components and arrangements of specific examples are described below. Furthermore, reference numerals and / or letters may be repeated in different examples. This repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. It should be noted that the components illustrated in the drawings are not necessarily drawn to scale. Descriptions of well-known components, processing techniques, and processes are omitted in this invention to avoid unnecessarily limiting the invention.

[0020] Example 1 To effectively address the problems of existing technologies, this invention proposes a novel multi-objective collaborative planning method for microgrids with carbon emission constraints, named the Barrel Theory-based Optimizer (BTO) algorithm. The core idea of ​​this invention is to abstract each microgrid capacity configuration scheme to be evaluated—a candidate solution to the optimization problem—as a barrel, and to abstract the three conflicting optimization objectives—economy (LCC), reliability (LOLP), and low carbon emissions (LCE)—as the three planks constituting this barrel. The height of the plank, or more precisely, its integrity, is directly related to the performance under that objective. The overall performance of a scheme, like the water capacity of a barrel, is limited by its shortest plank. Based on this metaphor, the BTO algorithm designed in this invention drives the entire barrel population to co-evolve towards a higher water level, i.e., better overall performance, by simulating three interlocking dynamic mechanisms: adaptive repair based on short plank identification, elitist-guided asymptotic convergence, and global balance perturbation adjustment.

[0021] like Figures 1-2 As shown, this invention provides a multi-objective collaborative planning method for microgrids with carbon emission constraints, comprising: S1. Construct a multi-objective collaborative planning model for microgrids. The multi-objective collaborative planning model for microgrids includes a set of decision variables to be optimized, a multi-objective function, and constraints used to define the feasible operating boundary of the microgrid. S2, based on the barrel theory optimization algorithm, solve the microgrid multi-objective collaborative planning model, normalize the multi-objective performance of each individual barrel in the barrel theory optimization algorithm and identify the performance bottleneck; S3 executes the iterative update mechanism of the barrel theory optimization algorithm, which drives the evolution and renewal of the barrel population by dynamically optimizing the shortcomings in recognition performance; S4, after satisfying the preset algorithm termination condition, outputs the obtained Pareto optimal solution set, and uses the Pareto optimal solution set as the microgrid capacity configuration planning scheme.

[0022] Wherein, step S1 corresponds to Figure 2 Step one in this process is fundamental to optimization, and its purpose is to accurately describe the practical engineering problem to be solved using mathematical language. This model is a typical two-layer optimization structure, with the outer layer being capacity planning optimization and the inner layer being operation scheduling simulation.

[0023] First, define the decision variables for the outer optimization layer. These variables are the core of the planning problem and represent the optimal parameter combination that the algorithm needs to find. In this invention, the decision variable vector... The capacity parameters of key equipment in a microgrid constitute at least the total installed capacity of the photovoltaic (PV) array, denoted as: , in units of peak watt kilowatts (kWp); the total installed capacity of the wind turbine (WT), denoted as , in units of kilowatts (kW); the rated energy capacity of the energy storage system (ESS), denoted as , in units of kilowatt-hours (kWh); the rated power of the power converter (PCS)配套 with the energy storage system, denoted as , in units of kilowatts (kW); and the rated power of the diesel generator (DG)作为备用电源, denoted as , in units of kilowatts (kW). Therefore, a complete decision variable vector can be expressed as . These variables take values within their respective reasonable engineering ranges.

[0024] Secondly, establish conflicting optimization objective functions. The present invention focuses on three core dimensions of economy, reliability, and environmental protection, and constructs three objective functions that need to be minimized simultaneously.

[0025] The first objective function is the minimization of the life cycle cost (LCC), denoted as . This objective aims to evaluate the economy of the solution. Its calculation formula is . In this formula, each cost needs to be converted to the present value at the initial stage of the project. is the initial investment cost, which is the sum of the product of the unit capacity cost of all equipment and its configured capacity. is the operation and maintenance cost during the life cycle, which is the cumulative value after discounting the annual operation and maintenance cost at the discount rate year by year to the initial year. is the equipment replacement cost. For equipment with a lifespan shorter than the project cycle (such as 20 or 25 years), such as energy storage batteries and inverters, the cost incurred in future replacement years needs to be calculated and converted to the present value. is the fossil fuel cost, mainly the fuel cost consumed by the diesel generator during the simulation operation, which also needs to be discounted and accumulated over the life cycle.

[0026] Among them, the life cycle operation and maintenance cost is not calculated by the simple equal annuity present value formula, but the sum of the non-equivalent costs in each year discounted based on the dynamic discount rate of its stage: ; where is a piecewise function, is the year, T is the total project cycle, when belongs to the stage, . Cost model, the calculation formula is ; where is the operation and maintenance cost in the base year, The fitting coefficients can more realistically reflect the cost changes during early equipment failures, mid-term stability, and later aging.

[0027] Equipment replacement cost Calculation, assuming equipment It needs to be replaced during its lifespan. The number of times the change was, and the years of change were respectively The calculation formula is: in, It is equipment Cost of replacement per transaction calculated at current prices. It is a cost discount function resulting from technological progress, an exponential decay model. , The same phased dynamic discount rate is used.

[0028] Fossil fuel costs ( ) calculation, Among them, annual fuel cost . It is based on the first The total fuel consumption was calculated based on the simulated operation of the diesel engine over the past year, including the number of starts, operating time, and load rate. It is the predicted number Annual fuel price per unit.

[0029] This invention relates to The discount accumulation method, compared to traditional single, static methods, differs in its systematic dynamic modeling: using phased discount rates. It replaces a single fixed discount rate, more accurately reflecting changes in risk and capital costs over a long period. It uses a nonlinear function based on equipment physical characteristics and market dynamics to predict the future. and It replaces simple constant or linear models. The calculation depends on a variety of factors, including the year's climate, load, equipment aging, efficiency decline, and electricity pricing strategies. Therefore, it varies from year to year. Each of these is a unique, non-linear calculation result.

[0030] The residual value of all equipment at the end of the project, calculated based on their remaining useful life, is deducted from the total cost as a benefit.

[0031] The second objective function is to minimize the system power failure probability (LOLP), denoted as This objective measures the power supply reliability of the system. Its calculation formula is as follows: .in, To assess the total duration of a cycle year, a year, or 8760 hours, is typically used. As an indicator function, in the inner layer simulation, if at the first... For one hour, the sum of all available power sources in the system, including renewable energy output, energy storage discharge, grid power purchase, and diesel generator generation, is still less than the load demand for that hour. That is, if a power deficit occurs, then The value is 1 if it is not 0 otherwise. The smaller the LOLP value, the lower the percentage of time the system is without power throughout the year, and the higher the reliability.

[0032] The third objective function is to minimize life-cycle carbon emissions (LCE), denoted as This objective is used to assess the environmental friendliness of the proposed solution. Its calculation spans the entire lifecycle of the facility, from cradle to grave, and the formula is: . The implicit carbon emissions during the equipment manufacturing and replacement phases are obtained by multiplying the carbon emission factor per unit capacity of each piece of equipment with its total configuration capacity (including the number of replacements) and summing the results. The direct carbon emissions during the system operation phase mainly come from two parts: first, the emissions generated by the diesel generator burning fuel; and second, the emissions obtained when the microgrid purchases electricity from the external power grid, which is calculated by multiplying the purchased electricity by the average carbon emission factor of the local power grid. Carbon emissions or carbon reductions are generated when equipment is disposed of or recycled at the end of its project life.

[0033] Among them, carbon emissions from equipment manufacturing and replacement ( The calculation is as follows: This involves calculating the implicit carbon emissions of all equipment in the production and manufacturing process. The total configuration capacity of each type of equipment in the system, such as photovoltaic panels, energy storage batteries, inverters, etc., including the capacity of all equipment replaced during the project period, is multiplied by their respective unit capacity manufacturing carbon emission factor, and then the calculation results of all equipment are summed up.

[0034] Direct carbon emissions from system operation ( The calculation is as follows: This refers to calculating the carbon emissions directly generated by a microgrid throughout its entire lifecycle. Fossil fuel emissions are calculated by multiplying the total fuel consumption of diesel generators by the carbon emission factor of the fuel; purchased electricity emissions are calculated by multiplying the total electricity purchased from the external power grid by the average carbon emission factor of the local power grid.

[0035] Carbon emissions from equipment disposal The calculation is as follows: This refers to the carbon emissions generated (or reduced) by the equipment during the scrapping or recycling process after the project's lifespan is completed. It is calculated by multiplying the total amount of each type of scrapped equipment by its corresponding unit disposal / recycling carbon emission factor, and then summing the results. If the recycling process effectively saves energy, this carbon emission factor may be negative, representing a carbon reduction.

[0036] Finally, the constraints that the inner-layer operation scheduling model must satisfy are established. These constraints ensure that the performance evaluation of any given capacity configuration scheme is conducted within the limits allowed by physical and operational rules. The main constraints include: system power balance constraints, i.e., at any given time... The total output of all power sources (photovoltaics, wind power, energy storage, diesel engines, and purchased electricity) must equal the total load (local load and electricity sales); the operating constraints of the energy storage system include the dynamic update equations of its state of charge (SOC). ( for Charge / discharge power (SOC) and upper and lower limits constraints In addition, there are constraints on charging and discharging power not exceeding the rated power of the converter; output constraints of each distributed power source, such as photovoltaic and wind power output being limited by current meteorological resources, and diesel engine output having its minimum starting power and maximum power limit; and power interaction constraints with the main power grid, that is, the power purchased and sold cannot exceed the upper limit of the transmission capacity of the interconnection line.

[0037] In step S2, that is Figure 2 Steps two, three, and four in the process, specifically step S2 includes: S21. Based on the set operating control parameters and the preset value range of the decision variables, a set of initial candidate solutions is generated. Each candidate solution represents a microgrid capacity configuration scheme. All candidate solutions together constitute an initial population of barrels, and each candidate solution is called a barrel individual. Specifically, initialize the Barrel Theory Optimization Algorithm (BTO): After constructing the mathematical model, the initial setup of the BTO algorithm begins. First, a series of hyperparameters controlling the algorithm's behavior need to be set, including the population size. For example, with 100 individuals, the maximum number of iterations for the algorithm. (e.g., generation 200), and the repair step size factor that will be used in subsequent update strategies. Elite guiding factors Disturbance intensity factor and strategy selection probability threshold And so on. Subsequently, the decision variables... Within the multidimensional solution space formed, a solution is generated through random sampling. There are 1 initial candidate solution vectors. Each solution vector... Each of these represents a specific microgrid capacity configuration scheme. The initial solutions together constitute the barrel population of the algorithm in generation 0. .

[0038] S22, evaluate the fitness of each individual barrel in the barrel population and calculate the value of each objective function corresponding to each individual barrel; Specifically, individual fitness assessment: This step serves as a bridge connecting outer-layer capacity optimization and inner-layer operational simulation. For the current population... Each individual wooden barrel The algorithm needs to accurately evaluate its performance. Specifically, this involves configuring the capacity parameters represented by the individual (…). The data (such as solar radiation, wind speed, etc.) are passed as input to the embedded operation scheduling simulation module. This module loads typical meteorological year data, including 8760 hourly points of solar radiation and wind speed data, as well as load year data. Then, based on a preset energy management strategy (EMS), it simulates the hourly operation of the microgrid configuration throughout the year. This EMS can be a rule-based heuristic strategy, such as prioritizing the use of renewable energy to meet the load, storing surplus electricity in energy storage, and selling electricity to the grid after the energy storage is fully charged; when renewable energy is insufficient, it is supplemented by discharging from the energy storage; when the energy storage capacity is insufficient, it purchases electricity from the main grid; when electricity purchase is limited or there is a grid failure, it starts a diesel generator. Through this year-long simulation, detailed operational data such as equipment output, energy storage status, electricity purchase and sale, fuel consumption, and whether power outages occur can be obtained at every moment. Finally, based on these simulation outputs and according to the objective function formula defined in step one, the individual "barrel" (referring to the microgrid's performance) can be accurately calculated. The three corresponding target values: , , and .

[0039] Step 3 embeds an operation scheduling simulation module, which uses a predefined energy management strategy to determine the optimal or suboptimal operation scheduling scheme of the microgrid system within a complete evaluation cycle under a given capacity configuration scheme. The energy management strategy can be a heuristic strategy based on expert rules, such as a load tracking strategy or a cyclic charging and discharging strategy, or it can be a mathematical optimization method, such as linear programming or dynamic programming. Its purpose is to provide accurate annual operating performance indicators for the outer capacity planning optimization, including but not limited to fuel consumption, electricity purchase and sale, power outages, etc., thereby ensuring the accuracy of the objective function evaluation.

[0040] S23, linearly normalize the function values ​​of all individual barrels in the barrel population on each optimization objective. For each individual barrel, the multiple objective function values ​​after normalization are regarded as the representation of the individual barrel's multidimensional performance. Among them, the objective dimension with the largest normalized value is identified as the performance bottleneck of the individual barrel.

[0041] Specifically, bottleneck identification: This step is the core innovation of the methodology of this invention, giving the algorithm the ability to identify the performance bottleneck of each candidate solution. After completing the fitness evaluation of all individuals in the population, the algorithm normalizes each objective dimension to eliminate the influence of differences in units and numerical ranges between different objective functions, thus making them comparable. First, it iterates through the current bottleneck population to find the maximum value of each of the three objectives. and minimum value ( Next, for each individual Each target value Apply the linear normalization formula Map it to the interval [0,1]. The normalized value. It intuitively reflects the individual In the target The degree of relative disadvantage in that objective: the closer the value is to 1, the worse the performance is relative to other members of the population; the closer the value is to 0, the better the performance. Finally, for the individual... By comparing its three normalized target values , , Find the largest value among the values. The target corresponding to this largest value is the individual identified as that individual. The most pressing performance weakness that needs improvement right now is its index. Depend on Confirmed. This process is crucial in transforming the abstract concept of the "barrel theory" into concrete, executable algorithmic operations.

[0042] Wherein, step S3 corresponds to Figure 2 Steps five and six; specifically: In step five, after identifying the weakest link, the algorithm enters the core stage of driving population evolution. For each individual in the population... The algorithm will select one of three specially designed, complementary update strategies based on a preset probability mechanism to update its position (i.e., the capacity configuration scheme it represents), thereby generating a new generation of individuals. .

[0043] Criteria for judging the execution of different dynamic optimization strategies: For each individual When performing an update, the algorithm first generates a random number rand in the interval [0, 1]. If rand < p1, then Strategy 1 is executed. If p1 <= rand < p2, then Strategy 2 is executed. If rand >= p2, then Strategy 3 is executed. p1 and p2 are preset probability thresholds, which jointly determine the usage frequencies of the three strategies, thereby balancing the exploration and optimization behaviors of the algorithm.

[0044] The specific descriptions of the three strategies are shown in the following table:

[0045] The execution of these three strategies is not based on the goodness or badness of individuals, but is invoked through a random selection mechanism. This mechanism enables the algorithm to focus on strengthening weaknesses (Strategy 1) and learning from the advanced (Strategy 2) most of the time to achieve efficient directional optimization and convergence; at the same time, it reserves a part of the opportunity for breakthroughs (Strategy 3) to prevent the algorithm from premature convergence and ensure the global search ability, and finally achieves a delicate dynamic balance between exploration and exploitation.

[0046] The first strategy is an adaptive repair strategy based on weakness identification. The goal of this strategy is to strengthen weaknesses, that is, to precisely improve the weakest performance dimension of an individual. When an individual is selected to execute this strategy, the algorithm first identifies the weakness target for it according to Step 4 . Then, the algorithm searches in the current entire population to find the individual that performs best (i.e., the target value is the smallest) on the target , denoted as . Then, the individual will move in the direction of to learn its excellent genes in this weakness dimension. The mathematical model for its position update is: . In this formula, is a random number in [0, 1] to increase the perturbation, is the repair step factor, which controls the intensity of learning. The intuitive meaning of this operation is: if the weakness of a solution is poor reliability (high LOLP), then it should imitate the equipment allocation of the solution with the best reliability (lowest LOLP) in the current population, for example, it may increase its energy storage or diesel engine capacity. This directional learning mechanism makes the optimization process highly targeted and avoids the blind mutation of traditional algorithms.

[0047] The second strategy is the elite-guided asymptotic convergence strategy. This strategy aims to leverage the high-quality solutions discovered by the algorithm during the search process to guide the evolutionary direction of the entire population, thereby accelerating convergence towards the global Pareto optimal front. At the beginning of each iteration, the algorithm first performs non-dominated sorting and crowding calculation on the current population, selecting individuals located at the Pareto first front (Rank 1) to form an elite archive. When a non-elite individual is selected to implement this strategy, the algorithm selects an elite individual from the elite archive through a method such as roulette or random selection. As a role model for its learning. The mathematical model for its position update is: .in, It is a random number between [0,1]. It is an elite guiding factor. This strategy ensures that the overall evolutionary trend of the population moves towards regions with better overall performance, demonstrating the algorithm's ability to utilize resources.

[0048] The third strategy is a global equilibrium perturbation adjustment strategy. The main purpose of this strategy is to maintain population diversity and prevent the algorithm from getting trapped in local optima due to excessive convergence, thereby ensuring its global exploration capability. This invention preferably uses the Lévy flight mechanism to achieve this goal. Lévy flight is a special random walk pattern whose step size follows a heavy-tailed (or long-tailed) probability distribution. This means that most of its step sizes are small, which helps in fine-grained searching near the current position, but at the same time, it produces very large step sizes with a non-negligible probability, allowing the individual to jump to a completely different, potentially better, distant region in the solution space. When an individual is chosen to execute this strategy, its position update mathematical model is: .in, It is a random step-size vector generated using the Lévy distribution. It is a scaling factor that controls the amplitude of the perturbation. This strategy simulates the impact of unexpected external shocks on the barrel system, injecting continuous innovative vitality into the algorithm.

[0049] These three strategies are not executed in isolation, but are dynamically scheduled through a probability switching mechanism. Specifically, a random number in the range [0,1] is generated for each individual. And set two thresholds. and .like If so, an adaptive repair strategy is executed; if If so, then implement an elite-guided strategy; if If so, a global perturbation strategy will be executed. This can be achieved through proper configuration. and The value (e.g.) This allows the algorithm to focus on addressing its shortcomings and learning from advanced technologies most of the time, while reserving some opportunities for breakthroughs, thus achieving a delicate dynamic balance between exploration and utilization.

[0050] Step Six: Updates and Iterations After all individuals in the population have undergone an update in step five, a completely new offspring population is formed. Before replacing the parent with offspring, a boundary check needs to be performed on the newly generated individuals. That is, check each individual. The decision variable values ​​are checked to see if they exceed the preset upper and lower limits. For variables that exceed the limits, various processing methods can be used, such as setting them to the nearest boundary value or randomly generating a new value in that dimension. After completing the boundary processing, the new generation population... The algorithm is now formally established. Subsequently, it checks the current iteration count. Has the preset maximum number of iterations been reached? If not, then the iteration counter If the algorithm returns to step three, it begins a new round of evaluation, identification, and updating. If so, the algorithm is considered to have converged sufficiently, and the iterative process terminates.

[0051] Step 7: Output the final planning scheme: After the algorithm loop terminates, the last generation of the population will be... This serves as the final candidate solution set. A final non-dominated sort is performed on this solution set to extract all non-dominated solutions. These solutions collectively constitute the Pareto optimal solution set obtained by the method of this invention. This solution set will be output in the form of data or graphs, such as Pareto front surfaces. It does not provide a single best answer, but rather presents decision-makers with a series of optimal combinations of solutions that represent different trade-offs between economy, reliability, and low carbon emissions, and are theoretically incomparable in terms of absolute superiority or inferiority. Decision-makers can select the solution that best suits their preferences from this solution set, based on the specific policy requirements of the project location, budget constraints, user sensitivity to power quality, and the degree of emphasis on environmental protection, as the final microgrid construction blueprint.

[0052] Reference Figure 1The entire process begins with step one, which involves constructing a precise mathematical model based on the specific circumstances of the microgrid project to be planned. This includes defining decision variables, three core optimization objective functions (LCC, LOLP, LCE), and a series of operational constraints. The process then proceeds to step two, where the BTO algorithm is initialized, including setting control parameters such as population size and maximum number of iterations, and randomly generating an initial "weakest link" population. Next, the algorithm enters its core iterative loop. In each loop, step three is executed first, evaluating the fitness of each individual in the population. This involves calling the inner simulation module to calculate the LCC, LOLP, and LCE values ​​corresponding to each configuration scheme. Following this is the key innovation of this invention, step four, which normalizes the objective values ​​of all individuals and identifies the performance bottleneck for each individual. In step five, the algorithm selects and executes one of three core update strategies for each individual based on a probability switching mechanism: step five a (adaptive repair), step five b (elite guidance), or step five c (global perturbation). After all individuals have been updated, step six involves boundary processing of the newly generated individuals, forming a new generation of the population. Step six is ​​a judgment step, checking whether the termination conditions, such as the maximum number of iterations, have been met. If not, the process returns to step three and begins a new round of iterations; if the conditions are met, the loop terminates, and the process proceeds to step seven, where the final population is sorted non-dominated and the final Pareto optimal solution set is output.

[0053] The implementation process of the present invention will be further illustrated through a specific embodiment. The application scenario of this embodiment is to plan a grid-connected microgrid system for a newly built industrial park far from the main urban area. The system needs to provide a highly reliable, low-cost, and green power supply for important loads such as office buildings and production lines within the park.

[0054] Reference Figure 3 The figure illustrates the microgrid system topology 200 used in this embodiment. The system mainly consists of the following components: a large-scale photovoltaic array 201 for converting solar energy into electrical energy; several wind turbines 202 for capturing wind energy; an energy storage system 203 containing a large-capacity lithium-ion battery pack and its bidirectional converter (PCS); a diesel generator 204 as an emergency backup power source; various electrical loads within the system 205; and a tie line connecting to the external power grid 207 via a point of common coupling (PCC) 206. The energy scheduling and optimized operation of the entire system are executed by a central controller 208.

[0055] Step 1: Model Building In this embodiment, the first step is to collect and organize all necessary basic data. This includes: typical meteorological data for the park's location, i.e., total solar radiation intensity, ambient temperature, and wind speed data for a full year with a time resolution of 1 hour; the park's expected hourly electricity load curve; electricity price information interacting with the external power grid, including time-of-use pricing and grid connection pricing; and detailed technical and economic parameters for various equipment (photovoltaic panels, wind turbines, energy storage batteries, PCS, diesel engines), such as unit investment cost, annual operation and maintenance cost coefficient, rated life, charge and discharge efficiency, energy density, and unit carbon emission factors at each stage of manufacturing, transportation, and disposal.

[0056] Based on this data, the decision variables and their value ranges are defined in this embodiment. The decision variable vector is... To ensure a reasonable search space, we set the following boundary: photovoltaic capacity. Values ​​are taken within the range [100, 1500] kWp; wind power capacity Values ​​are taken within the range [100, 1000] kW; energy storage battery capacity. Values ​​are taken within the range of [200, 3000] kWh; power of the energy storage converter. Values ​​are taken within the range of [100, 1500] kW; diesel engine power The values ​​are taken within the range of [0, 800] kW. To simplify the problem, these variables can be treated as discrete variables that change in steps of a certain size (e.g., 50kW or 100kWh).

[0057] The objective function is calculated according to the defined formula. The project lifecycle is set at 25 years, and the annual discount rate is 6%.

[0058] Step 2: BTO Algorithm Initialization We set the control parameters for the BTO algorithm as follows: Population size Maximum number of iterations The core strategy's control parameters are set as follows: Repair step size factor. Elite guiding factor Disturbance intensity factor Strategy selection threshold and Characteristic index of Levy flight distribution Subsequently, within the five-dimensional decision space defined above, the algorithm randomly generates 150 initial capacity configuration schemes, forming the initial barrel population.

[0059] Steps 3 to 7: Detailed Explanation of the Algorithm Iterative Optimization Process After the algorithm starts, it enters an iterative loop of up to 300 iterations. In any iteration, such as the t-th iteration, its internal operation is detailed as follows: The first step is fitness assessment. For the current population of 150 individuals, the algorithm processes each one individually. Taking one of these individuals... For example, the capacity configuration parameters it represents are fed into the operation simulation module. Based on the aforementioned energy management strategy, this module performs hourly simulations of the microgrid configuration for 8760 hours. After the simulation, the module outputs detailed breakdowns of total investment, operation and maintenance, replacement, fuel, and other costs for the scheme over a full year, total power outage hours, and key data such as total carbon emissions from diesel engines and purchased electricity. Based on this data, the algorithm can accurately calculate individual... The three objective function values .

[0060] Next comes the weakness identification. After the target values ​​for all 150 individuals have been calculated, the algorithm begins weakness identification. (Refer to...) Figure 4 This diagram visually illustrates the process. In the diagram, barrel 301 represents an individual solution, and the three planks 302, 303, and 304 represent the three objectives: LCC, LOLP, and LCE, respectively. Suppose that in a certain iteration, for the individual... The calculated initial target value is [LCC = 8.5 million USD, LOLP = 1.5%, LCE = 12,000 tons]. Simultaneously, after traversing the population, the algorithm finds that the current LCC ranges from [7.0 to 12.0] million USD, the LOLP ranges from [0.5% to 4.5%], and the LCE ranges from [8,000 to 20,000] tons. Therefore, for... Normalize the target value:

[0061]

[0062]

[0063] Comparing these three normalized values ​​[0.30, 0.25, 0.33], we find... The maximum. Therefore, the algorithm determines the individual. The current performance bottleneck is its life-cycle carbon emissions (LCE). This means that, compared to other options in the population, this option's environmental performance is its weakest link. The water level of 305 in the tank symbolizes the overall performance of the option, limited by its weakest link.

[0064] Then it enters the core update phase. The algorithm is for individuals. Generate a random number .

[0065] Scenario 1: If generated (less than) ), then for An adaptive repair strategy is implemented. Since its weakness is LCE (Limited Energy Constipation), the algorithm searches for the individual with the lowest LCE value in the current population; we call this the individual with the lowest LCE value. Then, the individual The location will be towards Update the direction: The practical significance of this operation is that the plan... They will learn from the configuration features of the most environmentally friendly solution, such as potentially significantly increasing the proportion of solar or wind power capacity while reducing reliance on diesel generators.

[0066] Scenario 2: If generated (between) and (between), then for The algorithm employs an elite-driven strategy. It first determines the current elite solution set (Pareto front) through non-dominated sorting. Then, it randomly selects an individual from this elite solution set. .individual The position will move closer to this elite solution with superior overall performance: .

[0067] Scenario 3: If generated (greater than) ), then for A global perturbation strategy is implemented. The algorithm will update its position using the Lévy flight formula: This update may cause... A large, non-linear change occurs in one or more decision variables, such as the energy storage capacity suddenly jumping from a low level to a very high level, giving the algorithm the opportunity to explore a completely new configuration region with potential advantages.

[0068] This process is repeated until all 300 iterations are completed.

[0069] Results analysis and verification refer to Figure 6The figure illustrates the convergence curves of the average values ​​of the three objective functions in the population during the operation of this embodiment. It can be seen that the LCC average curve (501), LOLP average curve (502), and LCE average curve (503) all exhibit a rapid downward trend in the early stages of iteration, indicating that the algorithm is quickly eliminating inferior solutions and searching for better regions. As the iteration progresses, the rate of descent of the curves gradually slows down and tends to stabilize in the later stages of iteration, signifying that the algorithm has found a relatively stable Pareto optimal region, i.e., the algorithm has converged. It is worth noting that the descent process of the three curves is relatively coordinated; there is no phenomenon where one objective stagnates while others continue to optimize. This, to some extent, reflects the BTO algorithm's ability to optimize various objectives in a balanced manner.

[0070] Reference Figure 5 The figure illustrates the final Pareto front 401 obtained by the BTO algorithm of this invention, and compares it with the results of two classic algorithms, NSGA-II (front 402) and MOPSO (front 403), running under the same conditions. For visualization purposes, the figure shows the two-dimensional projection of the two target dimensions, LCC and LCE. It is clearly observed from the figure that the Pareto front 401 obtained by the BTO algorithm of this invention is generally located to the lower left of the fronts obtained by the other two algorithms. This means that for any given carbon emission level, BTO can always find a lower-cost solution; or for any given cost budget, BTO can always find a solution with lower carbon emissions. This strongly demonstrates the significant advantage of the method of this invention in the quality (dominance) of the solution. Furthermore, the solution points obtained by BTO are more evenly and broadly distributed across the entire front, covering the entire trade-off spectrum from economic priority to environmental priority, while the solutions of the comparative algorithms may be sparse in some regions or not even touched upon.

[0071] Ultimately, the Pareto optimal solution set output by the algorithm may contain dozens of solutions. Decision-makers can select representative solutions for analysis, for example: Option A (Economy-Oriented): LCC = US$7.2 million, LOLP = 3.8%, LCE = 18,000 tons. This option is characterized by relatively conservative investment in renewable energy and energy storage, relying more on interaction with the main power grid and starting diesel generators during peak hours to reduce initial investment.

[0072] Option B (Balanced Development): LCC = US$9.5 million, LOLP = 0.9%, LCE = 10,000 tons. This option increases the penetration rate of renewable energy and the capacity of energy storage, significantly improves reliability and reduces carbon emissions, but the cost is increased.

[0073] Option C (Green Priority): LCC = US$12.8 million, LOLP = 0.6%, LCE = 6,500 tons. This option maximizes the installed capacity of renewable energy and features a very large-scale energy storage system, almost completely replacing the function of diesel engines and achieving extremely low operating carbon emissions, but it also has the highest initial investment. Park managers can make the most informed choice among these options based on their own financial situation, requirements for production continuity, and expectations of future carbon tax policies.

[0074] In summary, this invention combines profound systems theory (the barrel theory) with intelligent optimization algorithm design to create a novel, efficient, and robust multi-objective collaborative planning method for microgrids. Its technical solution is clear, highly feasible, and offers significant advantages over existing technologies. This invention not only provides a powerful tool for solving the specific engineering problem of microgrid planning but also offers a valuable new approach for solving complex, multi-objective black-box optimization problems in other fields.

[0075] The technical solution of this invention constructs a multi-objective collaborative planning model for microgrids. This model includes a set of decision variables to be optimized, a multi-objective function, and constraints to define the feasible operating boundary of the microgrid. Based on the barrel theory optimization algorithm, the multi-objective collaborative planning model is solved. The performance of each individual "barrel" in the barrel theory optimization algorithm is normalized according to multi-objectives, and performance bottlenecks are identified. The iterative update mechanism of the barrel theory optimization algorithm is executed, driving the evolution and updating of the barrel population through dynamic optimization of the identified performance bottlenecks. After satisfying the preset algorithm termination conditions, the obtained Pareto optimal solution set is output. This Pareto optimal solution set is used as the microgrid capacity configuration planning scheme, effectively solving the problem of low reliability in multi-objective collaborative planning of microgrids caused by existing technologies, and effectively improving the reliability of multi-objective collaborative planning of microgrids.

[0076] The technical solution of this invention employs a bottleneck identification and adaptive repair mechanism to keenly identify the performance bottlenecks of each candidate solution in the multi-dimensional target space and perform targeted, asymmetric optimization. This intelligent evolutionary approach, which addresses specific shortcomings, effectively avoids the phenomenon of traditional algorithms potentially exploiting strengths and mitigating weaknesses under random perturbations. This makes it easier to guide the population towards a region where performance is balanced across all target dimensions. Therefore, the solutions in the Pareto optimal solution set obtained by this invention typically possess higher overall performance, meaning they do not exhibit extreme performance deficiencies in any particular dimension, resulting in greater practicality and acceptability.

[0077] This invention constructs an efficient exploration-exploitation collaborative framework by organically combining and dynamically switching three strategies: adaptive repair, elite guidance, and global perturbation. The elite guidance strategy ensures the algorithm can quickly converge to the discovered high-quality region (Pareto front), demonstrating strong exploitation capabilities. The adaptive repair strategy, within a local region, deeply mines and finely adjusts the weak dimensions of the solution, representing a more purposeful exploration. The global perturbation strategy, based on Lévy flight, provides the algorithm with the ability to escape local optima and discover entirely new promising regions. The synergistic effect of these three strategies enables the algorithm to navigate efficiently in a complex, multi-modal solution space. Compared to traditional algorithms, it can converge to a Pareto front with superior dominance and distribution breadth with fewer iterations. Furthermore, through the synergistic effect of these three mechanisms, the algorithm's sensitivity to individual hyperparameters is relatively reduced, resulting in more stable and robust overall performance.

[0078] The technical solution of this invention cleverly transforms the abstract mathematical optimization problem into an intuitive and easy-to-understand physical process of repairing the weakest link in the system. This design makes it naturally and well-suited for typical high-dimensional, highly nonlinear, multi-constrained black-box optimization problems such as microgrid planning, where objective function evaluation requires time-consuming simulation.

[0079] The technical solution of this invention can track and record the dynamic evolution history of the shortcomings of each solution during the optimization process. By analyzing which objective (cost, reliability, or carbon emissions) more frequently becomes the shortcoming of most solutions in the population at different optimization stages, decision-makers can gain profound insights into the inherent contradictions of this specific planning problem. For example, if cost (LCC) generally becomes the shortcoming in the later stages of optimization, it indicates that after meeting certain reliability and environmental protection requirements, further cost reduction becomes extremely difficult, becoming the main bottleneck for improving system performance. This process-oriented diagnostic information transcends the limitations of traditional algorithms that only provide the final result, and can provide planners with valuable auxiliary decision-making basis for understanding the performance trade-off bottlenecks under different capacity configurations.

[0080] Example 2 like Figure 7 As shown, the present invention also provides a microgrid multi-objective cooperative planning system with carbon emission constraints, comprising: Module 101 is used to construct a multi-objective collaborative planning model for a microgrid. The multi-objective collaborative planning model for a microgrid includes a set of decision variables to be optimized, a multi-objective function, and constraints used to define the feasible operating boundary of the microgrid. The solution module 102 solves the microgrid multi-objective collaborative planning model based on the barrel theory optimization algorithm, and performs multi-objective performance normalization on the individual barrels in the barrel theory optimization algorithm and identifies the performance bottlenecks. Execution module 103 executes the iterative update mechanism of the barrel theory optimization algorithm, which drives the evolution and update of the barrel population by dynamically optimizing the shortcomings in recognition performance. After the preset algorithm termination condition is met, the output module 104 outputs the obtained Pareto optimal solution set and uses the Pareto optimal solution set as the microgrid capacity configuration planning scheme.

[0081] It should be noted that the implementation process of the construction module 101, the solution module 102, the execution module 103, and the output module 104 in this embodiment corresponds to the method steps in Embodiment 1, and will not be repeated here.

[0082] The technical solution of this invention constructs a multi-objective collaborative planning model for microgrids. This model includes a set of decision variables to be optimized, a multi-objective function, and constraints to define the feasible operating boundary of the microgrid. Based on the barrel theory optimization algorithm, the multi-objective collaborative planning model is solved. The performance of each individual "barrel" in the barrel theory optimization algorithm is normalized according to multi-objectives, and performance bottlenecks are identified. The iterative update mechanism of the barrel theory optimization algorithm is executed, driving the evolution and updating of the barrel population through dynamic optimization of the identified performance bottlenecks. After satisfying the preset algorithm termination conditions, the obtained Pareto optimal solution set is output. This Pareto optimal solution set is used as the microgrid capacity configuration planning scheme, effectively solving the problem of low reliability in multi-objective collaborative planning of microgrids caused by existing technologies, and effectively improving the reliability of multi-objective collaborative planning of microgrids.

[0083] The technical solution of this invention employs a bottleneck identification and adaptive repair mechanism to keenly identify the performance bottlenecks of each candidate solution in the multi-dimensional target space and perform targeted, asymmetric optimization. This intelligent evolutionary approach, which addresses specific shortcomings, effectively avoids the phenomenon of traditional algorithms potentially exploiting strengths and mitigating weaknesses under random perturbations. This makes it easier to guide the population towards a region where performance is balanced across all target dimensions. Therefore, the solutions in the Pareto optimal solution set obtained by this invention typically possess higher overall performance, meaning they do not exhibit extreme performance deficiencies in any particular dimension, resulting in greater practicality and acceptability.

[0084] This invention constructs an efficient exploration-exploitation collaborative framework by organically combining and dynamically switching three strategies: adaptive repair, elite guidance, and global perturbation. The elite guidance strategy ensures the algorithm can quickly converge to the discovered high-quality region (Pareto front), demonstrating strong exploitation capabilities. The adaptive repair strategy, within a local region, deeply mines and finely adjusts the weak dimensions of the solution, representing a more purposeful exploration. The global perturbation strategy, based on Lévy flight, provides the algorithm with the ability to escape local optima and discover entirely new promising regions. The synergistic effect of these three strategies enables the algorithm to navigate efficiently in a complex, multi-modal solution space. Compared to traditional algorithms, it can converge to a Pareto front with superior dominance and distribution breadth with fewer iterations. Furthermore, through the synergistic effect of these three mechanisms, the algorithm's sensitivity to individual hyperparameters is relatively reduced, resulting in more stable and robust overall performance.

[0085] The technical solution of this invention cleverly transforms the abstract mathematical optimization problem into an intuitive and easy-to-understand physical process of repairing the weakest link in the system. This design makes it naturally and well-suited for typical high-dimensional, highly nonlinear, multi-constrained black-box optimization problems such as microgrid planning, where objective function evaluation requires time-consuming simulation.

[0086] The technical solution of this invention can track and record the dynamic evolution history of the shortcomings of each solution during the optimization process. By analyzing which objective (cost, reliability, or carbon emissions) more frequently becomes the shortcoming of most solutions in the population at different optimization stages, decision-makers can gain profound insights into the inherent contradictions of this specific planning problem. For example, if cost (LCC) generally becomes the shortcoming in the later stages of optimization, it indicates that after meeting certain reliability and environmental protection requirements, further cost reduction becomes extremely difficult, becoming the main bottleneck for improving system performance. This process-oriented diagnostic information transcends the limitations of traditional algorithms that only provide the final result, and can provide planners with valuable auxiliary decision-making basis for understanding the performance trade-off bottlenecks under different capacity configurations.

[0087] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A multi-objective collaborative planning method for microgrids with carbon emission constraints, characterized in that, include: A multi-objective collaborative planning model for microgrids is constructed, which includes a set of decision variables to be optimized, a multi-objective function, and constraints used to define the feasible operating boundary of the microgrid. Based on the barrel theory optimization algorithm, the multi-objective collaborative planning model of microgrid is solved. The multi-objective performance of each individual barrel in the barrel theory optimization algorithm is normalized and the performance bottleneck is identified. The iterative update mechanism of the barrel theory optimization algorithm is implemented to drive the evolution and renewal of the barrel population by dynamically optimizing the shortcomings in recognition performance; After the preset algorithm termination condition is met, the Pareto optimal solution set is output, and the Pareto optimal solution set is used as the microgrid capacity configuration planning scheme.

2. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 1, characterized in that, The multi-objective function includes at least the following objectives: minimizing the total lifecycle cost, minimizing the probability of power outages in the microgrid system to maximize power supply reliability, and minimizing total lifecycle carbon emissions to reduce environmental impact; wherein, the total lifecycle cost objective... The calculation formula is: ； in, This refers to the initial investment cost; The annual operation and maintenance costs are calculated using the discount rate over the entire lifecycle. The cumulative value after each year is converted back to the initial year; The cost of equipment replacement is the cost incurred in future replacement years discounted to present value. To account for fossil fuel costs, the fuel costs consumed by the diesel generator during the simulation will be discounted and accumulated over its entire life cycle.

3. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 1, characterized in that, Based on the barrel theory optimization algorithm, a multi-objective collaborative planning model for microgrids is solved. The performance of each individual "barrel" in the barrel theory optimization algorithm is normalized according to multi-objectives, and performance bottlenecks are identified. Specifically, this includes: Based on the set operating control parameters and the preset value range of the decision variables, a set of initial candidate solutions is generated. Each candidate solution represents a microgrid capacity configuration scheme. All candidate solutions together constitute an initial population of weakest links, and each candidate solution is called a weakest link. The fitness of each individual barrel in the barrel population is evaluated, and the value of each objective function corresponding to each individual barrel is calculated. The function values ​​of all individual barrels in the barrel population on each optimization objective are linearly normalized. For each individual barrel, the multiple objective function values ​​after normalization are regarded as the representation of the individual barrel's multidimensional performance. Among them, the objective dimension with the largest normalized value is identified as the performance bottleneck of the individual barrel.

4. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 3, characterized in that, The function values ​​of all individual barrels in the barrel population on each optimization objective are linearly normalized. For each individual barrel, the normalized objective function values ​​are considered as a representation of the individual barrel's multidimensional performance. Specifically: For the Individual wooden barrels The corresponding original objective function value is Traverse the current barrel population to determine each target maximum value and minimum value ; Applying the linear normalization formula to individual barrels Each target value is normalized; Individual wooden barrels Shortcomings The corresponding target index Through function The operation determines this.

5. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 1, characterized in that, The iterative update mechanism of the barrel theory optimization algorithm, which dynamically optimizes the shortcomings in recognition performance, drives the evolution and update of the barrel population. Specifically: The iterative update mechanism of the barrel theory optimization algorithm is implemented by parallelly executing different dynamic strategies to dynamically optimize the identified performance bottlenecks, thereby driving population evolution. These dynamic strategies include: an adaptive repair strategy based on bottleneck identification, an elite-guided asymptotic convergence strategy, and a local equilibrium perturbation adjustment strategy. The adaptive repair strategy, based on bottleneck identification, guides each individual in the barrel to learn from the individual with the best performance in that bottleneck dimension, thereby achieving targeted enhancement of the weakest performance link. The elite-guided asymptotic convergence strategy uses the Pareto optimal solution in the current population as an elite guide to accelerate the convergence process of the entire population towards the global optimum. The global equilibrium perturbation adjustment strategy applies random perturbations with long-tailed distribution characteristics to some individuals to increase the diversity of the barrel population and help the algorithm avoid local optima traps.

6. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 5, characterized in that, The mathematical model for position updating in the adaptive repair strategy based on short-board identification is defined as follows: ； in, For the first The individual wooden barrels were in the first... The new position vector at the next iteration; The current position vector; This is the preset repair step size factor; It is a random number that is uniformly distributed in the interval [0,1]. Then it means in the first In the next iteration of the population, in the individual Shortcomings The position vector of the individual that performs best; The mathematical model for position update in the elite-guided progressive convergence strategy is defined as: ； in, The current position vector; Pre-set elite guidance factors; It is a random number that is uniformly distributed in the interval [0,1]. The position vector of an elite individual selected from the Pareto optimal solution set of the current iteration period through methods such as roulette wheel selection or random selection. The mathematical model for position update in the global balance disturbance regulation strategy is: ； in, A scaling factor to control the intensity of the disturbance; It is a random step-size vector generated by the Lévy distribution; This is the characteristic index of the Lévy distribution.

7. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 5, characterized in that, The specific switching mechanism for the execution allocation of the dynamic strategy is: Generate a random number in the interval [0,1] for each individual barrel. And based on two preset incremental probability thresholds and ; when At that time, an adaptive repair strategy based on short plank identification is applied to the individual barrel; when When, an elitist-guided asymptotic convergence strategy is applied to the individual barrel; when When this happens, a global balancing perturbation regulation strategy is applied to the individual.

8. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 1, characterized in that, The decision variables at least include: the total installed capacity of the photovoltaic array, the total installed capacity of the wind turbine, the rated energy capacity of the energy storage system, the rated power of the converter supporting the energy storage system, and the rated power of the diesel generator as a backup power source.

9. The multi-objective collaborative planning method for microgrids with carbon emission constraints according to claim 1, characterized in that, The constraint conditions include: system power balance constraint, energy storage system operation constraint, constraint that the charge and discharge power shall not exceed the rated power of the converter, output constraint of each distributed power source, and interactive power constraint with the large power grid.

10. A multi-objective collaborative planning system for microgrids with carbon emission constraints, characterized in that, Including: A construction module that constructs a multi-objective collaborative planning model for the microgrid. The multi-objective collaborative planning model for the microgrid includes a set of decision variables to be optimized, a multi-objective function, and constraint conditions for defining the feasible operation boundary of the microgrid; A solution module that solves the multi-objective collaborative planning model for the microgrid according to the barrel theory optimization algorithm, performs normalization processing on the multi-objective performance of each barrel individual in the barrel theory optimization algorithm, and identifies the performance short board; An execution module that executes the iterative update mechanism of the barrel theory optimization algorithm, and drives the evolution and update of the barrel population by dynamically optimizing the identified performance short board; An output module that outputs the obtained Pareto optimal solution set after meeting the preset algorithm termination condition, and uses the Pareto optimal solution set as the microgrid capacity configuration planning scheme.

Images