Distributed power supply management and scheduling evaluation index selection method based on AI enabling

Abstract

The invention discloses an evaluation index selection method for distributed power supply management and scheduling based on AI enabling. The method comprises the following steps: constructing a multi-dimensional dynamic evaluation system; comprising an economic dimension, an environmental protection dimension and a stability dimension, and the economic dimension comprises an investment and operation cost index, a power purchase cost index, a network loss cost index and a carbon emission cost index; the environmental protection dimension comprises a clean energy electric quantity permeability index and a renewable energy power generation consumption rate index; the stability dimension comprises a voltage fluctuation ratio index, a common coupling point power fluctuation index, a user average power failure time index and a system average power failure frequency index; the weight of each evaluation index is assigned in a mode of combining an analytic hierarchy process and an entropy evaluation method, including standardization and classification of expert scoring and subjective deviation elimination through clustering analysis, objective quantification of index information entropy based on the entropy evaluation method, and finally dynamic fusion of subjective and objective weights is realized.

Description

Technical Field

[0001] This invention relates to the field of smart grid management and dispatching, and in particular to a method for selecting evaluation indicators for distributed power source management and dispatching based on AI. Background Technology

[0002] Against the backdrop of the current energy transition, the proportion of distributed generation in the power system continues to rise, posing new challenges to traditional power grid operation models. Traditional evaluation systems primarily focused on economic efficiency are no longer adequate to meet the operational needs of new power systems, and their limitations are becoming increasingly apparent. On the one hand, solely pursuing economic efficiency may affect the safe operation of the system, as evidenced by the frequent voltage fluctuations observed in actual operation. On the other hand, the rapid development of renewable energy has continuously increased the importance of environmental indicators, but existing systems often place them in a secondary position. Furthermore, with increased user participation, social indicators such as demand response also need to be considered. Therefore, constructing a comprehensive evaluation system encompassing economic, environmental, technological, and social dimensions has become a crucial technical support for promoting the transformation of the power system.

[0003] Constructing a scientific and comprehensive multi-dimensional evaluation system faces numerous technical challenges. In terms of indicator system design, existing research generally suffers from incomplete dimensional coverage and a lack of in-depth analysis of the interactions between indicators. Weight determination methods rely excessively on expert experience, leading to subjective biases in evaluation results. The problem of insufficient dynamic adaptability is particularly prominent; existing methods struggle to respond promptly to real-time fluctuations in distributed power sources. To address these challenges, next-generation evaluation technologies are making breakthroughs in several directions: data-driven indicator selection methods can improve the scientific rigor of the indicator system; improved objective weighting algorithms can effectively reduce the influence of human factors; and dynamic evaluation frameworks integrating advanced computing technologies significantly enhance system response speed. These technological innovations lay the foundation for establishing a more scientific evaluation system.

[0004] In recent years, research on distributed power scheduling evaluation has shown a clear methodological evolution. Multi-objective optimization algorithms have been continuously improved, with increasing solution efficiency and accuracy; the application scope of objective weighting methods has gradually expanded, replacing traditional subjective weighting methods; and dynamic evaluation techniques have integrated various advanced algorithms, demonstrating good adaptability. However, some problems still urgently need to be addressed in this field: evaluation standards are not yet unified, and the indicator systems used in different studies vary significantly; the engineering application of complex algorithms faces computational efficiency challenges; and most methods remain in the theoretical verification stage. Solving these problems requires in-depth interdisciplinary collaboration and innovation.

[0005] Establishing a comprehensive multi-dimensional evaluation system has significant theoretical and practical implications. For power system operation, this system enables the synergistic optimization of multiple objectives, significantly improving overall system efficiency. In energy market development, it provides a scientific evaluation tool for distributed generation to participate in trading. For policymakers, this system allows for the objective assessment of the effectiveness of various policies. Notably, with the development of smart technologies, multi-dimensional evaluation is evolving towards digitalization and intelligence, which will profoundly change the management model of distributed generation and drive the power system towards greater efficiency and sustainability. This transformation will not only optimize grid operation but also provide crucial support for energy transition.

[0006] Direct scheduling and management of large-scale distributed power sources present significant challenges in terms of both economy and efficiency. Microgrids, as a novel type of micro-power system, avoid the direct interaction between distribution network dispatchers and a vast number of distributed power sources. This model allows for indirect management of distributed power sources through microgrids, reducing the impact of massive distributed power source access on grid operation and addressing challenges such as local consumption and grid-connected transmission of renewable energy. Traditional microgrid scheduling often focuses solely on minimizing operating costs, neglecting the environmental impact. The literature [Shi Chenlu. Research on Distributed Optimization Scheduling Strategies for Microgrid Clusters Considering Renewable Energy Volatility [D]. North China University of Technology, 2024.] proposes multi-objective scheduling for individual microgrids, minimizing both operating and environmental costs, achieving a win-win situation for both economic and environmental benefits. Regarding the objective function, previous optimization scheduling for microgrids has largely focused on single-objective planning. The literature [Ren Tiantian. Research on Multi-Objective Optimization Scheduling and Solution Algorithms for Grid-Connected Microgrids [D]. Xi'an University of Technology, 2024.] proposes setting the objective function to minimize both the economic cost of microgrid operation and the cost of environmental pollution control. With the continuous development of power systems, the demand for microgrid operation is also increasing. Scholars have begun to comprehensively consider multiple factors such as economy, reliability, and environment. Reducing the operation and maintenance costs of microgrids is the key to their promotion and development, and in my country's energy revolution process, energy conservation and emission reduction are regarded as crucial strategic goals. Constraining the models of each micro-power source can ensure the stability of the system during operation and ensure that the system operates in the optimal state, thereby reducing operating costs and improving energy utilization efficiency. The literature [Li Peng, Wu Difan, Li Yuwei, et al. Optimal scheduling strategy of multi-microgrid integrated energy system based on integrated demand response and master-slave game [J]. Proceedings of the CSEE, 2021.] proposes that the interest demand of multi-microgrid integrated energy system is to maximize profits, and the profit function is the difference between revenue and cost. The revenue from energy sales is equivalent to the user's energy purchase cost, and the operating cost consists of daily operating cost, daily equipment maintenance cost, and daily environmental penalty cost.

[0007] The literature [Gao Fangming. Research on Optimal Scheduling of Distributed Small Hydropower Microgrids [D]. Guangzhou University, 2024.] proposes to establish a scheduling model with the goal of minimizing the total operating cost of distributed small hydropower microgrids. The total operating cost includes the operation and maintenance costs of wind turbine generators, photovoltaic generators, distributed small hydropower generators, battery charging and discharging losses, the cost of purchasing electricity from the distribution network and the revenue from electricity sales, and the cost of flexible load interaction compensation. The literature [Zhu Wenqing. Research on the Dual-Layer Optimal Configuration of Distributed Power Sources and Energy Storage in Active Distribution Networks [D]. Nanchang University, 2024.] proposes a distributed power source optimization configuration model with the goal of minimizing the annual comprehensive cost of the distribution network. The annual comprehensive cost includes annual investment and operating costs, electricity purchase costs, network loss costs, and carbon emission costs. The literature [Lü Zhilin, Tan Ying, Li Jie, et al. Multi-objective capacity optimization configuration of distributed generation in independent hybrid microgrids based on Markov-ELM [J]. Proceedings of the CSEE, 2017.] proposes an optimal configuration model for distributed generation in independent microgrids, based on a given scheduling strategy and with system power supply reliability as a constraint, to minimize the system investment cost and environmental cost of the microgrid and maximize energy utilization. From the perspective of the interests of the power grid company, the distributed generation planning problem must consider the economics of distributed generation after grid connection. From the perspective of the active distribution system's ability to absorb distributed generation, it is also required that the capacity of distributed generation access be as large as possible. The literature [Liu Wenxia, ​​Xu Huiting. Optimal configuration of distributed generation considering voltage control cost [J]. Power System Technology, 2016.] proposes to establish an optimal configuration model for distributed generation access capacity under active voltage control, with the objectives of minimizing the annual comprehensive cost and maximizing the proportion of clean energy generation. The literature [Xiao Fang. Research on the Planning of Distributed Power Generation in Distribution Networks [D]. Shenyang Agricultural University, 2023.] argues that from an economic model perspective, it is necessary to reduce the cost of power dispatching while satisfying power flow constraints within the distribution network system, in order to maximize the economic benefits of the dispatching scheme. Therefore, to achieve the goal of maximizing the economic benefits of power dispatching, the optimization objectives are to maximize the benefits of power trading, reduce the supply of electricity, and minimize network losses. The literature [Feng Xianzheng. Research on Optimal Distributed Power Generation Dispatch Based on Optimal Power Flow Algorithm [D]. North China Electric Power University (Beijing), 2024.] evaluates the dispatching of distributed power generation from three aspects: operating cost, power system stability, and power quality, with the optimization objectives being the lowest operating cost per unit time period, the highest renewable energy consumption rate per unit time period, and the lowest voltage fluctuation per unit time period, respectively.The literature [Liu Junjia. Research on Multi-Objective Optimization Scheduling of Microgrids Based on Improved Dung Beetle Algorithm [D]. Anhui University of Science and Technology, 2024.] proposes to model the microgrid system from both economic and environmental perspectives, taking operating costs and environmental protection costs as objective functions. The specific operating costs include the generation costs of each distributed power source, the cost of equipment depreciation, and the cost of electricity purchased from the main grid. The environmental protection costs are the costs of treating pollutants that need to be treated when constructing any project.

[0008] The existing technology has the following relative shortcomings:

[0009] Traditional evaluation indicators are too simplistic and cannot fully reflect the overall benefits of scheduling.

[0010] Existing distributed power dispatch evaluation systems typically focus on a single dimension, such as minimizing economic costs or optimizing environmental performance as the optimization objective, while neglecting other key factors. For example, some studies only consider operation and maintenance costs, failing to incorporate indicators such as network losses, carbon emissions, or voltage stability. This can lead to dispatch schemes failing in practical applications due to grid fluctuations or environmental policy restrictions. Furthermore, optimization based on a single indicator is prone to getting trapped in local optima, making it impossible to achieve global balance in complex and dynamic power systems. For instance, pursuing only the lowest electricity purchase cost may result in a decrease in renewable energy integration rates or over-reliance on a particular type of power source (such as coal-fired power), increasing carbon emissions. This one-sidedness makes dispatch strategies ill-suited to the multi-objective collaborative optimization requirements of smart grids, especially in scenarios with large fluctuations in wind and solar power output and dynamic changes in load demand, where a single indicator system cannot provide a scientific basis for decision-making.

[0011] Weight allocation relies on subjective experience and lacks data-driven support.

[0012] Existing weighting methods (such as the Analytic Hierarchy Process, AHP) heavily rely on expert scoring, resulting in high subjectivity and low consistency. Experts, due to personal preferences or industry backgrounds, may assign different weights to the same indicator, leading to significant fluctuations in evaluation results. For example, environmental experts may overemphasize carbon emission indicators, while power grid maintenance personnel may focus more on voltage stability; this discrepancy can cause the final dispatch plan to deviate from actual needs. Furthermore, traditional methods cannot dynamically adjust weights, making it difficult to cope with real-time data changes. For instance, in extreme weather conditions, using preset stability weights may fail to respond quickly to voltage spikes or frequency instability. Although some studies have attempted to introduce objective weighting methods such as entropy methods, they have not yet effectively combined the advantages of both subjective and objective approaches, resulting in weighting allocation either relying too heavily on historical data (ignoring expert experience) or being entirely limited by subjective judgment (lacking data validation).

[0013] With the increasing demand for electricity, higher requirements are being placed on the power grid's distribution capacity. Distributed generation (DG) has received increasing attention and research due to its flexibility and environmental friendliness. However, determining the evaluation indicators for DG management and dispatch is a key issue in building a smart distribution network. To ensure the safe, stable, and economical operation of the power system, designing a comprehensive and scientific evaluation index for DG dispatch is particularly important. Currently, existing research suffers from problems such as overly simplistic evaluation indicators and highly subjective weight allocation. Therefore, this invention aims to address the following two issues:

[0014] 1) The problem of traditional single evaluation indicators: Existing distributed generation dispatch evaluation systems often focus only on single-dimensional indicators, such as simply aiming for the lowest economic cost or the minimum carbon emissions, lacking a systematic assessment of the comprehensive benefits of dispatch schemes. This one-sidedness leads to many contradictions in actual operation: the most economically optimal scheme may exacerbate grid fluctuations, while the most environmentally friendly scheme may significantly increase operating costs. More seriously, the existing indicator system fails to incorporate key elements under the new power system, such as the penetration rate of renewable energy, making the evaluation results unable to truly reflect the multidimensional value of distributed generation in promoting energy transition and improving power supply resilience. This single-dimensional evaluation model is no longer suitable for the increasingly complex decision-making needs of grid dispatch. Therefore, a comprehensive and scientific method for selecting evaluation indicators is needed to improve energy utilization, reduce costs, and enhance the overall economic benefits of the power system.

[0015] 2) The problem of strong subjectivity in weight allocation: Current evaluation methods rely excessively on expert experience when determining the weights of each indicator, lacking objective data support, leading to a serious disconnect between weight allocation and actual operational needs. For example, different experts may have significantly different priorities for economic efficiency and environmental protection, resulting in diametrically opposed conclusions for the same dispatching scheme under different evaluation systems. Furthermore, the fixed-weight model cannot adapt to the different characteristics of power grids in different regions: the weight requirements for economic efficiency and environmental protection should differ between densely industrialized areas and ecologically sensitive areas, but existing methods often use a uniform template allocation, causing evaluation results to deviate from actual benefits. This highly subjective and inflexible weight mechanism greatly weakens the scientific rigor and credibility of dispatching decisions. Therefore, a method that combines subjective and objective approaches to weight allocation is needed.

[0016] In view of the above-mentioned shortcomings of the existing technology, the present invention is proposed. Summary of the Invention

[0017] To address the aforementioned shortcomings in existing technologies, this invention aims to construct a multi-quantitative index system encompassing economic efficiency, environmental friendliness, and stability. Through multi-objective collaborative optimization, it ensures that the dispatching scheme reduces economic costs while simultaneously enhancing renewable energy absorption capacity and safeguarding grid security. For example, during peak photovoltaic power generation periods at midday, the system can automatically increase the weight of the absorption rate to reduce curtailment, while prioritizing voltage stability optimization during nighttime peak load periods. This dynamic balancing mechanism significantly improves the scientific rigor and adaptability of distributed power dispatching.

[0018] To achieve the objective of this invention, the technical solution provided by this invention is as follows:

[0019] A method for selecting evaluation metrics for AI-enabled distributed power management and scheduling.

[0020] The specific steps include:

[0021] Step S1: Construct a multi-dimensional dynamic evaluation system;

[0022] This includes economic, environmental protection, and stability dimensions, among which...

[0023] The economic dimension includes indicators of investment and operating costs, electricity purchase costs, network loss costs, and carbon emission costs, in order to quantify the contribution of distributed power dispatch to economic benefits.

[0024] The environmental protection dimension includes indicators such as clean energy power penetration rate and renewable energy power generation consumption rate, in order to assess the positive impact of dispatch schemes on environmental protection;

[0025] Stability dimensions include voltage fluctuation rate, power fluctuation at the point of common coupling, average outage time for users, and average outage frequency for the system, to ensure the stable operation of the power grid after the integration of distributed power sources.

[0026] Step S2 involves assigning weights to the various evaluation indicators from Step S1 using a combination of the analytic hierarchy process (AHP) and the entropy method. This includes standardizing and classifying expert scores through cluster analysis to eliminate subjective bias, and then objectively quantifying the information entropy of the indicators based on the entropy method. Ultimately, this achieves a dynamic fusion of subjective and objective weights. Specifically:

[0027] Subjective weights were determined using the analytic hierarchy process (AHP) based on cluster analysis.

[0028] First, cluster analysis is performed on the individual weight vectors of the experts;

[0029] Based on the characteristics of individual expert weight vectors, they are clustered into different classes and assigned different weight values;

[0030] Finally, the weight vectors of different experts are weighted and averaged to obtain the weight vector values ​​of each evaluation model.

[0031] The preferred technical solution provided by this invention is as follows:

[0032] In step S1,

[0033] In the aforementioned economic dimension, economic cost The function expression is:

[0034] ,

[0035] in, Annual investment and operating costs of microgrids; Indicates the cost of purchasing electricity. Indicates network loss cost, Indicates the cost of carbon emissions;

[0036] Annual investment and operating costs of microgrids The expression is:

[0037] ,

[0038] in, Indicates the number of nodes in the power distribution system; Represents a node Is DG installed at this location? Represents a node The capacity of the DG installation location; Indicates a fixed annual interest rate; Indicates the planning period; Represents a node Investment cost per unit capacity of DG; Represents a node Operating cost per unit capacity of DG;

[0039] Electricity purchase cost The expression is:

[0040] ,

[0041] in, Indicates the unit price of electricity purchased; Indicates the maximum annual power grid utilization hours; Indicates the first The load of each node;

[0042] Network loss cost The expression is:

[0043] ,

[0044] in, This indicates the unit electricity price for network losses; This indicates the maximum number of hours of power grid loss per year. Represents the set of all branches in the power distribution system; Indicates a branch The current amplitude; Indicates a branch The resistance value;

[0045] carbon emission costs The expression is:

[0046] ,

[0047] in, The environmental penalty is 45.61 yuan / ton, meaning a penalty of 45.61 yuan for every ton of CO2 produced. Represents the carbon emission factor, taking That is, 0.5703 kg of CO2 is emitted for every kilowatt-hour of electricity generated.

[0048] A further preferred technical solution provided by the present invention is as follows:

[0049] In step S1,

[0050] In the aforementioned environmental protection dimension, the clean energy electricity penetration rate The expression is:

[0051] ,

[0052] in, For the total scheduling time, The number of DGs connected to the power grid; Time period No. Each DG contributed its efforts; For time period node Active load; For time period of Active power loss of branch circuits;

[0053] Renewable energy power generation absorption rate It can be defined as:

[0054] ,

[0055] in, For nodes Maximum predicted output power of renewable energy For nodes Actual output power of renewable energy.

[0056] A further preferred technical solution provided by the present invention is as follows:

[0057] In step S1,

[0058] In the aforementioned stability dimensions,

[0059] Common point of coupling power fluctuation The expression is:

[0060] ,

[0061] in, express Electricity purchase volume during the time period; express Electricity sales volume during a given time period;

[0062] Voltage fluctuation The expression is:

[0063] ,

[0064] in, Represents a node The voltage on; Represents a node The rated voltage; , For nodes Maximum and minimum voltages;

[0065] System Average Interruption Frequency Index The expression for ) is:

[0066] ,

[0067] in, For nodes Average failure rate, For nodes The total number of users.

[0068] The expression for Customer Average Interruption Duration Index (CAIDI) is as follows:

[0069] ,

[0070] in, For nodes The average annual power outage time.

[0071] Another preferred technical solution provided by the present invention is as follows:

[0072] In step S2, the specific steps for assigning weights to each evaluation index using a combination of the analytic hierarchy process (AHP) and the entropy method include:

[0073] The number of samples in the individual weight vector is The 10 evaluation indicators mentioned in step S1 can be used to obtain the expert individual weight vector. , ,

[0074] The expert individual weight vector matrix, calculated from the judgment matrix obtained from the expert questionnaire, is as follows:

[0075] ,

[0076] First, the individual weight vectors of each expert are standardized, that is:

[0077] ,

[0078] ,

[0079]

[0080] After standardization, the standardized expert individual weight vector can be obtained. , ;

[0081] The standardized expert individual weight vectors are clustered using the hierarchical clustering method.

[0082] The class average method is used to calculate the distance between classes, and the Euclidean distance is used for the distance between samples.

[0083] Cluster analysis was performed on the samples using hierarchical clustering, and the samples were clustered into... kind;

[0084] The results of cluster analysis Individual expert weight vectors are used to calculate the number of samples in each class. ;

[0085] No. The individual weight vector of the nth expert belongs to the... Class, define the first The number of expert individual weight vectors contained in the class Total number of expert individual weight vectors The ratio is the expert individual weight vector The confidence factor, i.e.

[0086] ,

[0087] No. The weight coefficients of the individual expert weight vector Weighting of individual experts Confidence factor Proportional, that is

[0088] ,

[0089] because

[0090] ,

[0091] achievable

[0092] ,

[0093] ,

[0094] ,

[0095] By weighting each expert's individual weight vector with its corresponding weight value, we can obtain the weight vector of the evaluation model.

[0096] ,

[0097] Each sample is denoted as ,sample For indicators The evaluation value is recorded as ;

[0098] The evaluation values ​​of the samples are standardized, whereby...

[0099] The method for handling benefit-type indicators is as follows:

[0100] ,

[0101] The method for handling cost-related indicators is as follows:

[0102] ,

[0103] in, , , ;

[0104] According to the definition of entropy, the first The entropy value of each indicator is

[0105] ,

[0106] in, , ,

[0107] make

[0108] ,

[0109] Then the first The objective weights of each evaluation indicator are:

[0110]

[0111] in, .

[0112] Combining the analytic hierarchy process (AHP) and the entropy method, the comprehensive weights of each indicator are obtained as follows:

[0113] .

[0114] Compared with the prior art, the beneficial effects of the present invention include:

[0115] Compared with existing technical solutions, the present invention has the following advantages:

[0116] 1) This invention has significant comprehensive evaluation advantages compared to existing technologies. While existing technologies have attempted to construct evaluation systems from two or more dimensions, such as economy, environmental protection, or stability, they generally suffer from incomplete dimension coverage and a lack of systematic correlation between indicators. For example, they only consider three indicators: operating cost, absorption rate, and voltage fluctuation, without addressing key factors such as network losses and outage frequency; they focus only on economy and environmental protection, neglecting grid stability indicators. This invention, by constructing a multi-dimensional comprehensive evaluation indicator system covering economy, environmental protection, and stability, systematically integrates ten key indicators, including investment and operating costs, network loss costs, carbon emission costs, clean energy penetration rate, absorption rate, voltage fluctuation, common coupling point power fluctuation, system average outage frequency, and user average outage time. This achieves a comprehensive and multi-faceted evaluation of distributed power dispatching schemes, effectively avoiding evaluation biases caused by single or missing indicators.

[0117] 2) Regarding weight allocation, existing technologies largely rely on the Analytic Hierarchy Process (AHP) based on expert experience, which suffers from high subjectivity, low consistency, and a lack of responsiveness to real-time data. This invention innovatively proposes a weight allocation mechanism that integrates subjective and objective factors. First, it standardizes and optimizes expert opinions using an AHP method based on cluster analysis, eliminating individual subjective biases. Then, it introduces the entropy method to quantify the information entropy of each indicator based on historical and real-time operational data, dynamically adjusting the weights. This integrated approach retains the guidance of industry experience while enhancing the objectivity and dynamic adaptability of weight allocation, making the evaluation results more aligned with actual operational needs. Especially in complex scenarios such as extreme weather and sudden load changes, it can quickly adjust weight priorities, improving the scientific rigor and robustness of scheduling decisions.

[0118] Existing weight allocation methods have two major drawbacks: first, the AHP method, which relies on expert experience, is highly subjective, and differences in scores given by different experts for the same indicator may lead to scheduling schemes deviating from actual needs; second, the purely data-driven entropy method, while objective, ignores industry knowledge and is difficult to handle sudden scenarios (such as extreme weather). This invention innovatively combines AHP and the entropy method: first, cluster analysis is used to eliminate the subjective bias of expert scoring, and then the entropy method is used to quantify the information entropy of indicators based on historical data to dynamically adjust the weights. Specifically, the process begins with data preparation, collecting actual operational data for the key indicators over a specific period to form an initial data matrix. Then, standardization and entropy calculation are performed. The initial data matrix is ​​standardized to eliminate the influence of dimensions. Next, the weight of all data values ​​for each indicator is calculated, and based on this weight information, the entropy value for each indicator is calculated using the information entropy formula. This entropy value measures the dispersion of the corresponding indicator's data sequence. Next, entropy weights are determined. Based on the entropy values, the difference coefficients for each indicator are calculated and normalized to obtain objective weights driven entirely by data, referred to as 'entropy weights'. Finally, dynamic fusion is achieved by combining the objective 'entropy weights' obtained above with the subjective weights obtained through the AHP clustering method using a weighted average. The fusion coefficient can be adjusted according to the system's operational stage or specific scenarios. For example, during typhoon warnings, the system can automatically reduce the economic weight while simultaneously increasing the priority of stability indicators based on real-time meteorological data. This mechanism of subjective and objective coordination ensures that weight allocation aligns with industry experience and responds to actual data changes.

[0119] Terminology Explanation:

[0120] DG, Distributed Generation, distributed power generation

[0121] AHP, Analytic Hierarchy Process. Detailed Implementation

[0122] The technical solutions in the embodiments of this application will be clearly and completely described below.

[0123] The present invention provides a method for selecting evaluation metrics for distributed power management and scheduling based on AI, the specific steps of which include:

[0124] Step S1: Construct a multi-dimensional dynamic evaluation system;

[0125] This includes economic, environmental protection, and stability dimensions, among which...

[0126] The economic dimension includes indicators of investment and operating costs, electricity purchase costs, network loss costs, and carbon emission costs, in order to quantify the contribution of distributed power dispatch to economic benefits.

[0127] Regarding the economic dimension, this invention comprehensively considers investment and operating costs, network loss costs, electricity purchase costs, and carbon emission costs. Economic Costs The function expression is:

[0128]

[0129] in, Annual investment and operating costs of microgrids; Indicates the cost of purchasing electricity. Indicates network loss cost, This indicates the cost of carbon emissions.

[0130] In distributed generation dispatch, considering investment and operating costs is crucial, as it directly impacts the economics of power supply, optimal resource allocation, and system reliability. By comprehensively evaluating the cost-effectiveness of different power sources, the optimal dispatch scheme can be selected, avoiding resource waste and improving system performance. Simultaneously, cost considerations can promote the use of renewable energy, drive sustainable development in the power industry, and guide the market towards economical and environmentally friendly choices. Therefore, investment and operating costs are key factors that cannot be ignored in distributed generation dispatch. Annual investment and operating costs of microgrids. The expression can be defined as:

[0131]

[0132] in, Indicates the number of nodes in the power distribution system; Represents a node Is DG installed at this location? Represents a node The capacity of the DG installation location; Indicates a fixed annual interest rate; Indicates the planning period; Represents a node Investment cost per unit capacity of DG; Represents a node Operating cost per unit capacity of DG.

[0133] Electricity purchase cost is a significant component of power supply companies' costs, typically accounting for the majority, and directly impacts the economics and efficiency of power grid operation. By rationally considering electricity purchase cost, an economic balance can be achieved between distributed generation and grid-purchased electricity while ensuring power supply stability, thereby improving the overall efficiency and sustainability of the power system. Especially with the increasing proportion of distributed generation, scientifically scheduling distributed generation and effectively controlling electricity purchase cost are crucial for enhancing the economy and reliability of the distribution network. Electricity Purchase Cost The expression can be defined as:

[0134]

[0135] in, Indicates the unit price of electricity purchased; Indicates the maximum annual power grid utilization hours; Indicates the first The load of each node.

[0136] After distributed generation (DG) is connected to the distribution network, its power generation process may generate harmonics and unbalanced currents, causing losses to the distribution network. Furthermore, if the output power of the DG does not match the load of the distribution network, additional network losses will occur. Network losses not only increase energy waste during power transmission but also directly increase the operating costs of the power grid. Therefore, in DG dispatching, network loss costs must be fully considered, and the output power of DG must be adjusted in real time to reduce the impact and losses on the distribution network. This not only reduces network loss costs and improves the transmission capacity and stability of the power grid but also helps achieve efficient and stable grid operation and enhances the comprehensive management level of power companies. Network loss costs The expression can be defined as:

[0137]

[0138] in, This indicates the unit electricity price for network losses; This indicates the maximum number of hours of power grid loss per year. Represents the set of all branches in the power distribution system; Indicates a branch The current amplitude; Indicates a branch The resistance value.

[0139] Renewable energy sources such as solar and wind power have significant low-carbon emission advantages compared to traditional fossil fuel power generation. However, the technical characteristics, operating efficiency, and energy types of different distributed power sources all lead to variations in their carbon emissions. Fully considering carbon emission costs during dispatching can encourage dispatching decisions to favor low-carbon and environmentally friendly power sources, thereby effectively reducing the overall carbon emission level of the power system. This not only aligns with global goals of addressing climate change and achieving sustainable development but also helps enhance the environmental image of the power industry and strengthen its social responsibility. Furthermore, optimizing the dispatching of distributed power sources and reducing carbon emissions can also bring potential carbon trading revenue to enterprises, achieving a win-win situation for both economic and environmental benefits. Carbon emission costs The expression can be defined as:

[0140]

[0141] in, The environmental penalty is 45.61 yuan / ton, meaning a penalty of 45.61 yuan for every ton of CO2 produced. Represents the carbon emission factor, taking That is, 0.5703 kg of CO2 is emitted for every kilowatt-hour of electricity generated.

[0142] The environmental protection dimension includes indicators such as clean energy power penetration rate and renewable energy power generation consumption rate, in order to assess the positive impact of dispatch schemes on environmental protection;

[0143] In terms of environmental protection, this invention comprehensively considers the penetration rate of clean energy electricity and the consumption rate of renewable energy power generation.

[0144] Clean energy sources, such as solar, wind, hydro, and geothermal energy, are characterized by zero emissions, renewability, and abundant resources, making them ideal for energy transition and environmental sustainability. Increasing the penetration rate of clean energy can effectively reduce dependence on traditional fossil fuels, lower greenhouse gas emissions, and protect the ecological environment. Furthermore, distributed generation, installed near load-concentrated areas, can further reduce transmission losses and improve energy efficiency by using clean energy. Moreover, with the continuous advancement of clean energy technologies and cost reductions, the economic viability of clean energy generation is becoming increasingly apparent, contributing to improved overall efficiency of distributed generation dispatch. Therefore, fully considering the penetration rate of clean energy in distributed generation dispatch is a key measure to achieve the goals of sustainable power system development and environmental protection. Clean Energy Penetration Rate The expression can be defined as:

[0145]

[0146] in, For the total scheduling time, The number of DGs connected to the power grid; Time period No. Each DG contributed its efforts; For time period node Active load; For time period of Active power loss of branch circuits.

[0147] Renewable energy generation is intermittent, volatile, and uncertain. For example, the output of photovoltaic and wind power is significantly affected by weather and seasonal factors, leading to unstable power generation and posing challenges to grid dispatch. The absorption rate reflects the proportion of renewable energy integrated and effectively utilized in the power system, and is an important indicator of the power system's capacity to accommodate renewable energy. By focusing on the absorption rate, we can assess the actual absorption of distributed generation after grid connection, optimize power dispatch strategies, and ensure stable grid operation. At the same time, improving the renewable energy absorption rate is also an important way to promote energy structure transformation and achieve sustainable development. Therefore, considering the renewable energy generation absorption rate in distributed generation dispatch helps balance supply and demand, improve system flexibility, and promote the widespread use of clean energy. The renewable energy generation absorption rate... The expression can be defined as:

[0148]

[0149] in, For nodes Maximum predicted output power of renewable energy For nodes Actual output power of renewable energy.

[0150] Stability dimensions include voltage fluctuation rate, power fluctuation at the point of common coupling, average outage time for users, and average outage frequency for the system, to ensure the stable operation of the power grid after the integration of distributed power sources.

[0151] In terms of stability, this invention comprehensively considers voltage fluctuation rate, power fluctuation at the point of common coupling, average system outage frequency, and average user outage time.

[0152] Distributed power sources, especially intermittent energy sources such as wind and solar power, often have output power constrained by weather, environmental factors, and other uncertainties. When these distributed power sources are connected to the grid, their power fluctuations can affect the stability and reliability of the entire grid through the point of common coupling (PCC). Excessive PCC power fluctuations can lead to grid voltage and frequency instability, thereby affecting the power quality for other users and the safe operation of the grid. Therefore, in distributed power source dispatching, it is crucial to closely monitor and rationally control PCC power fluctuations to ensure the stable and reliable operation of the grid. The expression for PCC power fluctuations can be defined as:

[0153]

[0154] in, express Electricity purchase volume during the time period; express Electricity sales volume during a given time period.

[0155] When distributed generation (DG) sources are connected to the distribution network, their currents interact with the load currents, leading to voltage fluctuations in the distribution network. DG sources are significantly affected by climate and environment, and the uncertainty of their output can cause noticeable voltage fluctuations in the connected grid. Furthermore, factors such as the start-up and shutdown operations of DG sources, the power factor angle, and the impedance angle of the system's equivalent impedance also influence voltage fluctuations. Voltage fluctuations not only affect the stability and reliability of the power grid but can also lead to equipment damage and reduced production efficiency. Therefore, in distributed generation dispatching, monitoring and controlling voltage fluctuations can ensure the safe operation of the power grid and a high-quality power supply. The expression can be defined as:

[0156]

[0157] in, Represents a node The voltage on; Represents a node The rated voltage; , For nodes Maximum and minimum voltages.

[0158] System average outage frequency (SAIFI) represents the average number of outages per user powered by the system per unit time (usually one year), and is a key indicator for measuring the performance of a distribution system. Distributed power sources, as an important component of the modern energy system, directly impact users' electricity experience through effective dispatching. By considering SAIFI, dispatchers can assess the impact of different dispatching strategies on system outage frequency, thereby selecting strategies that reduce the number of outages and improve power supply reliability. This is crucial for ensuring the stability of power supply, meeting user needs, and improving the overall performance of the power grid. The expression for system average outage frequency can be defined as:

[0159]

[0160] in, For nodes Average failure rate, For nodes The total number of users.

[0161] Average User Outage Time (CAIDI) measures the average duration of a power outage experienced by each user supplied by the system over a given period, directly reflecting the continuity and reliability of power services. Distributed power sources, as a key resource for enhancing grid flexibility and resilience, require scheduling strategies optimized to closely align with improving user experience. By focusing on CAIDI, dispatchers can assess the potential outage risks to users under different scheduling schemes, and then take effective measures to shorten outage time and reduce the impact of outages on users' lives and production activities. This is not only related to user satisfaction but is also an essential requirement for improving power service quality and modernizing the power grid. The expression for Average User Outage Time can be defined as:

[0162]

[0163] in, For nodes The average annual power outage time.

[0164] Step S2 involves assigning weights to the various evaluation indicators from Step S1 using a combination of the analytic hierarchy process (AHP) and the entropy method. This includes standardizing and classifying expert scores through cluster analysis to eliminate subjective bias, and then objectively quantifying the information entropy of the indicators based on the entropy method. Ultimately, this achieves a dynamic fusion of subjective and objective weights. Specifically:

[0165] This invention uses a combination of the Analytic Hierarchy Process (AHP) and the entropy method to assign weights to each evaluation index.

[0166] Traditional analytic hierarchy process (AHP) averages the individual weights of each expert to obtain the final weight. However, since experts differ in their personal preferences, social experiences, and cultural backgrounds, averaging does not adequately reflect these characteristics. Therefore, this invention employs AHP based on cluster analysis to determine subjective weights. First, cluster analysis is performed on the individual expert weight vectors, grouping them into different classes based on their characteristics and assigning them different weight values. Finally, the weight vectors of different experts are weighted and averaged to obtain the weight vector value for each evaluation model. The calculation process is as follows.

[0167] Let the number of samples in the individual weight vector be... The 10 evaluation indicators mentioned in step S1 (four indicators in the economic dimension, two indicators in the environmental dimension, and four indicators in the stability dimension) are evaluated.

[0168] The individual expert weight vector can be obtained. , .

[0169] Let the expert individual weight vector matrix be calculated from the judgment matrix obtained from the expert questionnaire:

[0170]

[0171] To prevent irrationality caused by significant differences in the individual weight coefficients of different experts, the individual weight vectors of each expert are first standardized, that is:

[0172]

[0173]

[0174]

[0175] After standardization, the standardized expert individual weight vector can be obtained. , The standardized expert individual weight vectors are clustered using hierarchical clustering. This invention employs the class average method to calculate the distance between classes, and the distance between samples is calculated using Euclidean distance.

[0176] This invention uses hierarchical clustering to perform cluster analysis on samples, clustering the samples into... Class. The results of cluster analysis. Individual weight vectors for each class of experts; calculate the number of samples in each class. Let the first... The individual weight vector of the nth expert belongs to the... Class, define the first The number of expert individual weight vectors contained in the class Total number of expert individual weight vectors The ratio is the expert individual weight vector The confidence factor, i.e.

[0177]

[0178] for In a class of expert individual weight vectors, those within the same class share the same confidence factor, and the information they express can be considered similar. In classes with larger class sizes, the evaluation information expressed by the expert individual weight vectors aligns with the opinions of more evaluators, and the corresponding weight vectors should be assigned larger weight coefficients. Conversely, in classes with smaller class sizes, the weight vectors corresponding to the expert individual weight vectors should be assigned smaller weight coefficients. Let the i-th... The weight coefficients of the individual expert weight vector Weighting of individual experts Confidence factor Proportional, that is

[0179]

[0180] because

[0181]

[0182] achievable

[0183]

[0184]

[0185]

[0186] By weighting each expert's individual weight vector with its corresponding weight value, we can obtain the weight vector of the evaluation model.

[0187]

[0188] Assume there is There are 1 sample, denoted as _ . ,sample For indicators The evaluation value is recorded as To unify the range of variation of each indicator value, eliminate the influence of dimensions, and facilitate analysis, the evaluation values ​​of the sample must be standardized.

[0189] The method for processing benefit-type indicators

[0190]

[0191] The method for handling cost-type indicators

[0192]

[0193] in, , , .

[0194] According to the definition of entropy, the first The entropy value of each indicator is

[0195]

[0196] in, , In particular, when hour, But when hour, The value of is also 0, which violates the meaning of entropy. Therefore, let

[0197]

[0198] Then the first The objective weights of each evaluation indicator are:

[0199]

[0200] in, .

[0201] Combining the analytic hierarchy process (AHP) and the entropy method, the comprehensive weights of each indicator are obtained as follows:

[0202] .

[0203] The described embodiments are merely some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

Claims

1. A method for selecting evaluation metrics for AI-enabled distributed power management and scheduling, characterized in that, The specific steps include: Step S1: Construct a multi-dimensional dynamic evaluation system; This includes economic, environmental protection, and stability dimensions, among which... The economic dimension includes indicators of investment and operating costs, electricity purchase costs, network loss costs, and carbon emission costs, in order to quantify the contribution of distributed power dispatch to economic benefits. The environmental protection dimension includes indicators such as clean energy power penetration rate and renewable energy power generation consumption rate, in order to assess the positive impact of dispatch schemes on environmental protection; Stability dimensions include voltage fluctuation rate, power fluctuation at the point of common coupling, average outage time for users, and average outage frequency for the system, to ensure the stable operation of the power grid after the integration of distributed power sources. Step S2 involves assigning weights to the various evaluation indicators from Step S1 using a combination of the analytic hierarchy process (AHP) and the entropy method. This includes standardizing and classifying expert scores through cluster analysis to eliminate subjective bias, and then objectively quantifying the information entropy of the indicators based on the entropy method. Ultimately, this achieves a dynamic fusion of subjective and objective weights. Specifically: Subjective weights were determined using the analytic hierarchy process (AHP) based on cluster analysis. First, cluster analysis is performed on the individual weight vectors of the experts; Based on the characteristics of individual expert weight vectors, they are clustered into different classes and assigned different weight values; Finally, the weight vectors of different experts are weighted and averaged to obtain the weight vector values ​​of each evaluation model.

2. The method for selecting evaluation indicators for AI-enabled distributed power management and scheduling according to claim 1, characterized in that, In step S1, In the aforementioned economic dimension, economic cost The function expression is: ， in, Annual investment and operating costs of microgrids; Indicates the cost of purchasing electricity. Indicates network loss cost, Indicates the cost of carbon emissions; Annual investment and operating costs of microgrids The expression is: ， in, Indicates the number of nodes in the power distribution system; Represents a node Is a distributed generation (DG) installed at this location? Represents a node The capacity of the DG installation location; Indicates a fixed annual interest rate; Indicates the planning period; Represents a node Investment cost per unit capacity of DG; Represents a node Operating cost per unit capacity of DG; Electricity purchase cost The expression is: ， in, Indicates the unit price of electricity purchased; Indicates the maximum annual power grid utilization hours; Indicates the first The load of each node; Network loss cost The expression is: ， in, This indicates the unit electricity price for network losses; This indicates the maximum number of hours of power grid loss per year. Represents the set of all branches in the power distribution system; Indicates a branch The current amplitude; Indicates a branch The resistance value; Carbon emission costs The expression is: ， in, The environmental penalty is 45.61 yuan / ton, meaning a penalty of 45.61 yuan for every ton of CO2 produced. Represents the carbon emission factor, taking That is, 0.5703 kg of CO2 is emitted for every kilowatt-hour of electricity generated.

3. The method for selecting evaluation indicators for AI-enabled distributed power management and scheduling according to claim 1, characterized in that, In step S1, In the aforementioned environmental protection dimension, the clean energy electricity penetration rate The expression is: ， in, For the total scheduling time, The number of DGs connected to the power grid; Time period No. Each DG contributed its efforts; For time period node Active load; For time period of Active power loss of branch circuits; Renewable energy power generation absorption rate It can be defined as: ， in, For nodes Maximum predicted output power of renewable energy For nodes Actual output power of renewable energy.

4. The method for selecting evaluation indicators for AI-enabled distributed power management and scheduling according to claim 1, characterized in that, In step S1, In the aforementioned stability dimensions, Common point of coupling power fluctuation The expression is: ， in, express Electricity purchase volume during the time period; express Electricity sales volume during a given time period; Voltage fluctuation The expression is: ， in, Represents a node The voltage on; Represents a node The rated voltage; , For nodes Maximum and minimum voltages; System average power outage frequency The expression is: ， in, For nodes Average failure rate, For nodes The total number of users; The expression for the User Average Power Outage Time (CAIDI) is: ， in, For nodes The average annual power outage time.

5. The method for selecting evaluation indicators for AI-enabled distributed power management and scheduling according to claim 1, characterized in that, In step S2, the specific steps for assigning weights to each evaluation index using a combination of the analytic hierarchy process (AHP) and the entropy method include: The number of samples in the individual weight vector is The 10 evaluation indicators mentioned in step S1 can be used to obtain the expert individual weight vector. , , The expert individual weight vector matrix, calculated from the judgment matrix obtained from the expert questionnaire, is as follows: ， First, the individual weight vectors of each expert are standardized, that is: ， ， ， After standardization, the standardized expert individual weight vector can be obtained. , ; The standardized expert individual weight vectors are clustered using the hierarchical clustering method. The class average method is used to calculate the distance between classes, and the Euclidean distance is used for the distance between samples. Cluster analysis was performed on the samples using hierarchical clustering, and the samples were clustered into... kind; The results of cluster analysis Individual expert weight vectors are used to calculate the number of samples in each class. ; No. The individual weight vector of the nth expert belongs to the... Class, definition of the first The number of expert individual weight vectors contained in the class Total number of expert individual weight vectors The ratio is the expert individual weight vector The confidence factor, i.e. ， No. The weight coefficients of the individual expert weight vector Weighting of individual experts Confidence factor Proportional, that is ， because ， achievable ， ， ， By weighting each expert's individual weight vector with its corresponding weight value, we can obtain the weight vector of the evaluation model. ， Each sample is denoted as ,sample For indicators The evaluation value is recorded as ; The evaluation values ​​of the samples are standardized, whereby... The method for handling benefit-type indicators is as follows: ， The method for handling cost-related indicators is as follows: ， in, , , ; According to the definition of entropy, the first The entropy value of each indicator is ， in, , , make ， Then the first The objective weights of each evaluation indicator are: ， in, . Combining the analytic hierarchy process (AHP) and the entropy method, the comprehensive weights of each indicator are obtained as follows: 。

Images