Power grid intelligent customer service outbound multi-objective optimization method and system based on AGE-MOEA-II algorithm

Abstract

The invention provides a power grid intelligent customer service outbound multi-objective optimization method and system based on an AGE-MOEA-II algorithm. The method comprises the following steps: preprocessing historical outbound data; defining an objective function of the optimization problem; processing constraint conditions in the optimal outbound strategy; multi-target optimization search is executed through an AGE-MOEA-II algorithm, and an optimal outbound strategy is determined; and generating and outputting a final power grid intelligent customer service outbound strategy. According to the invention, the multi-target optimization algorithm is combined, so that the multi-dimensional targets can be balanced, the call-out scheduling strategy of the intelligent customer service of the power grid is optimized, and the efficiency and effect of the call-out strategy are improved; through optimal distribution of decision variables such as outbound time, outbound range, working time and the like, the call completing rate is maximized, meanwhile, possible transfer situations in the return visit process of a customer are reduced, the utilization rate of manual service resources is optimized, the economic cost of manual outbound is effectively reduced, and the economic benefit of the customer is improved. Therefore, on the premise of ensuring the service quality and improving the customer satisfaction, the operation cost is remarkably saved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid optimal dispatching. Specifically, it relates to a multi-objective optimization method and system for intelligent customer service outbound calls of the power grid based on the AGE-MOEA-II algorithm. Background Art

[0002] With the development of the power industry, the application of smart grids has gradually become an important means to improve power grid management efficiency and optimize resource allocation. In the smart grid system, the power grid intelligent customer service, as one of the important service supports, is responsible for providing users with services such as power-related information consultation, fault troubleshooting, and charge query. At present, most power grid intelligent customer service systems still rely on manual telephone return visits. During the manual outbound call process, the outbound call time and frequency are usually closely related to the labor cost, resulting in a large consumption of human resources and an increase in economic costs. In addition, the time and scope of manual outbound calls usually rely on manual experience, lacking scientific scheduling and optimization, leading to low customer service efficiency and difficult to guarantee customer satisfaction.

[0003] The artificial transfer rate is an important indicator to measure the efficiency of customer service work. In this context, how to optimize the outbound call strategy of the power grid intelligent customer service through intelligent means, improve service efficiency and reduce operating costs, has become a key problem to be solved urgently. In recent years, multi-objective optimization algorithms in the field of operations research, especially optimization methods based on evolutionary algorithms, have been widely applied to various scheduling optimization problems. Multi-objective optimization can find the optimal or near-optimal solution among multiple objectives, providing a scientific outbound call strategy for the power grid intelligent customer service.

[0004] However, most of the existing power grid customer service optimization methods focus on the optimization of a single objective or adopt simple rule-based scheduling, and fail to fully consider the collaborative optimization of multiple objectives such as connection rate, labor cost, and transfer rate. The patent document "A Power Grid Customer Service Work Order Intelligent Outbound Call System and Usage Method" (CN114205467A) discloses that by adding a return visit function, it realizes full-process intelligent return visits, automatically tracks user work order information, automatically initiates outbound calls, actively broadcasts the completion status of work orders, obtains user feedback information, and improves the customer service quality of power grid enterprises. However, it does not consider the optimal allocation of decision variables, and the efficiency and utilization rate are low.

[0005] The patent document "A Two - layer Multi - objective Optimal Scheduling Method for Active Distribution Networks" (CN118748436A) discloses a multi - objective snake - heron optimization algorithm designed based on the Pareto theory, which considers demand - side response, strengthens the interaction between the power supply side and the power consumption side, thereby improving the economy of the distribution network operation; by using well - performing ASBOA and MOSBOA for optimization, the lower - layer model and the upper - layer model are solved respectively, and finally the best active distribution network optimal scheduling strategy is obtained. However, it still relies on traditional evolutionary algorithms or single - objective weighting methods, lacks accurate modeling of the geometric characteristics of the Pareto front, has poor dynamic optimization ability for decision variables, and performs poorly in balancing the connection rate, labor cost, and transfer rate in the high - dimensional objective space.

[0006] Therefore, based on the multi - objective optimization algorithm, providing a set of intelligent outbound call strategies that can simultaneously improve the outbound call efficiency of the power grid customer service, reduce labor costs, and optimize the transfer rate, can effectively promote the digital transformation of the power industry while improving service quality, and promote the construction of smart grids, which has important practical significance. Summary of the Invention

[0007] Aiming at the defects in the prior art, the purpose of the present invention is to provide a multi - objective optimization method and system for the intelligent customer service outbound calls of the power grid based on the AGE - MOEA - II algorithm.

[0008] The multi - objective optimization method for the intelligent customer service outbound calls of the power grid based on the AGE - MOEA - II algorithm provided by the present invention includes:

[0009] Data processing step: Pre - process the historical outbound call data to generate an initial solution set;

[0010] Objective function step: Define the objective function of the AGE - MOEA - II algorithm, and iterate the initial solution set to obtain candidate solutions;

[0011] Optimization search step: Perform multi - objective optimization search through the AGE - MOEA - II algorithm, screen non - dominated solutions from the candidate solutions, and determine the optimal outbound call strategy;

[0012] Constraint handling step: Handle the constraint conditions in the optimal outbound call strategy;

[0013] Model output step: Generate the final intelligent customer service outbound call strategy of the power grid and output it.

[0014] Preferably, the historical outbound call data includes work order information, outbound call time, and connection status.

[0015] The pre - processing includes data cleaning and normalization processing, and key features are extracted through feature engineering.

[0016] The work order information includes work order ID and customer type.

[0017] In the data processing step, historical outbound call data is collected at the minimum interval between work orders to form an outbound call data record.

[0018] The output of the power grid intelligent customer service outbound call strategy includes the optimized outbound call time period, outbound call frequency, connection rate, labor cost, and transfer rate.

[0019] Preferably, the objective function f(x) optimizes the connection rate f1, labor cost f2(x), and transfer rate f3(x) by modeling the geometric shape of the Pareto front, maximizing f1(x), minimizing f2(x) and f3(x), and f(x) = Minimize{-f1(x), f2(x), f3(x)}.

[0020] The connection rate is calculated based on historical outbound call data. base_rate i = exp(-a j ·(t i -μ j1 )) 2 );

[0021] where priority_weight i is the priority weight of work order i;

[0022] t i is the outbound call time of the i-th work order;

[0023] μ j1 is the timestamp with the highest connection rate within the j-th time period;

[0024] a j is the attenuation coefficient within the j-th time period;

[0025] N represents the total number of work orders.

[0026] The labor cost base_cost i = C j + b j ·|t i -μ j2 |);

[0027] where priority_factor i is the priority influence coefficient of work order i;

[0028] base_cost i is the basic labor cost of work order i;

[0029] C j is the basic cost within the j-th time period;

[0030] b j is the coefficient related to the time deviation degree in the j-th time period;

[0031] μ j2 is the timestamp with the lowest labor cost in the j-th time period.

[0032] The transfer rate

[0033] where priority_effect i and base_prob i are both the transfer rate priority influence coefficients of work order i;

[0034] D j is the basic transfer rate within the j-th time period;

[0035] μ j3 is the timestamp with the lowest manual transfer rate in the j-th time period

[0036] k j1 and k j2 are respectively the amplitude of change in the control transfer rate and the steepness of the change in the transfer rate in the j-th time period.

[0037] Preferably, in the optimization search step, it is optimized by modeling the geometric shape of the Pareto front, the Pareto front is modeled by an adaptive geometric estimation method, and the candidate solutions are classified by non-dominated sorting to stratify all candidate solutions;

[0038]

[0039]

[0040] Use the Newton-Raphson iteration formula, with the initial value set to p0 = 1, Iterate until the error is less than the threshold;

[0041] where A and B are two points in the solution set;

[0042] a1,…a M and b,…b M respectively represent the M-dimensional coordinates of points A and B;

[0043] γ is the set of all smooth curves from A to B;

[0044] |γ ′ (t)| is the speed of the curve;

[0045] gd(A,B) is the geodesic distance between points A and B;

[0046] C ⊥ is the projection of the midpoint C of the straight line between points A and B;

[0047] ||·||2 represents the Euclidean distance;

[0048] p represents the geometric parameter of the Pareto front curvature;

[0049] a i is the value of the non-dominated solution on the i-th objective;

[0050] M is the number of target points;

[0051] p n+1 represents the curvature parameter value obtained by n + 1 iterative calculations;

[0052] p n represents the curvature parameter value at the n-th iteration.

[0053] Preferably, the constraint conditions include the minimum interval between work orders and the preferential work orders are preferentially called out.

[0054] The minimum interval between work orders is determined according to the work order priority and the calling time, and is embedded in the optimization model in the form of non-linear constraints.

[0055] The preferential work orders are preferentially called out by embedding the corresponding priority rules in the objective function.

[0056] In the constraint handling step, by judging the feasibility of the solution set, the search direction of the solution is adjusted or a penalty term is added to the objective function for correction.

[0057] where f(x) is the objective function;

[0058] g i (x) is the i-th constraint function;

[0059] m is the total number of constraint functions;

[0060] λ is the penalty factor.

[0061] In the model output step, according to the actual requirements, the relative importance weights of the connection rate, labor cost, and transfer rate target values are defined, the weighted target values are calculated, and the non-dominated solution with the optimal target value is selected through comprehensive evaluation to output the final power grid intelligent customer service outbound call strategy.

[0062] According to the present invention, a multi-objective optimization system for power grid intelligent customer service outbound calls based on the AGE-MOEA-II algorithm is provided, including a data processing module, a model construction module, and an outbound call strategy optimization module.

[0063] The data processing module preprocesses historical outbound call data to generate an initial solution set;

[0064] The model construction module defines the objective function of the AGE-MOEA-II algorithm and iterates the initial solution set to obtain candidate solutions;

[0065] The outbound call strategy optimization module performs multi-objective optimization search through the AGE-MOEA-II algorithm, screens non-dominated solutions from the candidate solutions, and determines the optimal outbound call strategy;

[0066] The model construction module processes the constraint conditions in the optimal outbound call strategy, generates the final outbound call strategy for the power grid intelligent customer service and outputs it.

[0067] Preferably, the historical outbound call data includes work order information, outbound call time, and connection status.

[0068] The preprocessing includes data cleaning and normalization processing, and extracts key features through feature engineering.

[0069] The work order information includes work order ID and customer type.

[0070] In the data processing module, historical outbound call data is collected at the minimum interval between work orders to form an outbound call data record.

[0071] The output of the outbound call strategy for the power grid intelligent customer service includes the optimized outbound call time period, outbound call frequency, connection rate, labor cost, and transfer rate.

[0072] Preferably, the objective function f(x) optimizes the connection rate f1, labor cost f2(x), and transfer rate f3(x) by modeling the geometric shape of the Pareto front, maximizes f1(x), minimizes f2(x) and f3(x), and f(x) = Minimize{-f1(x), f2(x), f3(x)}.

[0073] The connection rate is calculated based on historical outbound call data, base_rate i =exp(-a j ·(t i -μ j1 ) 2 );

[0074] where, priority_weight i is the priority weight of work order i;

[0075] t i is the outbound call time of the i-th work order;

[0076] μ j1 is the timestamp with the highest connection rate in the j-th time period;

[0077] a j is the attenuation coefficient in the j-th time period;

[0078] N represents the total number of work orders.

[0079] The labor cost base_cost i = C j + b j ·|t i - μ j2 |);

[0080] where priority_factor i is the priority influence coefficient of work order i;

[0081] base_cost i is the basic labor cost of work order i;

[0082] C j is the basic cost in the j-th time period;

[0083] b j is the coefficient related to the degree of time deviation in the j-th time period;

[0084] μ j2 is the time stamp with the lowest labor cost in the j-th time period.

[0085] The transfer rate

[0086] where priority_effect i and base_prob i are both the transfer rate priority influence coefficients of work order i;

[0087] D j is the basic transfer rate in the j-th time period;

[0088] μ j3 is the time stamp with the lowest manual transfer rate in the j-th time period

[0089] k j1 and k j2 are respectively the control transfer rate change amplitude and the steepness of the transfer rate change in the j-th time period.

[0090] Preferably, in the outbound call strategy optimization module, the geometric shape of the Pareto front is optimized by modeling, the Pareto front is modeled by an adaptive geometric estimation method, the candidate solutions are classified by non-dominated sorting, and all candidate solutions are stratified;

[0091]

[0092] Using the Newton - Raphson iteration formula, the initial value is set to p0 = 1, Iterate until the error is less than the threshold;

[0093] where A and B are two points in the solution set;

[0094] a1,…a M 、b,…b M represent the M - dimensional coordinates of points A and B respectively;

[0095] γ is the set of all smooth curves from A to B;

[0096] |γ ′ (t)| is the velocity of the curve;

[0097] gd(A,B) is the geodesic distance between points A and B;

[0098] C ⊥ is the projection of the mid - point C of the straight line between points A and B;

[0099] ||·||2 represents the Euclidean distance;

[0100] p represents the geometric parameter of the Pareto front curvature;

[0101] a i is the value of the non - dominated solution on the i - th objective;

[0102] M is the number of target points;

[0103] p n+1 represents the curvature parameter value obtained by n + 1 iterative calculations;

[0104] p n represents the curvature parameter value at the n - th iteration.

[0105] Preferably, the constraint conditions include the minimum interval between work orders and the preferential outbound calls for preferential work orders.

[0106] The minimum interval between work orders is determined according to the work order priority and the outbound call time, and is embedded in the optimization model in the form of non - linear constraints.

[0107] The preferential outbound calls for preferential work orders are processed by embedding the corresponding priority rules in the objective function.

[0108] In the model construction module, by judging the feasibility of the solution set, the search direction of the solution is adjusted or a penalty term is added to the objective function for correction,

[0109] Among them, f(x) is the objective function;

[0110] g i (x) is the i-th constraint function;

[0111] m is the total number of constraint functions;

[0112] λ is the penalty factor.

[0113] Define the relative importance weights of the connection rate, labor cost, and transfer rate target values according to actual requirements, calculate the weighted target values, select the non-dominated solution with the optimal target value through comprehensive evaluation, and output the final outbound strategy of the power grid intelligent customer service.

[0114] Compared with the prior art, the present invention has the following beneficial effects:

[0115] 1. By optimizing the allocation of decision variables such as outbound time, outbound range, and working hours, the present invention maximizes the connection rate while effectively reducing the economic cost of manual outbound calls, thereby significantly saving the operating cost on the premise of ensuring service quality and improving customer satisfaction.

[0116] 2. By optimizing the outbound sequence and time period, the present invention reduces the possible transfer situations that customers may encounter during the return visit process, optimizes the utilization rate of manual service resources, and further improves the overall efficiency of the customer service system.

[0117] 3. By combining the multi-objective optimization algorithm, the present invention can balance among multi-dimensional objectives (such as connection rate, labor cost, transfer rate), optimize the outbound scheduling strategy of the power grid intelligent customer service, and improve the efficiency and effect of the outbound strategy.

[0118] 4. The present invention adopts the AGE-MOEA-II algorithm, realizes the multi-dimensional global search of the outbound strategy by introducing the adaptive geometric estimation (AGE) and the curvature modeling of the Pareto front, and combining the geodesic distance to measure the diversity of the solution set. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] By reading the detailed description of the non-restrictive embodiments with reference to the following drawings, other features, objectives, and advantages of the present invention will become more obvious:

[0120] Figure 1 It is a schematic diagram of the power grid customer service outbound optimization process based on the AGE-MOEA-II algorithm.

[0121] Figure 2 It is a schematic diagram of the Pareto front based on the AGE-MOEA-II algorithm. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0122] The present invention will be described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, several changes and improvements can be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0123] According to the present invention, a multi-objective optimization method for outbound calls of power grid intelligent customer service based on the AGE-MOEA-II algorithm is provided. Combining the AGE-MOEA-II algorithm, a new and multi-dimensional optimization method is provided to solve problems such as high cost and low efficiency faced in the traditional manual outbound call process. While improving the service level of power grid customer service, it also provides technical support for the development of intelligent customer service in the power industry. Specifically, the AGE-MOEA-II architecture method includes the following steps: data input step, objective function step, optimization search step, constraint handling step, and result output step.

[0124] Data input step: Preprocess the historical outbound call data of the power grid intelligent customer service, mainly including cleaning and normalizing data such as work order information, outbound call time, connection status, etc. Extract key features through feature engineering and use these features as the input of the model. Especially for the priority of work orders, characteristics of outbound call time periods, etc., targeted processing is carried out to make the input data have high quality and usability. Through effective processing of the data, stable and accurate input data is provided for the subsequent optimization model.

[0125] Specifically, the targeted processing refers to custom analysis and optimization processing of key variables such as work order priority and outbound call time period according to the specific requirements and data characteristics of the power grid intelligent customer service outbound call task. Specifically, this processing includes the following steps:

[0126] Based on the priority of the work order (such as urgency or customer importance), classify and grade it and assign corresponding weights to reflect the importance differences in outbound call scheduling.

[0127] Extract outbound call time and priority data from historical outbound call data, and calculate and set the minimum interval between work orders in combination with actual operation requirements (such as avoiding repeated calls). This constraint, as an independent parameter, is generated through feature engineering and input into the optimization model together with information such as work order ID and customer type.

[0128] For the characteristics of the outbound call time period (such as peak historical connection rate periods or customer active periods), extract time-related feature laws through statistical analysis or pattern recognition.

[0129] These processes aim to enhance the representativeness and pertinence of the data, ensuring that the input data can fully reflect the multi-objective optimization requirements of the outbound call task (such as connection rate, cost, transfer rate). In this way, the quality and usability of the data are improved, providing more accurate basic support for the subsequent optimization model based on the AGE-MOEA-II algorithm.

[0130] The input of the constructed model includes the historical outbound call data of the power grid intelligent customer service and the time period characteristic information related to the outbound calls, providing accurate and high-quality data support for the AGE-MOEA-II algorithm to ensure the effectiveness of the outbound call strategy optimization.

[0131] First of all, the historical outbound call data is the core input of the model. It includes detailed information of each work order, such as work order ID, customer type, outbound call time, connection status, etc. These data are usually collected at fixed time intervals (such as every hour or every day) and form real outbound call data records in the operation process of the power grid intelligent customer service. Through these historical data, the model can capture the impact of different outbound call time periods on the connection rate and transfer rate, as well as the potential impact of the priority of the work order on the outbound call effect. In the data preprocessing stage, the original historical outbound call data is cleaned and normalized to eliminate outliers and the influence of different scales in the data, thus ensuring the consistency and accuracy of the data.

[0132] Secondly, the time period characteristic information related to the outbound calls is also an important part of the input. The optimization of the outbound call strategy not only depends on the specific content of the work order, but the time period characteristics are equally crucial. The impact of different outbound call time periods (such as weekdays and weekends, day and night) on the connection rate and transfer rate varies greatly, so the impact of these time periods needs to be considered in the input data. Through the analysis of the historical outbound call data, the model can identify the effect of making outbound calls during specific time periods and adjust the outbound call plan accordingly to improve the overall connection rate, reduce the transfer rate and labor costs.

[0133] Through the effective processing of these input data, the model can provide stable and accurate basic data for the subsequent optimization process. These data provide rich information for the AGE-MOEA-II algorithm, enabling the optimization process to find the best balance among multiple objectives and ultimately realizing the optimization of the outbound call strategy of the power grid intelligent customer service.

[0134] Objective function steps: As the core part of the AGE-MOEA-II algorithm, it is responsible for defining the objective function of the optimization problem, specifically including:

[0135] For the design of the objective function of the optimization model for the outbound calls of the power grid intelligent customer service, it follows the core idea of the AGE-MOEA-II algorithm, that is, to optimize multiple objectives by modeling the geometric shape of the Pareto front, rather than by weighted combination of multiple objective functions. Specifically, the model needs to optimize the following three objectives: maximizing the connection rate f1(x), minimizing the labor cost f2(x) and the transfer rate f3(x).

[0136] f(x) = Minimize{-f1(x), f2(x), f3(x)}

[0137] Maximizing the connection rate: According to the historical outbound call data, optimize the outbound call time and frequency to maximize the connection rate of outbound calls and improve customer satisfaction;

[0138] Specifically, the goal is to maximize the customer connection rate during the outbound calls of the power grid intelligent customer service. The calculation of the connection rate is based on historical outbound call data to evaluate the impact of different outbound call periods on the connection rate. This objective function aims to find the optimal outbound call time arrangement so that more customers can answer the phone smoothly.

[0139]

[0140] base_rate i = exp(-a j ·(t i - μ j1 ) 2 )

[0141] where base_rate i , priority_weight i are the basic transfer rate and priority weight of work order i (there are N in total), t i is the outbound call time of the i-th work order, μ j1 is the timestamp with the highest connection rate in the j-th time period, and a j is the attenuation coefficient in the j-th time period.

[0142] Minimizing the labor cost: According to different outbound call periods, optimize the outbound call plan, reduce the manual intervention time, and lower the labor cost;

[0143] Specifically, the goal is to minimize the labor cost required during the outbound calls. By optimizing the arrangement of the outbound call periods, the manual intervention time is reduced, thus reducing the cost. This objective function guides the adjustment of the outbound call strategy by measuring the costs of different outbound call periods and frequencies.

[0144]

[0145] base_cost i = Cj +b j ·|t i -μ j2 |)

[0146] where base_cost i and priority_factor i are the basic labor cost and the priority impact coefficient of work order i (there are N in total), C j is the basic cost in the j-th time period, b j is the coefficient related to the time deviation degree in the j-th time period, controlling the growth rate of the cost, t i is the outbound call time of the i-th work order, μ j2 is the time stamp with the lowest labor cost in the j-th time period.

[0147] Minimize the transfer rate: By reasonably arranging the outbound call sequence and time period, reduce the manual transfer caused by untimely outbound calls or unresolved problems, and improve the outbound call efficiency.

[0148] Specifically, the goal is to reduce the manual transfer caused by unresolved problems or unreasonable outbound call time periods. The optimization of the transfer rate reduces the number of work orders that need to be transferred by reasonably arranging the outbound call sequence and time period, thereby improving the service efficiency.

[0149]

[0150] where base_prob i and priority_effect i are the priority impact coefficients of work order i (there are N in total), D j is the basic transfer rate in the j-th time period, k j1 and k j2 in the j-th time period, respectively control the change range and the steepness of the change of the transfer rate, t i is the outbound call time of the i-th work order, μ j3 is the time stamp with the lowest manual transfer rate in the j-th time period.

[0151] Using the AGE-MOEA-II algorithm, by introducing adaptive geometric estimation (AGE) and curvature modeling of the Pareto front, and combining geodesic distance to measure the diversity of the solution set, a multi-dimensional global search for the outbound strategy is achieved. The AGE-MOEA-II algorithm models the geometric characteristics of the Pareto front and measures the diversity of solutions based on non-dominated sorting and geographical distance, ensuring that these three objectives can be taken into account simultaneously during the optimization process; it calculates the curvature of the Pareto front through the Newton-Raphson method, and then optimizes the diversity and convergence of the outbound strategy, thus finding the best balance among multiple optimization objectives.

[0152] This optimization process avoids combining multiple objectives by weighting them into a single objective. Instead, through the independent optimization of each objective and the geometric modeling of the Pareto front, the final solution achieves a good balance among multiple objectives.

[0153] By combining multi-objective optimization algorithms, it is possible to trade off among multi-dimensional objectives (such as connection rate, labor cost, transfer rate), optimize the outbound scheduling strategy of the power grid intelligent customer service, and improve the efficiency and effectiveness of the outbound strategy; by optimizing decision variables such as outbound time and outbound scope, the connection rate of outbound calls is increased, manual intervention is reduced, customer satisfaction is improved, and thus the overall service efficiency is enhanced. On the premise of ensuring service quality, the operation efficiency is greatly improved, costs are reduced, and the customer experience is optimized, thereby providing a practical solution for practical applications.

[0154] Optimization search steps: Perform multi-objective optimization search through the AGE-MOEA-II algorithm to find the optimal outbound strategy to achieve the multi-objective balance of maximizing the connection rate, minimizing the labor cost, and minimizing the transfer rate. The AGE-MOEA-II algorithm does not optimize multiple objectives by simply weighting the objective function, but realizes optimization by modeling the geometric shape of the Pareto front, can trade off among multiple objectives, and models the Pareto front through the adaptive geometric estimation method to help quickly converge to the optimal solution.

[0155] Specifically, the search process optimizes the parameters extracted and quantified in the data preprocessing step, including the connection rate characteristics of the outbound time, the outbound frequency, and the classification results of the work order priorities, and inputs them as decision variables into the AGE-MOEA-II algorithm. The minimum interval between work orders is embedded in a non-linear form to ensure that the outbound strategy meets the time interval requirements, thereby improving efficiency and reducing resource waste. Determine its optimal combination through global search. Use the objective function and constraints to guide the algorithm to converge to the Pareto front.

[0156] The objective function consists of three parts: the connection rate function (maximized, calculated based on historical data and time characteristics), the labor cost function (minimized, considering time allocation and basic costs), and the transfer rate function (minimized, evaluating the transfer probability). These functions define the multi-objective optimization problem in a mathematical form. The constraint conditions include the minimum interval between workstations (to avoid repeated calls) and the preferential outbound calls for priority work orders (to ensure that high-priority tasks are processed first). The constraint processing module embeds the algorithm to ensure that the search results meet the actual operation requirements. AGE-MOEA-II uses non-dominated sorting and geometric modeling to iteratively optimize the decision variables based on these functions and constraints until it converges to the optimal solution set.

[0157] Through geometric modeling and non-dominated sorting, the outbound call strategy of the power grid intelligent customer service is optimized, significantly improving the operation efficiency, reducing manual intervention, and lowering costs, thus bringing economic benefits to power grid enterprises. Especially in complex power grid customer service scenarios, while taking into account the non-linear relationship between multiple objectives, the flexibility and adaptability are more prominent. It can be extended to other intelligent customer service fields, such as the financial and telecommunications industries, further enhancing the ability to solve multi-objective optimization problems within the industry, and providing more efficient and accurate technical support for the optimization and development of intelligent customer service systems.

[0158] Adopting the Pareto front modeling technology can find the optimal balance point among multiple conflicting objectives, thus ensuring that while optimizing the outbound call time and frequency, the connection rate is maximized and the labor cost and transfer rate are minimized. This not only improves the outbound call efficiency but also ensures the flexibility and reliability of the outbound call process, with high adaptability and operability.

[0159] During the search process, AGE-MOEA-II first classifies the solution set through non-dominated sorting, dividing all solutions into multiple levels to ensure that each solution can find its own position on the Pareto front.

[0160] The solution set includes the data representation of multiple outbound call strategy schemes, and each scheme consists of a set of decision variables, specifically including the outbound call time (t i ), the outbound call frequency, and the work order priority weight (priority_weight). These variables correspond to the output values of objective functions such as the connection rate (f1(x)), the labor cost (f2(x)), and the transfer rate (f3(x)).

[0161] Randomly generate the initial solution set from the preprocessed historical outbound call data (such as work order information, time period characteristics); subsequently, through the evolutionary operations (crossover, mutation, etc.) of the algorithm combined with the evaluation of the objective function, iteratively generate new candidate solutions. Non-dominated sorting then stratifies these candidate solutions according to the objective values, screening out the non-dominated solutions on the Pareto front to ensure that the solution set reflects the diversity and convergence of multi-objective optimization.

[0162] Non - dominated sorting is a method used to calculate the Pareto front in multi - objective optimization problems. Specifically, a solution "dominates" another solution if it is better than the other solution in at least one objective and at least not worse than it in other objectives. The goal of non - dominated sorting is to stratify the solution set and find all non - dominated solutions, that is, the Pareto front. For two solutions x1 and x2, if x1 dominates x2 (denoted as x1 < x2), the following conditions are met (assuming the goal is to minimize):

[0163] There are a total of m objectives. For all objectives i = 1, 2,..., m, there is f i (x1) ≤ f i (x2) (that is, x1 is not worse than x2 in all objectives);

[0164] There exists at least one objective j (1 ≤ j ≤ m) such that f j (x1) < f j (x2) (that is, x1 is strictly better than x2 in at least one objective).

[0165] To ensure the diversity and convergence of optimization, the AGE - MOEA - II algorithm captures the curvature characteristics and uses the geodesic distance metric method. This method can measure the distribution of the solution set in the solution space, ensuring that the solution set can not only converge to the Pareto front but also maintain appropriate diversity among different objectives. Specifically, the diversity of the solution set is evaluated by calculating the geometric distance between solutions, which in turn guides the search process to avoid over - concentrating on a certain part of the solutions, thereby improving the quality of the entire solution set.

[0166]

[0167] Among them, A and B are two points in the solution set, Υ is the set of all smooth curves from A to B, |Υ ′ (t)| is the speed of the curve (or local geometric measure), and t represents the integration variable. Based on the geodesic distance, the diversity of solutions in the solution set can be effectively measured. Since it takes into account the non - linear relationship of solutions in the objective space, it can more accurately measure the differences between solutions, especially when the Pareto front presents a complex shape.

[0168]

[0169] Among them, a1,…a M , and b,…b M , respectively represent the M - dimensional coordinates of points A and B, gd(A, B) measures the geodesic distance between points A and B, C ⊥It is the projection of the midpoint C of the straight line between points A and B. p represents the geometric parameter of the curvature of the Pareto front, and ||·||2 represents the Euclidean distance, which is only used as a tool for calculating locally.

[0170] In the actual search process, the AGE-MOEA-II algorithm models the Pareto front through the Adaptive Geometry Estimation (AGE) method. By introducing a new geometric representation method, it not only improves the convergence speed of the solution but also enhances the algorithm's performance in the high-dimensional objective space. AGE-MOEA-II uses the Newton-Raphson iteration method for the geometric modeling of the Pareto front, accurately calculating the curvature p of the front, enabling the optimization process to finely explore the balance point between multiple objectives. This modeling method enables the solution set to converge more efficiently to the ideal Pareto front, achieving the optimal balance among various objectives.

[0171] AGE-MOEA-II determines p, a by solving the roots of the following non-linear equation i is the value of the non-dominated solution on the i-th objective, M is the number of objective points, and p controls the curvature of the front:

[0172]

[0173] Using the Newton-Raphson iteration formula, the initial value is set to p0 = 1 (flat manifold), and the iteration continues until the error is less than the threshold (such as |p n+1 - p n | ≤ 0.001). This ensures that p can quickly and accurately reflect the curvature of the front.

[0174]

[0175] where p n+1 is the curvature parameter value obtained by n + 1 times of iterative calculation, representing the updated estimate of the geometric shape of the Pareto front, and p n is the curvature parameter value at the n-th iteration, serving as the basis for the current estimate.

[0176] During the optimization process, at each iteration, the AGE-MOEA-II algorithm adjusts the search direction according to the geometric relationship between the current solution and other solutions, enabling the solution set to gradually evolve along the Pareto front. Finally, through multiple iterations, the algorithm can obtain a set of Pareto optimal solutions, which represent the trade-offs and balances between different objectives. This process ensures that the outbound call strategy of the power grid intelligent customer service can obtain the optimal decision-making scheme on multiple objectives. By combining the optimization theory of operations research with the actual needs of the power grid intelligent customer service, the efficiency and effect of the outbound call strategy are significantly improved.

[0177] Constraint handling steps: Handle various constraint conditions that need to be satisfied during the outbound call process. In the multi-objective optimization process, the handling of constraint conditions has an important impact on the optimization effect. The AGE-MOEA-II algorithm adopts a constraint handling method based on a geometric model to ensure that the solution set not only meets the requirements of the objective function but also follows the preset constraint conditions. In the present invention, the main constraint conditions involve the minimum interval between work orders and the scheduling problem of priority work orders. Specifically include:

[0178] Minimum interval between work orders: According to the priority of the work order and the outbound call time, ensure that there is enough time interval between work orders to avoid repeated calls;

[0179] Specifically, the minimum interval constraint between work orders ensures that there is enough time interval between different work orders during the outbound call process, avoiding overly intensive outbound call tasks from affecting the overall efficiency. The design of this constraint is based on the actual needs of the power grid intelligent customer service, avoiding repeated calls to the same customer within the same time period, resulting in resource waste or customer dissatisfaction. During the optimization process, this constraint is embedded in the optimization model in the form of a non-linear constraint to ensure that during each iteration, the algorithm will not violate this constraint during the search process.

[0180] Priority work orders are called first: Ensure that priority work orders are called earlier than ordinary work orders to ensure that customer needs are responded to in a timely manner.

[0181] Specifically, calling priority work orders first is another important constraint to ensure that high-priority work orders are processed first. This constraint requires that during the optimization search, the outbound call time of high-priority work orders is earlier than that of low-priority work orders to ensure that the urgent needs of customers are responded to in a timely manner. This constraint is processed by embedding the corresponding priority rules in the objective function to ensure that the outbound call plan of priority work orders is always prior to that of ordinary work orders.

[0182] To effectively handle these constraints, the AGE-MOEA-II algorithm judges the feasibility of the solution set. If a solution violates the constraint conditions, it will correct by adjusting the search direction of the solution or adding a penalty term to the objective function.

[0183]

[0184] Among them, f(x) is the objective function, g i (x) is the i-th (a total of m) constraint function, and λ is the penalty factor. When the constraint is not satisfied, the penalty term will affect the value of the objective function and push the optimization process towards the feasible solution region.

[0185] In this way, AGE-MOEA-II can continuously guide the solution set to search along the feasible solution space during the optimization process, ensuring that the finally output solutions meet all constraint conditions, so as to provide an optimized result that meets the actual operation requirements for the outbound call strategy of the power grid intelligent customer service.

[0186] Model output step: Based on the results of the AGE-MOEA-II optimization algorithm, generate the final outbound call strategy for the power grid intelligent customer service. Specifically, during the optimization process of the AGE-MOEA-II algorithm, the results output by the model directly affect the optimization effect of the outbound call strategy of the power grid intelligent customer service. The final output includes multiple indicators such as the optimized outbound call time period, outbound call frequency, connection rate, labor cost, and transfer rate. These indicators comprehensively reflect the optimal solution found among multiple objectives. Through comprehensive evaluation of the comprehensive performance of the solution set on the Pareto front, the outbound call plan that best meets the optimization objectives is output, providing an efficient outbound call strategy for the power grid intelligent customer service.

[0187] Specifically, the AGE-MOEA-II algorithm generates a set of non-dominated solutions, each solution corresponding to decision variables such as the outbound call time period, outbound call frequency, and their objective values (connection rate, labor cost, transfer rate). According to the actual requirements of the power grid intelligent customer service (such as giving priority to improving the connection rate or reducing costs), define the relative importance weights of each objective; then, calculate the weighted objective value of each solution or use the Hypervolume metric to quantify the coverage of the Pareto front; finally, select the solution with the optimal objective value or the largest Hypervolume contribution under the weight condition as the optimal outbound call plan. This method ensures that the output strategy achieves the best balance among multiple objectives and meets the actual application requirements.

[0188] Through the AGE-MOEA-II algorithm, the model can generate a set of Pareto optimal solutions, representing the best strategies for trade-offs among different objectives. These solutions not only maximize the connection rate and improve customer satisfaction, but also achieve ideal results in controlling labor costs and reducing the transfer rate. During the optimization process, each solution in the solution set represents an outbound call strategy that balances the requirements of multiple objectives, providing decision-makers with the option to select the most suitable plan according to specific circumstances.

[0189] In more preferred examples, the outbound call time period and outbound call frequency will be adjusted according to the optimization results to ensure effective outbound calls during the customer active period, thereby increasing the connection rate. In addition, the labor cost will be optimized according to the outbound call plan to avoid unnecessary manual intervention, thus achieving cost minimization. The minimization of the transfer rate is achieved by optimizing the outbound call order and priority handling to ensure that customer problems can be solved upon the first connection, reducing subsequent transfer requirements.

[0190] The present invention also provides a multi-objective optimization system for intelligent power grid customer service outbound calls based on the AGE-MOEA-II algorithm. The multi-objective optimization system for intelligent power grid customer service outbound calls based on the AGE-MOEA-II algorithm can be implemented by executing the process steps of the multi-objective optimization method for intelligent power grid customer service outbound calls based on the AGE-MOEA-II algorithm. That is, those skilled in the art can understand the multi-objective optimization method for intelligent power grid customer service outbound calls based on the AGE-MOEA-II algorithm as a preferred implementation manner of the multi-objective optimization system for intelligent power grid customer service outbound calls based on the AGE-MOEA-II algorithm.

[0191] According to a multi-objective optimization system for intelligent power grid customer service outbound calls based on the AGE-MOEA-II algorithm provided by the present invention, the outbound call strategy is multi-objectively optimized through an operations research algorithm, and the optimization objectives include the connection rate, labor cost, and manual transfer rate. By designing reasonable objective functions and constraint conditions and combining the advantages of evolutionary algorithms, the efficient scheduling of the intelligent power grid customer service outbound call strategy is realized.

[0192] Specifically, it includes three modules: data processing, model construction, and outbound call strategy optimization.

[0193] Data processing: Preprocess the historical outbound call data of the intelligent power grid customer service, including steps such as data cleaning, feature engineering, and normalization processing, to ensure the high quality of the input data and provide a stable basis for model construction.

[0194] Model construction: Based on the AGE-MOEA-II algorithm, construct a multi-objective optimization model suitable for optimizing the intelligent power grid customer service outbound calls. Through reasonable objective functions and constraint conditions, the collaborative optimization of multiple objectives such as the connection rate, labor cost, and transfer rate is realized.

[0195] Outbound call strategy optimization: Use the AGE-MOEA-II optimization algorithm to globally search the outbound call strategy of the power grid customer service, optimize decision variables such as the outbound call time period and outbound call frequency, and finally achieve the goals of maximizing the connection rate, minimizing the cost, and minimizing the transfer rate.

[0196] In more preferred examples, the historical outbound call data includes work order information, outbound call time, and connection status.

[0197] The preprocessing includes data cleaning and normalization processing, and key features are extracted through feature engineering.

[0198] The work order information includes the work order ID and customer type.

[0199] In the data processing module, the historical outbound call data is collected at the minimum interval between work orders to form an outbound call data record.

[0200] The output of the grid intelligent customer service outbound call strategy includes the optimized outbound call time period, outbound call frequency, connection rate, labor cost, and transfer rate.

[0201] In more preferred examples, the objective function f(x) optimizes the connection rate f1, labor cost f2(x), and transfer rate f3(x) by modeling the geometric shape of the Pareto front, maximizing f1(x), minimizing f2(x) and f3(x), and f(x) = Minimize{-f1(x), f2(x), f3(x)}.

[0202] The connection rate is calculated based on historical outbound call data. base_rate i =exp(-a j ·(t i -μ j1 ) 2 );

[0203] where priority_weight i is the priority weight of work order i;

[0204] t i is the outbound call time of the i-th work order;

[0205] μ j1 is the timestamp with the highest connection rate within the j-th time period;

[0206] a j is the attenuation coefficient within the j-th time period;

[0207] N represents the total number of work orders.

[0208] The labor cost base_cost i =C j +b j ·|t i -μ j2 |);

[0209] where priority_factor i is the priority impact coefficient of work order i;

[0210] base_cost i is the basic labor cost of work order i;

[0211] C j is the basic cost within the j-th time period;

[0212] b j is the coefficient related to the degree of time deviation within the j-th time period;

[0213] μ j2 is the timestamp with the lowest labor cost in the j-th time period.

[0214] The transfer rate

[0215] where priority_effect i and base_prob i are both the transfer rate priority influence coefficients of work order i;

[0216] D j is the basic transfer rate within the j-th time period;

[0217] μ j3 is the timestamp with the lowest manual transfer rate in the j-th time period

[0218] k j1 and k j2 are respectively the control range of the transfer rate change and the steepness of the transfer rate change in the j-th time period.

[0219] In more preferred examples, in the outbound call strategy optimization module, the geometric shape of the Pareto front is optimized by modeling, the Pareto front is modeled by an adaptive geometric estimation method, and the candidate solutions are classified by non-dominated sorting to stratify all candidate solutions;

[0220]

[0221] Use the Newton-Raphson iteration formula, with the initial value set to p0 = 1, Iterate until the error is less than the threshold;

[0222] where A and B are two points in the solution set;

[0223] a1,…a M and b,…b M respectively represent the M-dimensional coordinates of points A and B;

[0224] γ is the set of all smooth curves from A to B;

[0225] |γ ′ (t)| is the speed of the curve;

[0226] gd(A,B) is the geodesic distance between points A and B;

[0227] C ⊥ is the projection of the midpoint C of the straight line between points A and B;

[0228] ||·||2 represents the Euclidean distance;

[0229] p represents the geometric parameter of the Pareto front curvature;

[0230] a i is the value of the non-dominated solution on the i-th objective;

[0231] M is the number of target points;

[0232] p n+1 represents the curvature parameter value obtained from the (n + 1)-th iterative calculation;

[0233] p n represents the curvature parameter value at the n-th iteration.

[0234] In more preferred examples, the constraint conditions include the minimum interval between work orders and the priority work orders are called first.

[0235] The minimum interval between work orders is determined according to the work order priority and the call time, and is embedded in the optimization model in the form of non-linear constraints.

[0236] The priority work orders are called first by embedding the corresponding priority rules in the objective function.

[0237] In the model construction module, by judging the feasibility of the solution set, the search direction of the solution is adjusted or a penalty term is added to the objective function for correction.

[0238] where f(x) is the objective function;

[0239] g i (x) is the i-th constraint function;

[0240] m is the total number of constraint functions;

[0241] λ is the penalty factor.

[0242] Define the relative importance weights of the connection rate, labor cost, and transfer rate target values according to actual needs, calculate the weighted target values, select the non-dominated solution with the optimal target value through comprehensive evaluation, and output the final power grid intelligent customer service outbound call strategy.

[0243] Those skilled in the art know that, in addition to implementing the system and its various devices, modules, and units provided by the present invention in the form of pure computer-readable program code, it is entirely possible to achieve the same functions by logically programming the method steps so that the system and its various devices, modules, and units provided by the present invention are implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers, etc. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a kind of hardware component, and the devices, modules, and units included therein for implementing various functions can also be regarded as the structures within the hardware component; it can also be considered that the devices, modules, and units for implementing various functions are both software modules for implementing the method and the structures within the hardware component.

[0244] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A multi-objective optimization method for intelligent customer service outbound calls in the power grid based on the AGE-MOEA-II algorithm, characterized in that, It includes: Data processing step: preprocess historical outbound call data to generate an initial solution set; Objective function step: define the objective function of the AGE-MOEA-II algorithm, and iterate the initial solution set to obtain candidate solutions; Optimization search step: perform multi-objective optimization search through the AGE-MOEA-II algorithm, screen non-dominated solutions from candidate solutions, and determine the optimal outbound call strategy; Constraint handling step: handle the constraint conditions in the optimal outbound call strategy; Model output step: generate and output the final outbound call strategy for the power grid intelligent customer service.

2. The multi-objective optimization method for intelligent customer service outbound calls of power grid based on the AGE-MOEA-II algorithm according to claim 1, wherein The historical outbound call data includes work order information, outbound call time, and connection status; The preprocessing includes data cleaning and normalization processing, and key features are extracted through feature engineering; The work order information includes work order ID and customer type; In the data processing step, historical outbound call data is collected at the minimum interval between work orders to form outbound call data records; The output of the outbound call strategy for the power grid intelligent customer service includes optimized outbound call time slots, outbound call frequencies, connection rates, labor costs, and transfer rates.

3. The multi-objective optimization method for outbound calls of the power grid intelligent customer service based on the AGE-MOEA-II algorithm according to claim 2, characterized in that, The objective function f(x) optimizes the connection rate f1, labor cost f2(x), and transfer rate f3(x) by modeling the geometric shape of the Pareto front, maximizes f1(x), minimizes f2(x) and f3(x), and f(x) = Minimize{-f1(x), f2(x), f3(x)}; The connection rate is calculated based on historical outbound call data, base_rate i = exp(-a j ·(t i - μ j1 ) 2 ); Among them, priority_weight i is the priority weight of work order i; t i is the outbound call time for the i-th work order; μ j1 is the timestamp with the highest connection rate within the j-th time period; a j is the attenuation coefficient in the j-th time period; N represents the total number of work orders; The said labor cost base_cost i = C j + b j ·|t i - μ j2 |); where priority_factor i is the priority influence coefficient of work order i; base_cost i It is the basic labor cost for work order i; C j is the basic cost within the j-th time period; b j is the coefficient related to the time deviation degree in the j-th time period; μ j2 is the timestamp with the lowest labor cost in the j-th time period; The transfer rate Among them, priority_effect i and base_prob i are both the transfer rate priority influence coefficients of work order i; D j is the basic transfer rate within the j-th time period; μ j3 is the timestamp with the lowest manual transfer rate in the j-th time period k j1 and k j2 are the change range of the handover rate and the steepness of the change in the handover rate respectively in the j-th time period.

4. The multi-objective optimization method for outbound calls of the power grid intelligent customer service based on the AGE-MOEA-II algorithm according to claim 2, wherein, In the optimization search step, it is optimized by modeling the geometric shape of the Pareto front, the Pareto front is modeled through an adaptive geometric estimation method, and candidate solutions are classified through non-dominated sorting, and all candidate solutions are stratified; Use the Newton-Raphson iteration formula with the initial value set to p0 = 1, Iterate until the error is less than the threshold; Among them, A and B are two points in the solution set; a1,…a M 、b,…b M respectively represent the M-dimensional coordinates of points A and B; γ is the set of all smooth curves from A to B; |γ ′ (t)| is the velocity of the curve; gd(A, B) is the geodesic distance between points A and B; C ⊥ is the projection of the midpoint C of the straight line between points A and B; ||·||2 represents the Euclidean distance; p represents the geometric parameter of the curvature of the Pareto front; a i is the value of the non-dominated solution on the i-th objective; M is the number of target points; p n+1 represents the curvature parameter value obtained from the (n + 1)-th iterative calculation; p n Represents the curvature parameter value at the n-th iteration.

5. The multi-objective optimization method for outbound calls of the power grid intelligent customer service based on the AGE-MOEA-II algorithm according to claim 2, wherein The constraint conditions include the minimum interval between work orders and preferential outbound calls for preferential work orders; The minimum interval between work orders is determined according to the work order priority and outbound call time, and is embedded in the optimization model in the form of a non-linear constraint; The preferential outbound calls for preferential work orders are processed by embedding the corresponding priority rules in the objective function; In the constraint handling step, the feasibility of the solution set is judged, and the search direction of the solution is adjusted or a penalty term is added to the objective function for correction. Among them, f(x) is the objective function; g i (x) is the i-th constraint function; m is the total number of constraint functions; λ is the penalty factor; In the model output step, the relative importance weights of the target values of the connection rate, labor cost, and transfer rate are defined according to actual needs, the weighted target values are calculated, and the non-dominated solution with the optimal target value is selected through comprehensive evaluation to output the final outbound call strategy for the power grid intelligent customer service.

6. A multi-objective optimization system for intelligent customer service outbound calls in the power grid based on the AGE-MOEA-II algorithm, characterized in that, It includes a data processing module, a model construction module, and an outbound call strategy optimization module; The data processing module preprocesses historical outbound call data to generate an initial solution set; The model construction module defines the objective function of the AGE-MOEA-II algorithm, and iterates the initial solution set to obtain candidate solutions; The outbound call strategy optimization module performs multi-objective optimization search through the AGE-MOEA-II algorithm, screens non-dominated solutions from candidate solutions, and determines the optimal outbound call strategy; The model construction module processes the constraints in the optimal outbound calling strategy, generates the final outbound calling strategy for the power grid intelligent customer service, and outputs it.

7. The grid intelligent customer service outbound multi-objective optimization system based on the AGE-MOEA-II algorithm according to claim 6, wherein The historical outbound calling data includes work order information, outbound calling time, and connection status; The preprocessing includes data cleaning and normalization processing, and extracts key features through feature engineering; The work order information includes work order ID and customer type; In the data processing module, historical outbound calling data is collected at the minimum interval between work orders to form outbound calling data records; The output of the outbound calling strategy for the power grid intelligent customer service includes the optimized outbound calling time period, outbound calling frequency, connection rate, labor cost, and transfer rate.

8. The power grid intelligent customer service outbound multi-objective optimization system based on the AGE-MOEA-II algorithm according to claim 7, characterized in that, The objective function f(x) optimizes the connection rate f1, labor cost f2(x), and transfer rate f3(x) by modeling the geometric shape of the Pareto front, maximizes f1(x), minimizes f2(x) and f3(x), and f(x) = Minimize{-f1(x), f2(x), f3(x)}; The connection rate is calculated based on historical outbound call data, base_rate i = exp(-a j ·(t i - μ j1 ) 2 ); where priority_weight i is the priority weight of work order i; t i is the outbound call time for the i-th work order; μ j1 is the timestamp with the highest connection rate within the j-th time period; a j is the attenuation coefficient in the j-th time period; N represents the total number of work orders; The labor cost base_cost i = C j + b j ·|t i - μ j2 |); Among them, priority_factor i is the priority influence coefficient of work order i; base_cost i Is the basic labor cost for work order i; C j is the basic cost within the j-th time period; b j is the coefficient related to the degree of time deviation in the j-th time period; μ j2 is the timestamp with the lowest labor cost in the j-th time period; The transfer rate base_prob i = D j + k j1 · Among them, priority_effect i and base_prob i are both the transfer rate priority influence coefficients of work order i; D j is the basic transfer rate within the j-th time period; μ j3 is the timestamp with the lowest manual transfer rate in the j-th time period k j1 and k j2 are the change range of the control handover rate and the steepness of the change in the handover rate in the j-th time period, respectively.

9. The intelligent power grid customer service outbound multi-objective optimization system based on the AGE-MOEA-II algorithm according to claim 7, wherein, In the outbound calling strategy optimization module, it is optimized by modeling the geometric shape of the Pareto front, models the Pareto front through an adaptive geometric estimation method, classifies candidate solutions through non-dominated sorting, and stratifies all candidate solutions; Use the Newton-Raphson iteration formula with the initial value set to p0 = 1, and iterate until the error is less than the threshold value; Among them, A and B are two points in the solution set; a1,…a M 、b,…b M respectively represent the M-dimensional coordinates of points A and B; γ is the set of all smooth curves from A to B; |γ ′ (t)| is the velocity of the curve; gd(A, B) is the geodesic distance between points A and B; C ⊥ The projection of the midpoint C of the straight line between two points A and B; ||·||2 represents the Euclidean distance; p represents the geometric parameter of the curvature of the Pareto front; a i is the value of the non-dominated solution on the i-th objective; M is the number of target points; p n+1 represents the curvature parameter value obtained from the (n + 1)-th iterative calculation; p n Represents the curvature parameter value at the nth iteration.

10. The multi-objective optimization system for intelligent customer service outbound calls of power grid based on the AGE-MOEA-II algorithm according to claim 7, characterized in that The constraints include the minimum interval between work orders and preferential outbound calling for preferential work orders; The minimum interval between work orders is determined according to the work order priority and outbound calling time, and is embedded in the optimization model in the form of a non-linear constraint; The preferential outbound calling for preferential work orders is processed by embedding the corresponding priority rules in the objective function; In the model construction module, by judging the feasibility of the solution set, the search direction of the solution is adjusted or a penalty term is added to the objective function for correction. Among them, f(x) is the objective function; g i (x) is the i-th constraint function; m is the total number of constraint functions; λ is the penalty factor; Define the relative importance weights of the target values of the connection rate, labor cost, and transfer rate according to actual needs, calculate the weighted target values, select the non-dominated solution with the optimal target value through comprehensive evaluation, and output the final outbound calling strategy for the power grid intelligent customer service.

Images