Power-cut plan multi-objective dynamic optimization method and system based on hybrid intelligent algorithm

Abstract

The invention relates to the technical field of power grid dispatching, in particular to a power-cut plan multi-target dynamic optimization method and system based on a hybrid intelligent algorithm, and the method comprises the steps: obtaining the input data of a current power grid, and defining a schedulable time period set of equipment in a dispatching period in combination with the input data, continuously extracting time periods from the schedulable time period set, and generating a maximum schedulable power failure interval set of the equipment; based on the schedulable time period set and the maximum schedulable power failure interval, constructing all candidate power failure plan sets with scheduling feasibility, performing operation influence evaluation on the candidate power failure plan sets through an operation influence index function, and outputting an operation influence index set; constructing a non-dominated optimal plan set based on the candidate power-cut plan set and the operation influence index set; and based on the non-dominated optimal plan set and the scheduling priority score, generating a structured scheduling suggestion which can be submitted to the regulation and control system for execution.

Description

Technical Field

[0001] This invention relates to the field of power grid dispatching technology, and more particularly to a multi-objective dynamic optimization method and system for power outage plans based on hybrid intelligent algorithms. Background Technology

[0002] With the continuous expansion of the power grid, increasing load-side uncertainty, and frequent changes in operating modes due to the rapid integration of new energy sources, the formulation of power outage plans is facing unprecedented complexity. In receiving-end power grids, such as the Langfang power grid, due to weak power source support, complex power supply link structures, and critical nodes' high dependence on a single channel, equipment maintenance and infrastructure integration must be carried out through power outages. Furthermore, any power outage can lead to power flow redistribution, impact on critical users, or a decrease in grid redundancy. The formulation of power outage plans needs to simultaneously meet multiple conditions, including upper-level grid maintenance arrangements, restrictions on important users, control and operation modes, equipment-level safety constraints, and resource coordination capacity. These conditions often differ over time, resulting in dynamically changing outage windows. Existing technologies typically generate outage windows based on fixed rules or static feasibility models, then combine them using manual experience or simple heuristics. This approach struggles to address outage window failures caused by rapid changes in operating modes and cannot promptly identify structurally high-risk points in the combined plans.

[0003] Numerous real-world cases demonstrate that during months with a high concentration of maintenance shutdowns, previously deemed feasible outage combinations can suddenly become unfeasible due to conflicts in control resources, sudden user demands, or changes in operating modes. Dispatchers are forced to manually adjust the sequence of plans or cancel critical operations. Furthermore, the structural logic of inter-outage plans is complex, with implicit connections between lines, transformers, and power supply channels. Traditional static assessments cannot detect the sudden drop in backup path redundancy after disconnecting a particular device. In addition, the phenomenon of "unsolvable" outage plans due to drastic changes in operating modes frequently occurs, lacking a candidate plan generation and risk identification mechanism that can provide a unified assessment standard across different operating modes. Existing methods generally fail to provide structured risk explanations, making it difficult for dispatchers to determine whether candidate solutions truly meet the comprehensive requirements of system stability, control resources, and user needs.

[0004] Therefore, in the current power grid environment characterized by multiple constraints, high planning density, and complex structural relationships, there is an urgent need for a systematic technical solution that can build a bottom-level time window, generate candidate solutions, assess the combined impact, and make scheduling optimization decisions. This solution can achieve feasibility judgment for dynamic changes in operating modes, risk assessment for complex combined relationships, conflict identification for control resources, and structured output for scheduling execution, replacing the traditional approach that relies on human experience and static rules. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-objective dynamic optimization method and system for power outage planning based on hybrid intelligent algorithms, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A multi-objective dynamic optimization method for power outage planning based on hybrid intelligent algorithms includes:

[0008] The current power grid input data is obtained, and the set of schedulable time periods for the equipment within the scheduling cycle is defined in combination with the input data. Time periods are continuously extracted from the set of schedulable time periods to generate the maximum set of schedulable power outage intervals for the equipment. The set of schedulable time periods is used to determine the allowable power outage status of the equipment.

[0009] Based on the set of schedulable time periods and the maximum schedulable power outage interval, a set of all candidate power outage plans with scheduling feasibility is constructed. The set of candidate power outage plans is evaluated for operational impact using an operational impact index function, and a set of operational impact indices is output.

[0010] The set of operational impact indicators is input into the scheduling priority scoring model, which outputs a scheduling priority score. The scheduling priority score is then input into a multi-objective combined priority model. Guided by the scheduling priority score, the multi-objective combined priority model searches for and sorts the non-dominated solutions of the candidate power outage plan set, and outputs a set of non-dominated optimal plans. The scheduling priority scoring model includes a scheduling priority network, and the objective combined priority model includes an optimization objective function. The optimization objective function includes important user protection weights, resource conflict penalty coefficients, score offset penalty terms, sensitivity feedback coefficients, scheduling priority scores, and system operation sensitivity feedback terms.

[0011] The ranking value is calculated based on the scheduling priority score. The non-dominated optimal plan set is then sorted according to the ranking value. The sorted non-dominated optimal plan set is then divided into several scheduling task items. The comprehensive impact index of each scheduling item is calculated. The scheduling task items and comprehensive impact index are output in the form of structured scheduling suggestions.

[0012] Preferably, the maximum set of schedulable power outage intervals includes the start and end times of schedulable time periods, which are determined by linearly scanning an array of the set of schedulable time periods.

[0013] Preferably, the input data includes the power grid topology, operating mode level, scheduling time constraints, list of equipment to be inspected, and table of important user impacts of the equipment.

[0014] Preferably, the candidate power outage plan set includes candidate planned equipment, equipment outage time periods, number of equipment, and the intersection of all equipment outage time periods.

[0015] Preferably, the set of operational impact indicators includes redundancy indicators, important user coverage indicators, and resource conflict indicators.

[0016] Preferably, the scheduling priority network includes an input layer, a hidden layer, and an output layer.

[0017] Preferably, the operational impact index function includes the total number of power supply paths of the equipment in the power grid topology, the number of paths that still maintain power supply connectivity after the equipment is disconnected, the number of important users covered by the equipment, the total number of important users in the entire network, the maximum resource scheduling capacity, the structural redundancy penalty coefficient, the user impact weight coefficient, and the resource conflict penalty item.

[0018] Preferably, the scheduling task item includes the device number, power outage period, candidate plan set number, time window validity verification status field, sorting value, and risk interpretation field. The risk interpretation field includes the values ​​of important user coverage index and resource conflict index.

[0019] Preferably, the power grid topology includes equipment ID, start and end nodes, and associated substations; the operation mode levels include high load mode, dual power supply mode, and winter peak mode; the control time constraints include the time range of the superior dispatch plan, joint power outage restrictions for the same or adjacent equipment, and grid occupation restrictions for important projects; the list of equipment to be inspected includes the voltage level, equipment type, and associated substations; and the equipment important user impact table includes the power supply dependence relationship between the equipment and important users.

[0020] A multi-objective dynamic optimization system for power outage planning based on hybrid intelligent algorithms includes:

[0021] The scheduling time generation module is used to obtain the input data of the current power grid, combine the input data to define the set of schedulable time periods of the equipment within the scheduling cycle, continuously extract time periods from the set of schedulable time periods, and generate the maximum set of schedulable power outage intervals of the equipment. The set of schedulable time periods is used to determine the allowable power outage status of the equipment.

[0022] The scheduling plan generation and operation evaluation module is used to construct a set of all candidate power outage plans with scheduling feasibility based on the set of schedulable time periods and the maximum schedulable power outage interval. It evaluates the operation impact of the candidate power outage plan set through the operation impact index function and outputs the operation impact index set.

[0023] The scheduling plan orchestration and optimization module is used to input the set of operational impact indicators into the scheduling priority scoring model and output the scheduling priority score. Subsequently, the scheduling priority score is input into the multi-objective combined priority model. The multi-objective combined priority model searches for and sorts the candidate power outage plan set for non-dominated solutions based on the scheduling priority score and outputs the set of non-dominated optimal plans. The scheduling priority scoring model includes a scheduling priority network, and the objective combined priority model includes an optimization objective function. The optimization objective function includes important user protection weight, resource conflict penalty coefficient, score offset penalty term, sensitivity feedback coefficient, scheduling priority score, and system operation sensitivity feedback term.

[0024] The scheduling suggestion output module is used to calculate the ranking value based on the scheduling priority score, sort the set of non-dominated optimal plans according to the ranking value, split the set of non-dominated optimal plans into several scheduling task items, calculate the comprehensive impact index of each scheduling item, and output the scheduling task items and comprehensive impact index in the form of structured scheduling suggestions.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] This invention constructs a device-level dynamic outage window model based on power grid topology, operating mode, control constraints, and important user windows. This model enables hourly identification of the dispatchable boundaries of each device, ensuring that all subsequent plan generation is based on the actual dispatchable intervals. This solves the problem that traditional static window models are difficult to adapt to changes in actual operating modes. Furthermore, by jointly enumerating the dispatchable intervals of each device, a combination of candidate plans that can be actually executed is constructed. The operational impact of each candidate combination is quantified using power grid structural redundancy, the influence of important users, and control resource conflicts as core indicators, forming a combination feasibility identification mechanism oriented towards structural characteristics. Subsequently, by constructing a scheduling priority scoring network and a multi-objective optimization model, a hybrid intelligent orchestration strategy integrating scheduling experience and operational evaluation capabilities is formed. This allows the optimization process to dynamically balance multiple objectives, and the executability of the final scheduling results under different operating modes is improved by modeling the scoring offset and mode sensitivity. Finally, this invention transforms the optimized non-dominated optimal plan combination into a structured scheduling suggestion that can be directly accessed by the control system. This suggestion includes device operation items, time boundary criteria, scheduling ranking results, and impact explanation information, achieving seamless implementation from complex optimization results to engineering plans. This invention can systematically solve the problems of insufficient adaptability to dynamic operation modes, weak ability to identify combined risks, difficulty in coordinating multiple types of constraint conflicts, and unauditable planned output in existing technologies. It provides an interpretable, implementable, and adaptable intelligent scheduling method for power grid outage planning in highly complex operating environments. Attached Figure Description

[0027] Figure 1 A flowchart illustrating a multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm in the specific implementation of the invention;

[0028] Figure 2 This is a framework diagram of a multi-objective dynamic optimization system for power outage planning based on a hybrid intelligent algorithm, used in the specific implementation of the invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please refer to Figure 1 The present invention proposes a multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm, including:

[0031] Step 1: Obtain the current power grid input data, and define the set of dispatchable time periods for the equipment within the scheduling cycle based on the input data. Continuously extract time periods from this set to generate the maximum dispatchable outage interval set for the equipment. This set of dispatchable time periods is used to determine the permissible dispatchable outage status of the equipment, specifically including:

[0032] This step aims to determine the time period during which each piece of equipment under maintenance can be scheduled for power outage within the dispatch period, based on the current power grid topology, operating mode, upper and lower level dispatch plans provided by the power control system, and power supply constraints on the important user side. This outage window should fully reflect the actual constraints of engineering control, retaining only the truly available time set, which will serve as the boundary input for subsequent candidate plan combinations and dispatch optimization.

[0033] The input data includes the following aspects: power grid topology. This represents the connection relationships between all network devices and lines, exported from the control master station system (such as D5000). It is generally presented in topology table format (device ID, start and end nodes, substation to which it belongs), and the data collection cycle is hourly or triggered by switching operating modes. Operating mode level. Defined by dispatchers, this system reflects the power grid's operating mode during the current or expected dispatch period, such as high load mode, dual-source operation mode, and winter peak mode. It serves as an index to the mode library, linking to the section safety strategy model. This corresponds to a structured operational rules document. It also includes time constraint adjustments. This includes the time range of the superior scheduling plan exported from the planning system, joint power outage restrictions for the same or adjacent equipment, and grid occupation restrictions for important projects. This information is usually provided in structured plan records (plan number, restricted equipment ID, restricted time period). List of equipment to be inspected. Submitted by the maintenance unit or project owner, the list includes equipment ID, voltage level, equipment type (line, circuit breaker, busbar, etc.), and associated substation. This list is compiled by the maintenance management system or control cloud platform, with a daily time granularity. (Equipment Impact Table for Important Users) Provided by the power consumption information system or OMS, it maps the power supply dependency relationship between devices and important users, and provides the time window for important users to be allowed to have power outages. The data format is (device ID, user ID, permitted time period).

[0034] After considering the above information, define the device. The set of scheduleable time periods within the scheduling cycle that can be scheduled for power outages. :

[0035] ;

[0036] in, Indicates equipment At any moment Does the power outage meet the conditions for dispatchable operation? If so... This indicates that a power outage can be arranged. This indicates that it cannot be arranged. Indicates the breakability of the topology in the current operating mode. Next, disconnect the device. Whether it will cause structural breakage, islanding, or interruption of the main power supply path. This judgment is based on the topology connectivity matrix in the pre-built model and does not rely on real-time power flow simulation. This indicates a scheduling conflict check to determine the timing. Does it fall under the regulatory plan? The no-stopping zone in the text; if the equipment If a path has been designated as the primary supply path by its parent organization during a certain time period, then the function returns during that time period. . This indicates the user permission window's judgment, based on... The system determines the permitted time range for all users associated with the device, and only if all users allow access at time [time range missing]. Only during a power outage .

[0037] Set of schedulable time periods Calculations are performed at the hourly level, forming a binary time series. For example, equipment... During the first 10 days of December, power outages are permitted only on the 3rd and 7th days. The corresponding hours for these two days are The others are .from Extract continuous time periods and generate devices. Maximum set of dispatchable power outage sections The definition is as follows:

[0038] ;

[0039] in, It is equipment The maximum set of available consecutive scheduling time periods. and They represent the first The start and end times of a continuous schedulable time period. This partitioning process is performed through a linear scan. Array complete, find all values A continuous interval. For example, if the device of The period from hour 72 to hour 95 is continuous. Then generate This is recorded as a candidate power outage time period for that device. All devices ultimately form a set. This serves as an optional time boundary for subsequent scheduling plan combinations.

[0040] Step 2: Based on the set of schedulable time periods and the maximum schedulable outage interval, construct a set of all candidate outage plans with scheduling feasibility. Evaluate the operational impact of this set of candidate outage plans using an operational impact index function, and output a set of operational impact indices, specifically including:

[0041] This step involves the set of schedulable time periods for devices generated in the previous stage. Its maximum set of continuous stop time periods Based on this, all candidate outage plan combinations with scheduling feasibility are constructed, and a structured assessment of the operational impact of each combination is performed. This step is not only a crucial intermediate link connecting the equipment-level outage window with the final scheduling plan generation, but also an innovative interface that pushes the scheduling space from static input into the dynamic feasibility analysis process. Considering the complex grid operation mode, multi-level equipment correlation, and tight control resources in the scenario of this invention, a combined assessment mechanism that simultaneously considers grid structural redundancy, risk exposure of important users, and resource conflict degree is proposed, laying the foundation for a high-quality candidate set for the next step of multi-objective optimization.

[0042] The input includes two key variables from the previous stage's output, the first being... , indicating equipment At any moment Whether a power outage can be scheduled is determined by factors such as topology analysis, scheduling constraint identification, and important user permission windows, and serves as the time constraint boundary for candidate schedule combinations. Secondly... , indicating equipment The largest set of continuous schedulable time periods available for planning throughout the entire scheduling cycle, after eliminating all infeasible fragmented time periods, forms the core input domain of the combined plan.

[0043] The generation of candidate plans Based on this, a plan segment is formed by cross-combining equipment sets within a limited time window. To control the size of the combination, a single plan can contain a maximum of no more than [number missing] simultaneous items. Constraints of individual devices ( To set an upper limit (generally 3-5), and to combine them only when there is overlap in the scheduling time periods of all devices. Each set of candidate power outage plans... A candidate scheduling arrangement is defined as follows:

[0044] ;

[0045] in Indicates the first in the candidate plan One device Indicates device The selected power outage period This represents the number of devices in the combination. This represents the intersection of the power outage periods for all equipment, ensuring the unified scheduling feasibility of the combined plan. This design constraint considers the capability boundary of the scheduling system to perform synchronous control, avoiding the problem of logically feasible but unsynchronized scheduling in the combination.

[0046] To improve the quality of the plan and reduce the burden of subsequent optimization, it is necessary to analyze each candidate plan combination. An operational impact assessment is conducted. This step designs a set of structural indicators strongly correlated with the scenario of this invention to reflect the impact of the combined plan on security, power supply capacity, and resource coordination under the current power grid structure. One key innovation is the introduction of a structural redundancy suppression term, which integrates power grid topology and cross-sectional redundancy capabilities, unifying physical connection structure and risk space modeling into the combined scoring. The operational impact indicator function is defined as follows:

[0047] ;

[0048] in, For candidate plans The operational impact score is determined by the number of points; a higher score indicates greater risk and poorer schedulability, and can be used as part of the objective function or as a screening reference in subsequent optimization. Indicates device In topology The total number of power supply paths is obtained through a path search algorithm that starts at the main power supply substation and ends at the equipment power supply node. Indicates disconnecting the device The number of paths that maintain power supply connectivity is obtained by removing nodes from the topology graph and then recalculating the connectivity. Indicates device The number of important users covered is determined by The provided device-user dependency mapping relationship was obtained; The number of important users across the entire network; For equipment The number of conflicts in the resources (such as switch stations, maintenance teams, control channels, etc.) during the current time period. This value is extracted from the resource scheduling module of the control platform. This represents the maximum resource scheduling capacity and is used for normalization. , , These are the structural redundancy penalty coefficient, the user influence weight coefficient, and the resource conflict penalty term, respectively. The set values ​​can be obtained by adjusting the control strategy or fitting the training data.

[0049] The first term of this indicator function is the structural redundancy penalty term, which can approach 1 when path redundancy is extremely low, directly reflecting the risk of isolated points or path failures in the system after the device is disconnected. The second term is user coverage, which weighs the impact of the plan on the quality of power grid service. The third term is the resource conflict term, which is a model of the actual problems of conflicts arising under limited scheduling resources such as maintenance units and control command channels in the scenario of this invention, which is different from the traditional approach of only looking at the physical model.

[0050] The innovation of this function also lies in its strong interpretability and direct quantification; it can be embedded in subsequent optimization functions as a soft constraint, and can also be used as one of the conditions for eliminating candidate plans. When the value exceeds a certain control threshold, the system will automatically mark the combined plan as "low scheduling priority" or "unschedulable", thereby completing a pre-screening of the structural hierarchy before optimization and greatly reducing interference from invalid solutions.

[0051] Step 3: Input the set of operational impact indicators into the scheduling priority scoring model and output the scheduling priority score. Then, input the scheduling priority score into the multi-objective combined priority model. Guided by the scheduling priority score, the multi-objective combined priority model searches for and sorts the candidate power outage plans using non-dominated solutions, and outputs the set of non-dominated optimal plans. The scheduling priority scoring model includes a scheduling priority network, and the multi-objective combined priority model includes an optimization objective function. The optimization objective function includes important user protection weights, resource conflict penalty coefficients, score offset penalty terms, sensitivity feedback coefficients, scheduling priority scores, and system operation sensitivity feedback terms, specifically including:

[0052] This step builds upon the candidate plan set output in the previous stage. Its corresponding set of operational impact indicators Building upon this foundation, and addressing the task of "intelligent scheduling of plans under multiple constraints of source, grid, and load," an optimization module based on a hybrid intelligent algorithm is constructed. This module aims to select the optimal non-dominated combination from numerous feasible plans, ensuring controllable risk, strong coordination of regulatory resources, acceptable operational impact, and prioritization. This step is not only a crucial part of the core strategy for autonomous optimization of the scheduling system in this invention, but also forms the intelligent scheduling control center integrating scheduling experience, operational evaluation, and resource characteristics. In traditional methods, multi-objective scheduling optimization often relies on a unified scoring model or a single heuristic strategy, which cannot effectively coordinate the asymmetric distribution of various structural conflicts and resource constraints in the power grid. Especially in scenarios with high scheduling loads and intensive maintenance, situations prone to scheduling deviations, the exclusion of important tasks, or even system unsolvability can easily occur.

[0053] Input is and , It is the set of feasible plans formed by cross-combining time windows in the previous step, each Composed of several Composition, indicating equipment Scheduled in time slot power failure. It is a candidate plan The operational impact is quantified by metrics, including redundancy metrics. Key user coverage metrics Resource conflict index ,in , The scheduling range for equipment power outages is derived from the range established in step one. All combined plans are generated within the legal scheduling range, and there are no illegal scheduling behaviors.

[0054] This step is accomplished through two sub-models: a scheduling priority scoring model and a multi-objective combinatorial optimization model. The scheduling priority scoring model identifies combinatorial plans with higher scheduling value; the multi-objective combinatorial optimization model is responsible for searching and ranking non-dominated solutions under the guidance of the scoring. First, a scheduling priority scoring network is constructed. It is implemented using a structure-aware three-layer feedforward neural network, with the structure of an input layer (3 nodes) – a hidden layer (6 nodes, ReLU activation) – an output layer (1 node, linear). The input vector is... The output is a score. The model uses the actual execution plans adjusted in the first three months as training samples, and constructs labels by reverse engineering the actual execution success rate, scheduling smoothness, and frequency of subsequent plan adjustments, employing mean squared error loss for training. The innovation lies in introducing a structure regularization term during training. This is used to prevent scoring results from clustering towards resource-intensive plans, thus maintaining portfolio diversity and system flexibility.

[0055] After scoring, the model proceeds to a multi-objective combinatorial optimization model. This model uses a non-dominated sorting genetic algorithm (NSGA-II) for candidate selection. The optimization objective function is as follows: The design fully considers the three types of scheduling conflicts addressed by this invention: control resource constraints, power grid structure risks, and user-side limitations. It innovatively introduces a "system sensitivity feedback term" as an additional penalty component to enhance the system's ability to identify the feasibility of plans under extreme operating conditions in advance. The final objective function is designed as follows:

[0056] ;

[0057] in: , , , These are the important user protection weight, resource conflict penalty coefficient, score offset penalty item, and sensitivity feedback coefficient, which can be preset by the scheduling strategy. Assign a priority score to the candidate combination plan output by the scoring module. The highest score among all candidate plans; This is the system operation sensitivity feedback term, representing the current planned combination. The fluctuation intensity of operating indicators under different operating modes is defined as different The variance of the evaluation index is calculated (actually obtained in advance from the simulation results of the method library and tabulated).

[0058] Rating offset item As a regularization structure, the algorithm prioritizes high-value combinations with higher scores to prevent low-scoring plans from accidentally entering the optimal set in non-dominated search; the sensitivity feedback term prevents plan combinations from becoming infeasible after mode switching from the perspective of system stability, and builds the "anti-disturbance" judgment of plans in advance.

[0059] The optimization process employs the standard NSGA-II population evolution mechanism, performing crossover mutation, non-dominated sorting, and crowding distance sorting in each generation, retaining the previous generation. The non-dominated solutions constitute the optimal boundary set. After optimization, select... On the border of Middle Pareto The largest front The items form the final recommended set of execution plans.

[0060] Step 4: Calculate the ranking value based on the scheduling priority score, and sort the set of non-dominated optimal plans according to the ranking value. Then, break down the sorted set of non-dominated optimal plans into several scheduling tasks, calculate the comprehensive impact index for each scheduling item, and output the scheduling tasks and comprehensive impact index in the form of structured scheduling suggestions, specifically including:

[0061] This step is based on the set of non-dominated optimal plans output from the previous optimization result. and its corresponding scheduling priority scoring sequence This process generates structured scheduling suggestions that can be submitted to the control system for execution. This step is a crucial link in the entire invention, realizing the transition from "intelligent recommendation to engineering implementation." It receives the output from the intelligent optimization system and completes the systematization, standardization, and sequential transformation of the plan, ensuring that the control system can accurately understand, review, and execute the final scheduling instructions. Considering the high rigor required for the scheduling execution stage and its need to interface with the OMS and D5000 systems, this step designs a complete scheduling information structure template, integrating multiple types of information such as equipment attributes, time boundaries, scheduling risk information, and priority scores. The plan is then sorted and numbered through a scheduling logic sequence control mechanism.

[0062] The input consists of two main variables: first, the set of nondominated optimal plans. Each set of non-dominated optimal plans Composed of multiple Composition, indicating equipment In time period Internal maintenance power outages were arranged; each of them All are derived from the output of step one. This window is the sole source of information regarding the legitimacy of scheduling periods. The next step is to optimize the scoring sequence. It is through a structural scoring network to evaluate each The scoring system indicates the priority ranking of the combination based on factors such as structural risk, resource coordination, and impact on key users. Furthermore, this step requires backtracking. The evaluation indicators include and This is used to generate risk explanation fields in the output.

[0063] The generation of scheduling recommendations consists of three sub-processes. The first step is to sort the optimal combinations of schedules. Based on... The values ​​are sorted in descending order, with higher scores indicating a higher level of system recommendation and corresponding to earlier plan numbers. The sorted values ​​form the groups. Generate sort number This is used to identify the priority of scheduled execution. The sorting value is calculated using a normalized scoring method.

[0064] ;

[0065] This method ensures that the scheduling numbers are evenly distributed within the range of 0 to 1, facilitating the control platform to manage scheduling priority weights using floating-point numbers. Taking a real-world scenario as an example, if the system generates 10 optimal combinations within a certain scheduling cycle, and one of the plans... of The total score is Then the plan This value can be directly passed as the execution priority in the actual control and sorting interface.

[0066] The second step is to generate scheduling suggestion entries. Each It will be broken down into several scheduling task items, each corresponding to Each task item must include the following fields: (1) Device number The number comes from the D5000 master station equipment registration database; (2) power outage period (3) The combination number of the plan to which it belongs is recorded in the combination structure. (4) Time window validity verification The status field is used to mark whether the planned time period accurately hits the schedulable boundary; (5) scheduling sorting weight (6) Risk explanation field, including and The values ​​can be selected from all fields. All fields are output in standard JSON or CSV format via structured configuration, and can be automatically imported into the control plan management platform.

[0067] The third step is to generate a comprehensive impact index for scheduling entries. To improve the interpretability of the plan and enhance the assistance of manual approval, this step introduces an impact index. This is used to identify the degree of disruption the plan causes to the overall operation of the system.

[0068] ;

[0069] in, Indicates a combination plan The intensity of the impact of operation This indicates the extent of significant user impact involved in the plan's portfolio. This indicates the degree of resource conflict during scheduling. Weight and This can be set by regulators; for example, during peak summer periods when more attention is paid to the impact on key users, it can be... Set it to 0.7. The value is set to 0.3 to reflect the trade-offs of different scheduling strategies in plan selection.

[0070] Each generated entry will be output as a structured scheduling suggestion, containing not only information about the task operation itself, but also which optimal combination the task belongs to, the scheduling priority of that combination among all combinations, and why the task is recommended (through...). (Indicator Explanation) Does this task cause resource conflicts or user impact in the system? (via...) (Detailed field explanations in the document). This structured information will be uniformly formatted into a scheduling suggestion document and simultaneously generated into an electronic plan that interfaces with the standard interface of the control system. This allows for automatic import into the OMS or control cloud plan approval platform, supporting manual review and linkage with the plan execution system.

[0071] refer to Figure 2 As shown, in a second aspect of the present invention, a multi-objective dynamic optimization system for power outage planning based on a hybrid intelligent algorithm is proposed, comprising:

[0072] The scheduling time generation module is used to obtain the input data of the current power grid, combine the input data to define the set of schedulable time periods of the equipment within the scheduling cycle, continuously extract time periods from the set of schedulable time periods, and generate the maximum set of schedulable power outage intervals of the equipment. The set of schedulable time periods is used to determine the allowable power outage status of the equipment.

[0073] The scheduling plan generation and operation evaluation module is used to construct a set of all candidate power outage plans with scheduling feasibility based on the set of schedulable time periods and the maximum schedulable power outage interval. It evaluates the operation impact of the candidate power outage plan set through the operation impact index function and outputs the operation impact index set.

[0074] The scheduling plan orchestration and optimization module is used to input the set of operational impact indicators into the scheduling priority scoring model and output the scheduling priority score. Subsequently, the scheduling priority score is input into the multi-objective combined priority model. The multi-objective combined priority model searches for and sorts the candidate power outage plan set for non-dominated solutions based on the scheduling priority score and outputs the set of non-dominated optimal plans. The scheduling priority scoring model includes a scheduling priority network, and the objective combined priority model includes an optimization objective function. The optimization objective function includes important user protection weight, resource conflict penalty coefficient, score offset penalty term, sensitivity feedback coefficient, scheduling priority score, and system operation sensitivity feedback term.

[0075] The scheduling suggestion output module is used to calculate the ranking value based on the scheduling priority score, sort the set of non-dominated optimal plans according to the ranking value, split the set of non-dominated optimal plans into several scheduling task items, calculate the comprehensive impact index of each scheduling item, and output the scheduling task items and comprehensive impact index in the form of structured scheduling suggestions.

[0076] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A multi-objective dynamic optimization method for power outage planning based on hybrid intelligent algorithms, characterized in that, include: The current power grid input data is obtained, and the set of schedulable time periods for the equipment within the scheduling cycle is defined in combination with the input data. Time periods are continuously extracted from the set of schedulable time periods to generate the maximum set of schedulable power outage intervals for the equipment. The set of schedulable time periods is used to determine the allowable power outage status of the equipment. Based on the set of schedulable time periods and the maximum schedulable power outage interval, a set of all candidate power outage plans with scheduling feasibility is constructed. The set of candidate power outage plans is evaluated for operational impact using an operational impact index function, and a set of operational impact indices is output. The set of operational impact indicators is input into the scheduling priority scoring model, which outputs a scheduling priority score. The scheduling priority score is then input into a multi-objective combined priority model. Guided by the scheduling priority score, the multi-objective combined priority model searches for and sorts the candidate power outage plans using non-dominated solutions, and outputs a set of non-dominated optimal plans. The scheduling priority scoring model includes a scheduling priority network, and the objective combined priority model includes an optimization objective function. The optimization objective function includes important user protection weights, resource conflict penalty coefficients, score offset penalty terms, sensitivity feedback coefficients, scheduling priority scores, and system operation sensitivity feedback terms. The ranking value is calculated based on the scheduling priority score. The non-dominated optimal plan set is then sorted according to the ranking value. The sorted non-dominated optimal plan set is then divided into several scheduling task items. The comprehensive impact index of each scheduling item is calculated. The scheduling task items and comprehensive impact index are output in the form of structured scheduling suggestions.

2. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The maximum set of schedulable power outage intervals includes the start and end times of schedulable time periods, which are determined by linearly scanning an array of the set of schedulable time periods.

3. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The input data includes the power grid topology, operating mode level, scheduling time constraints, list of equipment to be inspected, and table of important user impacts of the equipment.

4. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The candidate power outage plan set includes candidate plan equipment, equipment outage time periods, number of equipment, and the intersection of all equipment outage time periods.

5. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The set of operational impact indicators includes redundancy indicators, important user coverage indicators, and resource conflict indicators.

6. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The scheduling priority network includes an input layer, a hidden layer, and an output layer.

7. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The operational impact index function includes the total number of power supply paths of the equipment in the power grid topology, the number of paths that still maintain power supply connectivity after the equipment is disconnected, the number of important users covered by the equipment, the total number of important users in the entire network, the maximum resource scheduling capacity, the structural redundancy penalty coefficient, the user impact weight coefficient, and the resource conflict penalty item.

8. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 1, characterized in that, The scheduling task item includes the device number, power outage period, candidate plan set number, time window validity verification status field, sorting value, and risk interpretation field. The risk interpretation field includes the values ​​of important user coverage index and resource conflict index.

9. The multi-objective dynamic optimization method for power outage planning based on a hybrid intelligent algorithm according to claim 3, characterized in that, The power grid topology includes equipment ID, start and end nodes, and associated substations. The operating mode levels include high load mode, dual power supply mode, and winter peak mode. The control time constraints include the time range of the superior dispatch plan, joint power outage restrictions for the same or adjacent equipment, and grid occupation restrictions for important projects. The list of equipment to be inspected includes the voltage level, equipment type, and associated substations. The table of important user impacts of the equipment includes the power supply dependence relationship between the equipment and important users.

10. A multi-objective dynamic optimization system for power outage planning based on a hybrid intelligent algorithm, characterized in that, include: The scheduling time generation module is used to obtain the input data of the current power grid, combine the input data to define the set of schedulable time periods of the equipment within the scheduling cycle, continuously extract time periods from the set of schedulable time periods, and generate the maximum set of schedulable power outage intervals of the equipment. The set of schedulable time periods is used to determine the allowable power outage status of the equipment. The scheduling plan generation and operation evaluation module is used to construct a set of all candidate power outage plans with scheduling feasibility based on the set of schedulable time periods and the maximum schedulable power outage interval. It evaluates the operation impact of the candidate power outage plan set through the operation impact index function and outputs the operation impact index set. The scheduling plan orchestration and optimization module is used to input the set of operational impact indicators into the scheduling priority scoring model and output the scheduling priority score. Subsequently, the scheduling priority score is input into the multi-objective combined priority model. The multi-objective combined priority model searches for and sorts the candidate power outage plan set for non-dominated solutions based on the scheduling priority score and outputs the set of non-dominated optimal plans. The scheduling priority scoring model includes a scheduling priority network, and the objective combined priority model includes an optimization objective function. The optimization objective function includes important user protection weight, resource conflict penalty coefficient, score offset penalty term, sensitivity feedback coefficient, scheduling priority score, and system operation sensitivity feedback term. The scheduling suggestion output module is used to calculate the ranking value based on the scheduling priority score, sort the set of non-dominated optimal plans according to the ranking value, split the set of non-dominated optimal plans into several scheduling task items, calculate the comprehensive impact index of each scheduling item, and output the scheduling task items and comprehensive impact index in the form of structured scheduling suggestions.

Images