CN-121998385-A - Power-preserving resource optimization scheduling method for power distribution network in disaster weather

Abstract

The invention relates to the technical field of power distribution network resource scheduling and discloses a power distribution network electricity-retaining resource optimal scheduling method under disaster weather, which comprises the steps of obtaining real-time disaster forecast data and equipment topology data and generating a disaster dynamic characteristic sequence with space-time labels; the method comprises the steps of inputting sequences into a space-time attention neural network to obtain fault probability distribution of each grid area, searching historical similar disaster scenes, extracting actual fault rules, quantitatively fusing the two to generate a dynamic fault risk level map, constructing an optimization function with minimized power failure time as a target by combining real-time states of resources based on the map, introducing urgent penalty factors according to real-time change rates of the risk levels, obtaining a resource throwing scheme of a multi-time window by adopting iterative optimization solution, converting the scheme into a scheduling instruction, and issuing the scheduling instruction, and carrying out closed-loop feedback with disaster evolution by real-time tracking. The invention realizes real-time adaptation of resource scheduling and disaster evolution, effectively solves the problem of resource mismatching, and shortens the power failure time and range to the maximum extent.

Inventors

• ZHANG YING
• SUN CHANGWEN
• WU MINGFENG
• WU XIAOJUN
• ZHANG GUANGWEI
• YANG JIEYUN
• ZHU YING
• SHEN DONGMING
• WU MING
• XU BINGYAN
• GUO LEI
• YAO WEI
• ZHAO YINGYING
• GAO SONGYAO

Assignees

• 国网山西省电力有限公司太原供电分公司
• 国网山西省电力有限公司电力科学研究院
• 国网上海市电力公司金山供电公司
• 华东电力试验研究院有限公司

Dates

Publication Date

20260508

Application Date

20260409

Claims (10)

1. 1. The power-preserving resource optimization scheduling method for the power distribution network in disaster weather is characterized by comprising the following steps of: Acquiring real-time disaster forecast data of a meteorological monitoring system and equipment topology data of a power grid geographic information system, fusing a plurality of dimensions of disaster intensity, influence range, duration and development trend, and generating a disaster dynamic characteristic sequence with a time stamp and space coordinates; Inputting the disaster dynamic characteristic sequence into a neural network based on a space-time attention mechanism for processing, and calculating to obtain the fault probability distribution of each grid area of the power distribution network in a future preset period; Searching matched historical similar disaster scenes from a historical fault database based on the disaster dynamic characteristic sequence, and extracting actual fault distribution rules corresponding to the historical similar disaster scenes; quantitatively fusing the fault probability distribution and the actual fault distribution rule to generate a dynamic fault risk level map of each line and each station of the power distribution network in a future period; Constructing a resource scheduling optimization function aiming at minimizing power failure time by taking a dynamic fault risk level map as a basis and combining real-time adjustable state information of a repair team, an emergency power generation car and material reserves, wherein the resource scheduling optimization function dynamically introduces a time urgency penalty factor according to the real-time change rate of the fault risk level of each region so that the region with the rapidly-rising risk level obtains higher resource allocation priority; solving a resource scheduling optimization function in an iterative optimization mode, and simulating a resource allocation scheme for a plurality of times to approach the lowest power failure time, so as to finally obtain an optimal scheduling scheme aiming at target throwing positions and action sequences of each rush repair team, emergency power generation car and materials in a plurality of time windows in the future; And converting the optimal scheduling scheme into a scheduling instruction, transmitting the scheduling instruction to a corresponding mobile terminal and resource management system, and tracking the instruction execution state and disaster evolution condition in real time so as to perform closed-loop feedback.
2. 2. The method for optimizing and scheduling electricity-retaining resources of power distribution network in disaster weather according to claim 1, wherein the disaster dynamic feature sequence comprises the following steps: Extracting a core index representing disaster intensity, boundary coordinates of a forecast influence area, a predicted duration and a forecast path or evolution direction from disaster forecast data issued by a meteorological monitoring system; And splicing according to time sequence to form a characteristic vector sequence comprising a time stamp, a grid number, an intensity normalization value, an influence range mark, a duration, an evolution direction angle, a moving speed, a line type code, a tower density normalization value and a commissioning period normalization value.
3. 3. The method for optimizing and scheduling electricity-retaining resources of power distribution network in disaster weather according to claim 1, wherein the neural network based on the space-time attention mechanism comprises an input layer, a space attention layer, a time attention layer and an output layer, wherein: The input layer receives a disaster dynamic characteristic sequence; the spatial attention layer calculates the spatial attention weight between each grid region and all other grid regions, and is calculated according to the Euclidean distance and the geographic distance of the feature vectors of the two regions through a softmax function and used for capturing the mutual conduction relation of disaster influences among different regions; The time attention layer calculates time attention weights between each time window and a plurality of time windows in front and back, is calculated through a softmax function according to the change rate of the feature vector along with time and is used for capturing the evolution trend of disaster intensity; And multiplying the space attention weight and the time attention weight by original features respectively, adding to obtain enhanced feature representation, and inputting the enhanced feature representation to an output layer to calculate the fault probability of each grid region.
4. 4. The method for optimizing and scheduling electricity-retaining resources of power distribution network in disaster weather according to claim 1, wherein the method is characterized by retrieving matched historical similar disaster scenes from a historical failure database based on a disaster dynamic characteristic sequence, and specifically comprises the following steps: Calculating cosine similarity between the current disaster dynamic feature sequence and the feature sequence of each historical disaster record in the historical fault database; sorting cosine similarity according to the sequence from big to small, and selecting a history disaster record with the top three similarity ranks as a matched history similar disaster scene; and extracting the actually-occurring line fault positions, fault types and fault numbers in the three historical disaster records within 24 hours after the occurrence of the disasters, and taking the actually-occurring line fault positions, fault types and fault numbers as actual fault distribution rules.
5. 5. The method for optimizing and scheduling electricity-retaining resources of the power distribution network in disaster weather according to claim 1, wherein the method for generating the dynamic fault risk level map is characterized by comprising the following steps: Sequencing all grid areas from large to small according to the values of the fault probability distribution to obtain the fault probability distribution value corresponding to each grid area; Sequencing all grid areas from large to small according to the total times of faults in each historical similar disaster scene to obtain a historical experience risk value corresponding to each grid area; Adding the fault probability distribution value of each grid region with the historical experience risk value to obtain a comprehensive risk value of the grid region; Marking lines and areas corresponding to the grid areas with the top 10% of the ranks as high risk levels, marking lines with the ranks between 10% and 30% as medium risk levels, and marking 70% of the ranked lines as low risk levels according to the order of the comprehensive risk values from large to small; And rendering the high-risk level region in red, the medium-risk level region in yellow and the low-risk level region in green on a map of the geographic information system of the power distribution network to form a visual dynamic fault risk level map.
6. 6. The method for optimizing and scheduling electricity-retaining resources of power distribution network in disaster weather according to claim 1, wherein the real-time adjustable state information of the rush-repair team, the emergency power generation car and the material reserve comprises the following steps: The method comprises the steps of acquiring current longitude and latitude coordinates of each repair team and each emergency power generation car in real time through a vehicle-mounted global positioning system terminal, acquiring the warehouse position, the material type and the available quantity of each batch of materials currently located through an electronic ledger of a resource management system, associating the position information with a dynamic fault risk level map by adopting the same grid numbering system, and establishing a mapping relation between the resource position and a risk area.
7. 7. The method for optimizing and scheduling electricity-saving resources of power distribution network in disaster weather according to claim 1, wherein constructing a resource scheduling optimization function targeting minimizing power outage time comprises: The constraint condition of the resource scheduling optimization function comprises that each repair team can only send to one fault area at the same time, each emergency power generation car can only support one platform area at the same time, and the quantity of materials called by each warehouse does not exceed the available stock quantity of the warehouse.
8. 8. The method for optimizing and scheduling electricity-retaining resources of a power distribution network in disaster weather according to claim 1, wherein the resource scheduling optimization function dynamically introduces a time urgency penalty factor according to the real-time change rate of the fault risk level of each region, and the method comprises the following steps: Periodically calculating the comprehensive risk value of each grid area according to a preset time interval, and recording the risk value variation of the adjacent time interval; When the variation of the risk value of a certain grid area in a plurality of continuous time intervals is positively increased, the accumulated increase amplitude exceeds 10% of the original comprehensive risk value, and the area is marked as a risk acceleration deterioration area; When resources are allocated to the risk acceleration deterioration area, an urgency multiplier larger than 1 is introduced into the resource demand calculation, the specific value is positively correlated with the accumulated increase amplitude of the risk value, and the urgency multiplier takes a higher value as the increase amplitude is larger; By introducing an urgency multiplier, the optimization function tends to preferentially allocate available resources to risk acceleration exacerbation areas when solving.
9. 9. The method for optimizing and scheduling electricity-retaining resources of the power distribution network in disaster weather according to claim 1, wherein solving the resource scheduling optimization function by adopting an iterative optimization mode comprises the following steps: Firstly, generating a group of initial solutions, namely, randomly distributing a target area for each rush-repair team and each emergency power generation car; Trimming the current solution, exchanging target areas of two teams or changing the throwing place of a certain material, and calculating the weighted power failure time corresponding to the new solution; If the power failure time of the new solution is lower than that of the original solution, the new solution is accepted, otherwise, the new solution is accepted with a certain probability; repeating the fine tuning and comparing processes until the power failure time is not reduced after the continuous multiple iterations, and outputting the current optimal solution as a final scheduling scheme.
10. 10. The method for optimizing and scheduling electricity-saving resources of power distribution network in disaster weather according to claim 1, wherein the method for tracking instruction execution state and disaster evolution condition in real time for closed-loop feedback comprises the following steps: monitoring whether each scheduling instruction is executed or not through the confirmation information uploaded by the mobile terminal, and if the execution confirmation is not received within 30 minutes after a certain instruction is sent out, marking the state of the resource as abnormal and re-participating in the next round of scheduling calculation; And re-acquiring the latest meteorological monitoring data every 10 minutes, comparing the newly generated optimal scheduling scheme with the previous round of scheme, and only sending a changed instruction to the resource terminal to keep the continuous execution of the unchanged instruction.

Description

Power-preserving resource optimization scheduling method for power distribution network in disaster weather Technical Field The invention relates to the technical field of power distribution network resource scheduling, in particular to a power distribution network electricity-retaining resource optimal scheduling method in disaster weather. Background The power distribution network bears vital responsibility in guaranteeing the living, industrial and agricultural production of residents and reliable electricity utilization of users, and extreme meteorological events such as typhoons, storm and ice disasters form serious threats to power distribution network facilities, so that large-scale power failure accidents are often caused. How to quickly restore power supply under the extreme condition, reduce the power outage range as much as possible, shorten the power outage time, and optimize and schedule the power-saving resources of the power distribution network in disaster weather has become one of the most focused core research subjects in the field of emergency management of power systems. Currently, the electricity-saving resource scheduling aiming at disaster weather mainly depends on two modes, namely a fixed emergency plan formulated based on historical experience and artificial resource allocation by scheduling personnel according to field feedback. These methods can play a role in conventional fault handling and predictable overhaul scenarios, but tend to cope with debilitation when faced with real disaster impact. The disaster is caused by the disaster, and the occurrence and development of the disaster have high uncertainty, namely, the disasters with different types, different intensities and different moving paths have huge differences in damage modes and fault distribution characteristics on the power distribution network. Typhoons can cause wide-range pole falling and line breakage, rainstorm can cause waterlogging to cause equipment to dip, and ice disaster can cause line icing to swing. The existing fixed plan or manual scheduling mode is difficult to quickly and accurately adjust according to the real-time evolution situation of disasters, the phenomenon that resources are put in and seriously misplaced with actual demands often occurs, and the contradiction that the emergency repair force of the disaster area is insufficient and the resources of the edge area are idle sometimes occurs. Analysis in greater depth may reveal that the disaster itself is characterized by a complex system of dynamic changes in multiple dimensions. The disaster intensity level directly determines the severity of equipment damage, the influence range determines the number of lines and areas to be covered, the duration determines the urgent windows of fault repair and power restoration, and the development trend further affects the future trend of the three dimensions. These four dimensions are interrelated and dynamically evolve, which together determine the true scale and priority of electricity conservation resource requirements. If these dimensions cannot be comprehensively considered and dynamically quantized as an organic whole, it cannot be accurately judged how many rush repair teams, how many emergency power generation cars and how many material reserves are needed at a certain moment and in a certain area. If the scheduling mechanism is still running in an initial large-scale average allocation, resource mismatch dilemma must occur. The problem of resource demand judgment misalignment caused by insufficient identification and real-time tracking of dynamic association relations among disaster multidimensional features has become the most prominent technical bottleneck in the current electricity-retaining scheduling field. Disclosure of Invention Therefore, the invention aims to overcome the defects that the resource scheduling is carried out by relying on a fixed plan or artificial experience in the prior art, and cannot be adjusted in real time according to the dynamic changes of multidimensional characteristics such as disaster intensity, range, duration, development trend and the like, so that the resource delivery and actual fault requirements are seriously misplaced. In order to solve the technical problems, the invention provides a power-saving resource optimization scheduling method for a power distribution network in disaster weather, which comprises the following steps: Acquiring real-time disaster forecast data of a meteorological monitoring system and equipment topology data of a power grid geographic information system, fusing a plurality of dimensions of disaster intensity, influence range, duration and development trend, and generating a disaster dynamic characteristic sequence with a time stamp and space coordinates; Inputting the disaster dynamic characteristic sequence into a neural network based on a space-time attention mechanism for processing, and calculating to obtain the fault probabi