CN-122000818-A - Method, system and storage medium for cable arrangement in power supply and distribution system based on reinforcement learning

CN122000818ACN 122000818 ACN122000818 ACN 122000818ACN-122000818-A

Abstract

The invention discloses a cable arrangement method, a system and a storage medium in a power supply and distribution system based on reinforcement learning, wherein the method comprises the following steps of S1, obtaining information of a cable and an arrangement environment and constraints which need to be met by cable arrangement, S2, abstracting a bridge system into a weighted undirected topological graph, S3, calling a path planning algorithm based on reinforcement learning, calculating an optimal cable arrangement path by taking the minimum comprehensive cost function comprising path length cost, volume rate cost and expensive penalty items as an optimization target under the condition that the constraints which need to be met by cable arrangement are met, S4, calculating the central coordinate of each cable to be arranged in each bridge by adopting a greedy algorithm according to the optimal cable arrangement path, and accordingly arranging the cable.

Inventors

LIAO PENG
ZHAO SHUAI
CAO SHUN
JIANG TAO
ZHU ZAN
WANG BIN

Assignees

中国电子系统工程第二建设有限公司

Dates

Publication Date: 20260508
Application Date: 20251217

Claims (10)

1. The cable arrangement method in the power supply and distribution system based on reinforcement learning is characterized by comprising the following steps of: S1, acquiring cable related information, starting and end equipment selected by a user and constraints to be met by cable laying; S2, abstracting a bridge system into a weighted undirected topological graph, wherein nodes in the topological graph comprise starting and ending points of the bridge, bridge connecting pieces and electric connecting points of starting and ending point equipment, and edges are the bridge and carry the length, the available cross-sectional area and the occupied cross-sectional area information of the bridge; S3, invoking a path planning algorithm based on reinforcement learning, and calculating an optimal cable laying path by taking the minimum comprehensive cost function comprising path length cost, volume rate cost and expensive penalty term as an optimization target under the condition of meeting the constraint that cable laying needs to meet; And S4, calculating the center coordinates of each cable to be arranged in each bridge by adopting a greedy algorithm according to the optimal cable laying path, and accordingly arranging the cables.
2. The method of claim 1, wherein the cabling constraints include bridge volume rate constraints, cable bend radius constraints, cable voltage class isolation constraints, and cable strong and weak electrical isolation constraints.
3. The method of claim 1, wherein the expression of the composite cost function is , Wherein, the A comprehensive cost function for the P-th path; 、 And The sum of the weight coefficient and the three is 1; 、 And The length cost function, the volume rate cost function and the violation penalty function of the P-th path are respectively adopted; the expression of (2) is , Wherein, the For the ith bridge frame, the first bridge frame, Is the length of the ith bridge; the expression of (2) is , Wherein n is the number of bridges included in the path P, The i-th bridge has a sectional area, For the original available cross-sectional area of the ith bridge, The sectional area of the cable to be arranged is; the expression of (2) is 。
4. The method according to claim 3, wherein the step of And And substituting the normalized value into the comprehensive cost function for calculation.
5. The method of claim 1, wherein the reinforcement learning is modeled using Markov decisions, wherein the states are real-time capacity and length information of a current node and its associated bridge, the actions are moving from the current node to any adjacent node, and the rewards include target achievement rewards, single-step guided rewards, and negative offending rewards.
6. The method according to claim 5, wherein the target achievement rewards correspond to a functional expression of , Wherein, the A bonus function is achieved for the purpose of achieving, And The normalized values of the length cost function and the volume rate cost function of the P-th path are respectively obtained; the single step guided prize is expressed as a function of , Wherein, the The bonus function is directed for a single step, The bridge length selected for this step is chosen, And The bridge frame selected in the step occupies a cross section area and an original available cross section area respectively; and giving a preset negative value reward to any action against the constraint to be met by cabling, and ending the round of exploration.
7. The method of claim 2, wherein the bridge volume constraint is expressed as , Wherein, the Allowing the cable cross-section to be the maximum proportion of the original available cross-section for the ith bridge; the bending radius of the cable is constrained to be not smaller than the minimum bending radius for any turning node on the path; The cable voltage class isolation constraint is that cables with different voltage classes are required to be arranged in different bridges or separated in the same bridge by metal partition plates; the cable strong and weak electric isolation constraint is that the power cable and the information or control cable are required to be arranged in the bridges with different voltage classes.
8. The method of claim 1, wherein the greedy algorithm calculates the center coordinates of each cable to be routed in each bridge in a delta arrangement.
9. A reinforcement learning-based cable routing system in a power supply and distribution system, comprising: The information acquisition module is used for acquiring cable related information, and the initial equipment and the terminal equipment selected by a user and the constraint which needs to be met by cable laying; The topology construction module is used for abstracting a bridge system into a weighted undirected topology graph, wherein nodes in the topology graph comprise a bridge starting point, a bridge ending point, a bridge connecting piece and an electric connection point of starting and ending point equipment, and edges are the bridge and carry the length, the available cross section area and the occupied cross section area information of the bridge; The laying path planning module is used for calling a path planning algorithm based on reinforcement learning, and calculating an optimal cable laying path by taking the minimum comprehensive cost function comprising the path length cost, the volume rate cost and the expensive penalty term as an optimization target under the condition of meeting the constraint that the cable laying needs to meet; And the arrangement module is used for calculating the center coordinates of each cable to be arranged in each bridge frame by adopting a greedy algorithm according to the optimal cable laying path, so as to arrange the cables.
10. A computer readable storage medium storing one or more programs, comprising one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-8.

Description

Method, system and storage medium for cable arrangement in power supply and distribution system based on reinforcement learning Technical Field The invention relates to cable arrangement of a power supply and distribution system, in particular to a cable arrangement method, a system and a storage medium in the power supply and distribution system based on reinforcement learning. Background At present, in the field of building electrical design and construction, the distribution design of power supply and distribution system cables mainly depends on manual operation of designers in Building Information Model (BIM) software. This approach requires the designer to drag the cables one by one from the starting equipment to the end equipment in a complex three-dimensional space and manually specify the bridge path through which they traverse. The process is low in efficiency and high in working strength, and due to complexity of a three-dimensional space, a designer is difficult to optimize from a global view, and a finally obtained cable path is often only a 'connected' function, but is an 'optimal' path with both economy and safety. This directly results in an unnecessary increase in the amount of cable material used, increasing project costs and potential for subsequent operation and maintenance. In order to solve the problem of low manual efficiency, an automatic laying scheme based on an algorithm appears in the prior art. For example, chinese patent publication No. CN116720294a discloses an automatic cabling method based on BIM technology and local dynamic domain search algorithm. According to the method, the cable channel routing network is established by reading the power grid identification information, and the Floyd algorithm based on dynamic programming is adopted to automatically seek paths so as to realize optimal arrangement of cables. However, the prior art solution still has the obvious limitation that firstly, the optimization target is single, the path is mainly pursued to be shortest, and the key engineering constraint of 'bridge volume rate' is not dynamically and prospectively considered in the path optimization process. This tends to result in excessive concentration of the planned cable in some bridge segments, causing localized congestion, violating electrical design specifications, and difficulty in system heat dissipation and future capacity expansion. Secondly, the traditional graph search algorithm such as Floyd is essentially used for carrying out global optimal solution calculation aiming at a static network, is difficult to adapt to actual engineering scenes of sequential cable laying and dynamic change of bridge residual capacity, and lacks the capability of on-line adjustment and adaptation to complex dynamic environments. Disclosure of Invention The invention aims to provide a cable arrangement method, a system and a storage medium in a reinforced learning-based power supply and distribution system, which can plan optimal laying and arrangement paths on the premise of conforming to various physical constraints of cable laying and can be automatically and dynamically adapted to different laying scenes. The cable arrangement method in the power supply and distribution system based on reinforcement learning comprises the following steps: S1, acquiring cable related information, starting and end equipment selected by a user and constraints to be met by cable laying; S2, abstracting a bridge system into a weighted undirected topological graph, wherein nodes in the topological graph comprise starting and ending points of the bridge, bridge connecting pieces and electric connecting points of starting and ending point equipment, and edges are the bridge and carry the length, the available cross-sectional area and the occupied cross-sectional area information of the bridge; S3, invoking a path planning algorithm based on reinforcement learning, and calculating an optimal cable laying path by taking the minimum comprehensive cost function comprising path length cost, volume rate cost and expensive penalty term as an optimization target under the condition of meeting the constraint that cable laying needs to meet; And S4, calculating the center coordinates of each cable to be arranged in each bridge by adopting a greedy algorithm according to the optimal cable laying path, and accordingly arranging the cables. By sequentially executing the steps of information acquisition, topology abstraction and intelligent planning, the invention converts the discrete experience-dependent traditional design mode into a systematic data-driven intelligent decision process. The method comprises the steps of obtaining cable related information and constraint, definitely determining a design target and conditions to be met, abstracting a bridge system into a weighted undirected topological graph, converting a complex space geometric problem into a graph theory problem which can be processed by a computer, lay