CN-121983958-A - Large-model-driven power distribution network line loss accurate calculation and optimization method

CN121983958ACN 121983958 ACN121983958 ACN 121983958ACN-121983958-A

Abstract

The application discloses a large-model-driven accurate calculation and optimization method for line loss of a power distribution network, and relates to the technical field of power distribution networks, comprising the steps of collecting multi-source data of the power distribution network, cleaning, denoising and normalizing to construct a basic data set; the method comprises the steps of generating a structured feature set based on a corpus fine tuning model in the power distribution network field, extracting key features such as equipment parameters and load power, distinguishing technology and management line loss, extracting a multi-dimensional feature construction matrix, inputting the feature matrix into the large model, generating an accurate line loss value by fusing a tide calculation method, identifying key influencing factors through feature importance assessment, constructing a multi-objective optimization model, generating optimization strategies such as reactive compensation and network reconstruction, and outputting a final optimization scheme and an implementation path through large model reasoning verification and operation constraint verification. The method improves the calculation accuracy of the line loss of the power distribution network and the scientificity of an optimization strategy, can dynamically respond to the change of the running state, adapts to different scene requirements, and improves the power supply reliability and the energy utilization efficiency.

Inventors

YANG OU
HUANG BO
SUN ZHANGCAI
CHEN LU
ZHAO ZUCHENG
WU BO
LIU XIAOMIN
LI JINQIU
LI JUN
ZUO BING
ZHANG XIBO

Assignees

云南电网有限责任公司普洱供电局

Dates

Publication Date: 20260505
Application Date: 20260114

Claims (10)

1. The large-model-driven power distribution network line loss accurate calculation and optimization method is characterized by comprising the following steps of: s1, collecting multi-source data of a power distribution network, realizing synchronous extraction of heterogeneous data through a unified data interface, performing data cleaning, denoising and normalization processing, and constructing a line loss calculation basic data set; S2, inputting a basic data set based on a corpus fine tuning model in the power distribution network field, extracting equipment characteristics through named entity recognition and relation extraction, and generating a structural characteristic set; s3, distinguishing technical line loss and management line loss, extracting characteristics of transmission line loss, transformer loss and reactive loss from the technical line loss, extracting metering error and electricity stealing suspicion characteristics from the management line loss, and constructing a multidimensional characteristic matrix by combining space-time dimensions; s4, inputting the feature matrix into the trimmed large model, fusing an improved tide calculation method, establishing a line loss calculation model, and outputting accurate line loss values of each period, each line and each device; S5, identifying factors influencing line loss through feature importance evaluation of the large model, and generating an influence factor ordering report; s6, constructing a multi-objective optimization model based on influence factors, and generating optimization strategies including reactive power compensation configuration, network reconstruction, load transfer and equipment upgrading; And S7, verifying the feasibility and effectiveness of the optimization strategy through large model reasoning, and combining the operation constraint conditions of the power distribution network to verify, and outputting a final line loss optimization scheme and an implementation path.
2. The method for precisely calculating and optimizing the line loss of the power distribution network driven by the large model according to claim 1, further comprising a step of dynamically calculating the line loss by integrating the time sequence load data, wherein the calculation formula is as follows: Wherein the method comprises the steps of For the total technical line loss in the period T, T is the calculated period duration, i is the line or equipment number, n is the total number of the lines and equipment participating in the line loss calculation, For the instant of the current of the ith element at time t, And (3) for the rated resistance value of the ith element, the line loss accumulation effect under load fluctuation is characterized by integral operation of the product of square of time-series current and resistance.
3. The method for precisely calculating and optimizing the line loss of the distribution network driven by the large model according to claim 1, further comprising a load time sequence prediction smoothing step, wherein the load data precision is improved through integral smoothing and change rate correction, and the prediction formula is as follows: Wherein the method comprises the steps of For the predicted load power at time t, deltat is the smoothed time window length, And (3) for the actual load power monitoring value at the moment tau, k is a load change rate correction coefficient, monitoring noise is eliminated through integration and smoothing of historical load data, and the load change trend is corrected by combining the derivative of the integration result of the previous period.
4. The method for precisely calculating and optimizing the line loss of the distribution network driven by the large model according to claim 1, further comprising a multi-objective optimization integral objective function construction link, wherein the objective function is as follows: where F is the multi-objective optimization total target value, 、、 Respectively the line loss weight, the investment cost weight and the power supply reliability weight, and meets the following requirements + + =1, For the line loss power at time t, In order to optimize the total investment cost of the scheme, And (3) integrating the time sequence line loss and the power supply reliability index to realize multi-objective quantization balance and global optimization by integrating the time sequence line loss and the power supply reliability index for the time t with the power supply reliability coefficient ranging from 0 to 1 and T being the time length of the optimization evaluation period.
5. The large-model-driven power distribution network line loss accurate calculation and optimization method is characterized by further comprising a large-model field adaptation fine adjustment link, a special training data set is constructed based on technical standards, historical data and typical cases related to power distribution network line loss calculation, a LoRA low-rank self-adaptation fine adjustment mode is adopted, a large-model basic weight is frozen, only low-rank matrix parameters related to line loss calculation are trained, and expert rules in the line loss calculation field are introduced as constraint conditions to optimize feature extraction and reasoning logic of a model.
6. The method for precisely calculating and optimizing the line loss of the distribution network driven by the large model according to claim 1, further comprising a step of precisely calculating the management line loss, wherein the large model is used for analyzing operation data of metering equipment, electricity consumption behavior data of a user and historical line loss abnormal records, a metering error identification model and a electricity stealing behavior detection model are constructed, and the metering error identification model is used for quantifying error values of all metering points based on voltage-current phase differences, load curve consistency and metering equipment operation age characteristics.
7. The method for precisely calculating and optimizing the line loss of the distribution network driven by the large model according to claim 1, further comprising a line loss optimization strategy dynamic adjustment link, wherein the real-time collection of the operation data of the distribution network comprises load change, equipment state and meteorological conditions, the line loss calculation result and the influence factor sequencing are updated in real time through the large model, when the influence factors of the key change remarkably or the implementation effect of the optimization strategy is not expected, an optimization strategy regeneration flow is automatically triggered, and reactive compensation configuration parameters, network reconstruction paths and load transfer schemes are adjusted based on the latest operation data.
8. The method for precisely calculating and optimizing the line loss of the distribution network driven by the large model according to claim 1, further comprising a topological structure line loss sensitivity analysis link, wherein the step of simulating different topological structure adjustment schemes through the large model comprises line addition, tie switch addition and net rack reconstruction, and the line loss variable quantity corresponding to each adjustment scheme is calculated to construct a topological structure and line loss sensitivity matrix.
9. The method for precisely calculating and optimizing the line loss of the distribution network driven by the large model according to claim 1, further comprising a reactive compensation configuration optimization refinement step, wherein based on reactive load distribution data of each time period and each node output by the large model, a reactive compensation configuration strategy of division, time division and grading is adopted in combination with a line loss dynamic calculation result, a fixed reactive compensation device is configured at a node with concentrated reactive load and high line loss sensitivity, a dynamic reactive compensation device is configured at a node with large load fluctuation, and the switching time and compensation capacity of the dynamic reactive compensation device are optimized through integral operation, so that reactive power of each time period is balanced in situ.
10. The method for precisely calculating and optimizing the line loss of the power distribution network driven by the large model according to claim 1, further comprising a step of predicting and evaluating the implementation effect of the optimization scheme, wherein the digital twin model of the power distribution network is constructed through the large model, the optimization scheme is input into the digital twin model for simulation operation, and key indexes after implementation are predicted.

Description

Large-model-driven power distribution network line loss accurate calculation and optimization method Technical Field The invention relates to the technical field of power distribution networks, in particular to a large-model-driven power distribution network line loss accurate calculation and optimization method. Background The power distribution network is used as a key link for connecting a power supply and a user of the power system, and the line loss level of the power distribution network is directly related to the energy utilization efficiency, the power supply cost and the power supply reliability. Along with the rapid growth of diversified loads such as large-scale grid connection of renewable energy sources and charging piles of electric vehicles, the topology structure of a power distribution network is increasingly complex, load fluctuation presents strong randomness and space-time imbalance, and the traditional line loss calculation and optimization method faces serious challenges. The line loss is divided into technical line loss and management line loss, wherein the technical line loss relates to inherent loss of equipment such as a power transmission line, a transformer and the like, the management line loss is related to factors such as metering error, electricity stealing and the like, and accurate splitting and accounting of the technical line loss and the management line loss become core preconditions of line loss management. The traditional method is based on tide calculation and empirical formula estimation, relies on manual input equipment parameters and operation data, is complex in flow and low in efficiency, and is easy to cause larger calculation errors due to data isomerism and untimely update of topology change. The line loss calculation scheme based on the simple algorithm is difficult to effectively fuse dynamic factors such as weather, load time sequence change and the like, and line loss distribution characteristics of different time periods and different areas cannot be accurately captured. The verification of the management line loss is lack of effective means, only rough estimation can be carried out through statistical analysis, and specific loss sources such as metering errors, electricity larceny and the like are difficult to identify, so that the line loss management pertinence is insufficient. With the application of artificial intelligence technology in the electric power field, some schemes attempt to introduce a machine learning model to assist line loss calculation, but still have significant drawbacks. The existing models are mostly single task models, complex association features among equipment parameters, topological relations and line losses are difficult to accurately extract without deep adaptation in the field of power distribution networks, and optimization strategies generate multi-dimensional requirements such as multi-dimensional requirements for line loss reduction, investment cost, power supply reliability and the like, which are not balanced, so that feasibility and economical efficiency of an optimization scheme are unbalanced. Meanwhile, the existing method lacks real-time response capability to factors such as dynamic load, equipment aging and the like, an optimization strategy is difficult to adapt to dynamic changes of the running state of the power distribution network, a closed-loop treatment system for calculation, analysis, optimization and verification cannot be formed, and the intelligent level and actual effect of the line loss treatment of the power distribution network are restricted. Disclosure of Invention The invention provides a large-model-driven accurate calculation and optimization method for the line loss of a power distribution network, which aims to solve the problems in the prior art. In order to achieve the purpose, the invention adopts the following technical scheme that the large-model-driven power distribution network line loss accurate calculation and optimization method comprises the following steps: S1, acquiring multi-source data of a power distribution network, wherein the multi-source data comprises equipment parameters, operation measurement data, topological structure data and meteorological data, synchronous extraction of heterogeneous data is realized through a unified data interface, data cleaning, denoising and normalization processing are carried out, and a line loss calculation basic data set is constructed; S2, inputting a basic data set based on a corpus fine tuning model in the power distribution network field, extracting key characteristics such as resistance reactance, load power, topological connection relation and voltage level of equipment through named entity recognition and relation extraction, and generating a structural characteristic set; s3, line loss sub-term feature extraction steps are used for distinguishing technical line loss and management line loss, sub-term features such as power tran