CN-122026313-A - Source network load interactive user energy space-time characteristic analysis method adapting to differentiated scene

CN122026313ACN 122026313 ACN122026313 ACN 122026313ACN-122026313-A

Abstract

The invention relates to a source network load interactive user energy space-time characteristic analysis method adapting to a differential scene, which comprises the following steps of S100, obtaining multi-source basic data of a medium-low voltage distribution network by adopting a data acquisition technology, S200, constructing a multi-effect factor quantification system, S300, establishing a user energy trend prediction model integrating the multi-source basic data and the quantification system data, S400, obtaining a trained user energy trend prediction model, S500, reasoning source-network-load interactive data in different areas, different DG permeability, different load types and different time scales, generating a space-time distribution characteristic map of user energy, and finishing the accurate analysis of the source network load interactive user energy space-time characteristic under the differential scene. The method and the device can solve the technical problems of insufficient suitability, insufficient quantification of multiple influencing factors and lower prediction precision in the energy space-time characteristic analysis of the users of the medium-low voltage distribution network.

Inventors

QI YU
WANG TING
GUO ZHUOHUI
GUO WENXUAN
HAN WEIHENG
TAO WENBIAO
SHI YUXIN
LI RUI
WANG LIANG
ZHAO JIN
CHEN DANYANG
HAO WEI

Assignees

国网山西省电力有限公司电力科学研究院

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (8)

1. A source network load interactive user energy space-time characteristic analysis method adapting to a differential scene is characterized by comprising the following steps: Step S100, acquiring multi-source basic data of a medium-low voltage distribution network by adopting a data acquisition technology; Step S200, constructing a multi-influence factor quantification system according to the multi-source basic data acquired in the step S100 and combining regional characteristics, social factors and population density distribution; step S300, based on a convolutional neural network and a long-term and short-term memory network, establishing a user energy trend prediction model integrating multi-source basic data and quantitative system data; Step S400, training the user energy consumption trend prediction model established in the step S300 by adopting a training data set, and iteratively optimizing model parameters through a loss function to determine an optimal model structure so as to obtain a trained user energy consumption trend prediction model; And S500, utilizing the trained user energy trend prediction model in the step S400 to infer source-network-load interaction data in different areas, different DG permeabilities, different load types and different time scales, generating a space-time distribution characteristic map of user energy, and completing the accurate analysis of the space-time characteristics of the source network-load interaction type user energy in a differentiated scene.
2. The method for analyzing energy space-time characteristics of source network load interaction type users adapting to differentiated scenes according to claim 1, wherein the specific steps of the step S100 include: collecting historical meteorological data, wherein the historical meteorological data comprise irradiance, wind speed, temperature and humidity, the collecting period is not less than 1 year, and the sampling frequency is 5 minutes/time; collecting output data of a distributed power supply, wherein the distributed power supply comprises distributed photovoltaic, wind power and a micro gas turbine, the collecting period is not less than 1 year, and the sampling frequency is 5 minutes/time; collecting charging and discharging data of the electric automobile, wherein the charging power, the discharging power and the charging and discharging duration are included, and the collecting frequency is 5 minutes/time; the user electricity load data are collected and collected according to the classification of industrial users, commercial users and resident users, and the collection frequency is 5 minutes/time.
3. The method for analyzing energy space-time characteristics of source network load interaction type users adapting to differentiated scenes according to claim 1, wherein the specific steps of the step S200 include: (1) Carrying out standardized processing on historical meteorological data and distributed power supply output data by adopting principal component analysis, and extracting space-time core characteristics; The normalized formula is: ; Where x i is the raw data, μ is the mean of the data, Data standard deviation, x i ' is normalized data; PCA solves the eigenvalue lambda j and the eigenvector e j through a covariance matrix, and selects the cumulative variance contribution rate The first k principal components of (a) are used as space-time core features; (2) Respectively calculating the association degree of regional characteristics, social factors, population density distribution, electric vehicle charging and discharging data and user electricity load type distribution by adopting the spearman correlation coefficient; The spearman correlation coefficient formula is: ; where di is the rank difference of the ith sample over two variables and n is the number of samples; according to the space-time core characteristics and the association degree obtained by calculation in the step (1) and the step (2), the weight of each influence factor on the power load of the user is quantized through a binary Gauss Copula function, and a multi-influence factor quantization system is constructed; The Gauss Copula density function is: ; Wherein ρ is the Pearson correlation coefficient between the two variables, A standard normal distribution bit dividing function, u, v E (0, 1) is the cumulative probability of edge distribution; The factor weights W j were obtained by joint density normalization after Copula fitting.
4. The method for analyzing energy consumption space-time characteristics of the source network load interaction type user adapting to the differentiated scene according to claim 1, wherein the factors in the multi-effect factor quantification system of the step S200 comprise irradiance, distributed power supply output fluctuation, population density, people average GDP, time-of-use electricity price and intelligent electricity meter coverage rate.
5. The method for analyzing energy space-time characteristics of source network load interaction type users adapting to differentiated scenes according to claim 1, wherein the specific steps of the step S300 include: constructing a multi-layer convolution layer and a pooling layer through a convolution neural network, extracting spatial correlation characteristics of multi-source basic data and multi-influence factor quantization system data, and outputting spatial characteristic vectors; The convolution operation formula of the convolution neural network is as follows: ; Wherein, the Is the first The layer convolution kernel weights are used to determine, As a result of the bias term, K is the convolution kernel size, which is the ReLU activation function; Constructing a plurality of hidden layers through a long-short-term memory network, capturing long-short-term time sequence dependency relationship between multi-source basic data and quantitative system data, and outputting time sequence feature vectors; The updating formula of the LSTM is as follows: ; ; ; ; ; ; wherein f t is forget gate output, i t is input gate output, o t is output gate output, As candidate cell states, C t is a cell state, h t is a hidden layer output, W f 、W i 、W C 、W o is a weight matrix, b f 、b i 、b C 、b o is a bias term, x t is a current time input, and h t-1 is a hidden layer output at a previous time; The space feature vector and the time sequence feature vector are spliced and then input into a full connection layer, a complete framework of a user energy trend prediction model is built by combining an activation function, the construction of the user energy trend prediction model which fuses multi-source basic data and quantitative system data is further completed, and three types of predicted values are output, namely distributed power supply output Electric automobile payload General energy for users 。
6. The method for analyzing energy space-time characteristics of source network load interaction type users adapting to differentiated scenes according to claim 1, wherein the specific steps of the step S400 include: dividing the training data set into a training set, a verification set and a test set according to the proportion of 7:2:1; adopting average absolute percentage error as a loss function, and iteratively updating model parameters through a back propagation algorithm; The MAPE formula is: ; Wherein, the As a result of the fact that the value, N is the number of samples for the predicted value; Setting an initial learning rate, and dynamically adjusting the learning rate by adopting a self-adaptive moment estimation optimizer; And stopping training when the loss function value is converged to a set threshold value or the iteration number reaches a set maximum value, and determining the optimal structure of the model to obtain a trained user energy trend prediction model.
7. The method for analyzing energy space-time characteristics of source network load interaction type users adapting to differentiated scenes according to claim 6, wherein the initial learning rate is set to be 0.001, the set threshold of the loss function is set to be 0.05, and the set maximum value of the iteration times is 1000.
8. The method for analyzing energy space-time characteristics of source network load interaction type users adapting to differentiated scenes according to claim 1, wherein the specific steps of the step S500 include: And (3) reasoning the source-network-load interaction data under different areas, different DG permeabilities, different load types and different time scales by using the trained user energy consumption trend prediction model in the step (S400), and outputting a distributed power output prediction result, an electric vehicle charge-discharge load prediction result and a user energy consumption trend prediction result.

Description

Source network load interactive user energy space-time characteristic analysis method adapting to differentiated scene Technical Field The invention belongs to the technical field of power systems and automation thereof, relates to a source-network-load interactive user energy space-time characteristic analysis method, and in particular relates to a source-network-load interactive user energy space-time characteristic analysis method adapting to a differentiated scene. Background Along with the promotion of 'double carbon' targets and the construction of a novel power system, the medium-low voltage distribution network presents the characteristics of 'source network load' limit fuzzification, user demand diversification and rapid promotion of distributed energy permeability. On one hand, DG output of distributed photovoltaic, wind power and the like has strong randomness and fluctuation, space-time uncertainty exists in charging and discharging loads of electric vehicles, so that the peak-valley difference of the loads of the distribution network is enlarged, the running stability faces challenges, and on the other hand, the influence mechanism of multiple factors such as regional geographic characteristics, economic level, population density, electricity utilization policy and the like on the electricity utilization load distribution of users is complex, and the prior art lacks a quantitative analysis means of a system. The existing user energy space-time characteristic analysis method has the following defects: Firstly, the DG output data is not enough in time sequence and distribution, so that the DG output prediction error is larger, and the dynamic scheduling of the power distribution network is difficult to support; secondly, the multidimensional influences of regional characteristics, social factors and the like are not fully quantized, and the adaptability of user energy trend prediction and actual scenes is poor; Thirdly, the prediction model adopts a single neural network structure, so that the spatial relevance and time sequence dependency relationship of data cannot be captured at the same time, and the prediction precision is difficult to meet the requirements of a differential scene. Therefore, it is highly desirable to provide a source network load interactive user energy space-time characteristic analysis method capable of integrating multi-source data, quantifying multiple influencing factors and adapting to different scenes, which provides accurate data support for access optimization configuration and extension planning of users of a medium-low voltage power distribution network, and improves the running stability of the power distribution network. No prior art publication is found, either the same or similar to the present invention, upon retrieval. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a source network load interactive user energy space-time characteristic analysis method adapting to a differentiated scene, which can solve the technical problems of insufficient suitability, insufficient quantification of multiple influencing factors and lower prediction precision in the energy space-time characteristic analysis of a user of a medium-low voltage distribution network. The above object of the present invention is achieved by the following technical solutions: a source network load interactive user energy space-time characteristic analysis method adapting to a differentiated scene comprises the following steps: Step S100, acquiring multi-source basic data of a medium-low voltage distribution network by adopting a data acquisition technology; Step S200, constructing a multi-influence factor quantification system according to the multi-source basic data acquired in the step S100 and combining regional characteristics, social factors and population density distribution; step S300, based on a convolutional neural network and a long-term and short-term memory network, establishing a user energy trend prediction model integrating multi-source basic data and quantitative system data; Step S400, training the user energy consumption trend prediction model established in the step S300 by adopting a training data set, and iteratively optimizing model parameters through a loss function to determine an optimal model structure so as to obtain a trained user energy consumption trend prediction model; And S500, utilizing the trained user energy trend prediction model in the step S400 to infer source-network-load interaction data in different areas, different DG permeabilities, different load types and different time scales, generating a space-time distribution characteristic map of user energy, and completing the accurate analysis of the space-time characteristics of the source network-load interaction type user energy in a differentiated scene. Further, the specific steps of the step S100 include: collecting historical meteorological data, wherein the histo