CN-122026427-A - Load adjustable characteristic evaluation and interactive regulation method for users in three-way industry

CN122026427ACN 122026427 ACN122026427 ACN 122026427ACN-122026427-A

Abstract

The invention provides a load adjustable characteristic evaluation and interaction regulation method for third industry users, which aims to improve the resource adjustable capacity of a demand side and support the supply and demand cooperative regulation under a novel power system. Firstly, collecting typical three-product user load and multi-source heterogeneous data, preprocessing and extracting features, adopting kernel principal component analysis to realize feature dimension reduction, and completing typical user load clustering through a K-Means clustering algorithm. And secondly, constructing a multidimensional user image evaluation system based on the clustering result, and quantitatively evaluating the typical user load adjustable potential by combining a multidimensional time sequence load model. And finally, establishing a multi-constraint optimization scheduling model, introducing a particle swarm optimization algorithm to solve, and outputting a targeted regulation strategy and regulation instruction of the time-division and equipment-division type. The invention effectively realizes the fine evaluation and the rapid dispatching optimization of the load adjustable potential of the third-generation users.

Inventors

ZHU XIAOMIN
LIU YINGHUI
CHEN KE
ZHOU YAJUN
LI YUHUA
YU XIAOHAN
ZONG XUHUI

Assignees

北京交通大学

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (8)

1. The load adjustable characteristic evaluation and interactive regulation method for the users in the three-way industry is characterized by comprising the following five steps: S1, preprocessing load data, namely collecting multi-source heterogeneous data of typical third industry users and preprocessing load related data; S2, feature dimension reduction and clustering, namely performing feature extraction and dimension reduction on the preprocessed data, and obtaining a typical user load clustering result of a typical three-product industry through a K-means clustering algorithm; S3, constructing a multi-dimensional user portraits evaluation system based on the clustering result, and quantitatively evaluating the load adjustable potential of each typical user by combining a multi-dimensional time sequence load model; S4, establishing an interactive optimization scheduling model, namely establishing a power grid and user side resource friendly interactive optimization scheduling model based on the evaluation result of the load adjustable potential, wherein the model simultaneously considers the total load adjustment amount, the user comfort and the response speed as multiple constraints; and S5, outputting a model solving and regulating strategy, namely solving the interactive Optimization scheduling model by adopting a Particle Swarm Optimization (PSO) algorithm, and outputting an optimal load regulating strategy for different time periods and different devices.
2. The load adjustable characteristic evaluation and interactive control method for users in three industries according to claim 1, wherein the multi-source heterogeneous data in step S1 comprises the following steps: (1) Industry types include real estate industry, wholesale and retail industries; (2) The load related data includes electrical load data, device parameter data, and environmental impact data.
3. The load adjustable characteristic evaluation and interactive control method for users in the three-way industry according to claim 1, wherein the preprocessing in step S1 comprises the following steps: (1) Filling the missing value by adopting an interpolation method; (2) By passing through Removing abnormal values by using a criterion; (3) The load value fluctuations were smoothed by a Moving Average method (Moving Average).
4. The load adjustable characteristic evaluation and interactive control method for users in the three-way industry according to claim 1, wherein the feature extraction in the step S2 comprises the following sub-steps: (1) Extracting time domain features including daily average load Peak-valley difference of daily load Load factor Peak load time ; (2) Extracting frequency domain features including load fluctuation frequency And amplitude of fluctuation ; (3) Performing nonlinear feature dimension reduction on the feature matrix by adopting a kernel principal component analysis (KERNEL PRINCIPAL Components Analysis, KPCA) method; (4) Selecting a front part with a cumulative contribution rate not lower than 85% The main components are used for obtaining the feature matrix after dimension reduction 。
5. The load adjustable characteristic evaluation and interactive control method for users in three industries according to claim 1, wherein the multi-dimensional user portrait evaluation system in step S3 comprises the following steps: (1) Information labels, namely industry, geographic position and scale attribute of the user; (2) Load labels, namely labels obtained according to the characteristics of a load curve, including midday peak type, double peak type and stable type; (3) Psychological price labels, which are the degree of sensitivity of users to electricity price, can be divided into price-sensitive and price-insensitive; (4) Response behavior labels, namely willingness and historical behavior of users to participate in demand response, can be divided into an active response type and a potential participation type.
6. The method for evaluating and interactively controlling the load adjustable characteristics of users facing the three-way industry according to claim 1, wherein the evaluation of the load adjustable potential in the step S3 comprises the following steps: S31, constructing a Long Short-Term Memory (LSTM) model as the multidimensional time sequence load prediction model; S32, training an LSTM model through a Adam (Adaptive Moment Estimation) optimizer, and minimizing a load prediction Mean Square Error (MSE); s33, calculating the maximum adjustable load, the adjustment response speed and the continuous adjustment duration as core adjustable indexes based on a load curve predicted by an LSTM model; s34, determining each index weight by adopting a hierarchical analysis (ANALYTIC HIERARCHY Process, AHP) , , After weighted summation, the final load adjustable potential evaluation result is obtained 。
7. The load adjustable characteristic evaluation and interactive regulation method for users in the three-way industry according to claim 1, wherein the friendly interactive scheduling model in step S4 has the constraint conditions that: (1) Load regulation constraints: , Is the first Time of day (time) The class user adjusts the total load amount, Is the first Moment aggregator maximum adjustable load; (2) User comfort constraints: , wherein, , , Is the first Class user presence The time of day due to the comfort loss value actually generated by the load adjustment, A maximum comfort loss threshold for a class j user; (3) Response speed constraint: , For the overall response time of the aggregator, Is the first The response speed of the class user.
8. The load adjustable characteristic evaluation and interactive regulation method for users in three industries according to claim 1, wherein the load adjustment instruction output in the step S5 is a differential strategy, and specifically comprises: (1) Preferentially adjusting the temperature setting and fresh air proportion of a central air conditioner for commercial buildings; (2) Adjusting illumination power and optimizing charging time periods of electric vehicles for wholesale and retail industries; (3) Dynamically adjusting non-core device loads for the cultural sports and entertainment industries; (4) The heat pump and kitchen ventilation equipment are linked in the lodging and catering industry to cut peaks and fill valleys.

Description

Load adjustable characteristic evaluation and interactive regulation method for users in three-way industry Technical Field The invention relates to the technical field of power systems and demand side management, in particular to a method for analyzing and interactively regulating load characteristics of users in a third industry. Background In the global energy structure transformation and the large background of 'carbon peak, carbon neutralization' of China, the permeability of renewable energy sources represented by wind energy and solar energy in an electric power system is rapidly increasing. However, the inherent intermittence and uncertainty of new energy power generation presents a serious challenge for safe and stable operation of the power system. To effectively address these challenges, deep mining and utilization of Demand-side resources (DR) has become a critical direction of power industry development. In ultra-large cities, such as Beijing city, the industrial structure thereof has evolved to be dominant by a third industry, and the power consumption ratio thereof continuously rises, so that the ultra-large cities become a main component of urban power grid load. The third industry covers industry types and is various, and electric equipment (such as a central air conditioner, illumination, a data center and an electric automobile charging pile) is high in isomerism, so that the overall load characteristic of the electric equipment presents high complexity and dynamic performance. At present, research on the regulation technology of industrial load is mature, but research on the third industrial load characteristic is still in the primary stage. The existing load analysis method often has difficulty in accurately describing the refined electricity utilization behaviors and internal rules of three-product users, so that the load adjustment potential of the third industry cannot be accurately estimated and effectively quantized. In addition, a cooperative regulation and control strategy for multi-component heterogeneous equipment in the three-production industry is lacking, so that flexible load resources at a user side are difficult to form a large-scale and schedulable aggregation effect to participate in power grid interaction. Therefore, the invention provides a load adjustable characteristic evaluation and interactive regulation method for users in the three-way industry to solve the problems. Disclosure of Invention Aiming at the situation, the load adjustable characteristic evaluation and interactive regulation method for the third industry user can realize the fine characterization of the load characteristic of the user, the quantitative evaluation of the adjustable potential and the interactive regulation under the multi-constraint condition. The technical scheme comprises the following steps: s1, preprocessing load data And aiming at third industry users, systematically collecting multi-source heterogeneous data comprising historical electricity load data, equipment parameter data and external environment influence factor data, and preprocessing the load data, namely filling missing values, removing abnormal values and smoothing load fluctuation. S2, load characteristic dimension reduction and clustering The method comprises the steps of carrying out feature engineering on preprocessed data, extracting statistical features capable of reflecting dynamic characteristics of the preprocessed data from two angles of a time domain and a frequency domain, carrying out nonlinear feature dimension reduction by adopting kernel principal component analysis (KERNEL PRINCIPAL Components Analysis, KPCA) to obtain core features, and determining optimal cluster numbers by taking a contour coefficient maximization as a target through a K-Means clustering algorithm to realize division of typical user load modes. S3, typical user portrayal construction and potential evaluation Aiming at the typical user category formed by each cluster, a refined user image evaluation system is constructed from a plurality of dimensions of information, load, psychological price and response behaviors, a multidimensional time sequence load prediction model constructed by combining a Long Short-Term Memory (LSTM) is used for quantitatively evaluating the maximum adjustable load, the adjustment response speed and the continuous adjustment duration core index of each typical user based on a predicted load base line, and a hierarchical analysis method (ANALYTIC HIERARCHY Process, AHP) is used for weighted summation to obtain a final load adjustable potential evaluation result. S4, establishing an interactive optimization scheduling model Based on the evaluation result of the adjustable potential of the load, introducing a load aggregator as a market subject, and carrying out integrated management on the evaluated distributed user resources with the adjustable potential. And (3) establishing a friendly interactive optimizatio