CN-117172589-B - Industry typical user power demand response potential evaluation method

Abstract

The invention provides an assessment method for power demand response potential of an industry typical user, which comprises the steps of S1, extracting industry user demand response characteristics based on integrated empirical mode load decomposition and demand response willingness, and S2, assessing the industry typical user demand response potential based on the industry user demand response characteristics and a convolution self-attention mechanism. The validity of the method is verified by carrying out example analysis on a demand response case of a metal processing machinery manufacturing industry in certain city of Zhejiang province, and the result proves that the demand response potential evaluation result obtained based on the method accords with the actual response quantity of a user. The invention solves the problem of modal aliasing, not only can capture long-term and short-term dependency in time sequence, but also can ensure higher accuracy by considering local context information of each time period characteristic, and can provide support for an electricity selling company or a load aggregator to fully utilize user side adjustable resources.

Inventors

• LIN ZHENZHI
• MA YUANQIAN
• WANG YUNCHU
• LU FENG
• YAN YONG
• ZHANG TIANHAN
• CHEN CHANGMING
• QIU WEIQIANG

Assignees

• 浙江大学

Dates

Publication Date

20260512

Application Date

20230830

Claims (3)

1. 1. An industry typical consumer power demand response potential assessment method, comprising: s1, extracting industry user demand response characteristics based on a load decomposition result and demand response will; s2, based on industry user demand response characteristics and a convolution self-attention mechanism, evaluating industry typical user demand response potential; The step S1 specifically includes: s11, decomposing a load sequence of an industry typical user based on an integrated empirical mode decomposition method, and extracting an index vector reflecting objective demand response characteristics of the industry typical user based on a load decomposition result; S12, establishing an index system reflecting user demand response will based on historical response electricity quantity, effective response times, subscription capacity and investigation statistical data of user quotation of the user, and extracting subjective demand response characteristics of the user; step S2 is specifically that according to the extracted demand response characteristics of the industry typical user, multi-period characteristics are obtained based on local context information of the characteristics of each period and causal convolution, and the demand response potential of the industry typical user is evaluated based on the multi-period characteristics and a multi-head self-attention mechanism; In the step S2, when the demand response potential of the industry typical user is evaluated based on the multi-period feature and the multi-head self-attention mechanism, the relative sequential relationship between the demand response features of different periods is considered, each period is regarded as a position in the demand response feature sequence, and the demand response features of the industry typical user are position-coded, which specifically includes the steps of: The demand response feature sequence is provided to include a number of time periods of N, i.e., there are N positions to be encoded, denoted p os = 1,2, n., N represents the position of the demand response feature in the demand response feature sequence; The position embedding vector dimension of the demand response feature is d, and the demand response feature is calculated to obtain ; Determination of Corresponding position coding: when the vector dimension d is an odd number, When the vector dimension d is an even number,
2. 2. The method for evaluating the power demand response potential of an industry typical consumer according to claim 1, wherein the step S11 specifically comprises: Firstly, dividing a certain industry into M electricity types based on a normal cloud model and an improved density peak value rapid clustering algorithm, representing the industry as a set C T = {1,2 }, M, wherein the typical load curve identification result of the M electricity type is represented as P m ＝{P m,1 ,P m,2 ,…,P m,j ,…,P m,J , The J-th moment load of the typical load curve m is the total sampling point number, and the user set in the m-th electricity type of the industry is Representing a b-th user in the m-th electricity type; Is the load sequence of (2) Wherein, the Representing a user Load at j-th moment, and typical load sequence by using integrated empirical mode decomposition method Decomposing to obtain an eigenmode function component and the rest: Wherein G IMF is the total number of eigenmode functions; the trend component is a remainder, called a typical load sequence m, is a deterministic part, and represents the general variation trend of typical load of industry users; Is a typical load sequence The g-th eigenmode function is obtained through integrated empirical mode decomposition; then, on the basis of neglecting high-frequency components of the load sequence, extracting objective demand response characteristics of industry users: Wherein, the Removing the trend component and the residual periodic component after the high-frequency component for the original typical load sequence; then, the residual periodic component is filtered by S-G Smoothing and filtering: According to the difference of production flow and working procedure, the residual periodic component is added according to time period The switching start and stop of the combined electric equipment of the industry user corresponds to the rising and falling of the residual periodic components of each period, which is called a step, and the characteristic that the industry user has the demand response potential of filling the valley or clipping the peak in the period is expressed as follows: Wherein, the An average value of the remaining period component s segments of the typical load curve m; Steps representing the theta-th hour residual period component of an industry typical load curve m, and a sequence formed by the steps is called a step sequence when Indicating that the industry user has peak clipping demand response potential at the theta hour when Indicating that the industry user has a demand response potential of filling the valley at the theta hour; Finally, objective demand response feature vector for typical load m Expressed as: Wherein, the Response characteristic sequences for peak clipping and valley filling demands of industry users, namely
3. 3. The method for evaluating the power demand response potential of an industry typical user according to claim 1, wherein in the step S12, the extracted subjective demand response characteristics of the user include an actual response power amount ratio, an effective response number ratio, an offer response capacity ratio, and a demand response subsidy power rate ratio; (1) Ratio of actual response power Wherein, the The actual response electric quantity ratio of the theta-th hour of the typical load m represents the average value of the actual response electric quantity of each user in the industry electricity consumption type m in the period and the response electric quantity ratio declared by the user; And Respectively users Actual response power (kW) and declared response power (kW) for the θ -th hour; (2) Effective response time ratio Wherein, the The effective response time ratio of the typical load m theta hour represents the average value of the ratio of the historical effective response times to the total offer times of each user in the industry electricity type m in the period, and N iv (b, theta) is the average value of the ratio of the effective response times to the total offer times of the users in the industry electricity type m in the theta hour of the electricity company Total offer times, n i is 0-1 variable, characterizing the user Whether the response is valid or not is judged at the ith time, if the response is valid, the value of n i is 1, otherwise, the value of n i is 0; (3) Subscription response capacity ratio Wherein, the For the typical load m theta hour, the average value of the contract demand response capacity and the capacitance ratio of the user is represented by the contract response capacity ratio of each user and the electricity company theta hour in the industry electricity consumption type m, and Q contr (b, theta) and Q cap (b) are respectively the users Subscription capacity (kWh) and total capacitance (kWh) at hour θ; (4) Ratio of DR patch unit price Wherein, the A rate of a subsidy price for a demand response of a typical load mθ hour, representing a mean value of a ratio of a subsidy price for participation in a demand response declaration of each user in the industry electricity type m θ hour to a declaration price upper limit defined for the subsidy price in an electricity company demand response policy, p cus (b, θ) and p hat (θ) are respectively users of the θ hour Quotation (Yuan/kWh) and electric company subsidy price upper limit (Yuan/kWh); in summary, the subjective demand response feature vector of the typical load m is: (9) The subjective and objective demand response feature vector of the typical load m is expressed as:

Description

Industry typical user power demand response potential evaluation method Technical Field The invention relates to the field of power demand response, in particular to a method for evaluating power demand response potential of a typical user in the industry. Background Clean new energy sources such as wind power, photovoltaic and the like are developed greatly to become a necessary way for energy transformation in China, but the new energy sources bring great risks to safe and stable operation of a power grid due to the characteristics of randomness, uncontrollability and the like, and bring higher requirements to the regulation capability of the power grid. The demand response can actively change the self electricity consumption behavior by exciting a user, so that the uncertainty caused by the high permeability of the new energy is overcome, and the method is an effective means for improving the power grid regulation capability. The electricity selling company clusters industry users, extracts typical electricity utilization types, matches the required demand response indexes according to the demand response potential of each typical electricity utilization type, initiates demand response offers to the industry users of the corresponding electricity utilization types by adopting various excitation strategies, and the users can decide whether to participate in demand response according to the load characteristics and demand response willingness of the users. However, the existing method fails to extract objective demand response characteristics and subjective demand response willingness characteristics of users from industry user load data at the same time, so that the accuracy of the existing method in evaluating typical demand response potential of industry users is not high. Disclosure of Invention The invention aims to provide an industry typical user power demand response potential evaluation method, which comprises the steps of firstly decomposing a user load sequence by using an integrated empirical mode decomposition method, overcoming the mode aliasing problem, accurately representing the characteristics of the load sequence by the obtained eigenmode function components, effectively improving the accuracy of extracting objective demand response characteristic sequences, then establishing a mapping relation between demand response characteristics and user demand response potential by adopting convolution self-attentions based on industry user demand response characteristics, capturing long-term and short-term dependency relations in time sequences, considering local context information of characteristics of each period, better incorporating the local context information into a multi-head self-attentions mechanism, more effectively extracting the demand response potential characteristics of users, improving the accuracy of evaluation results, providing a new idea for accurate scheduling of electric power companies, and having good practical application value. The invention is realized by the following technical scheme: An industry typical consumer power demand response potential assessment method comprises the following steps: s1, extracting industry user demand response characteristics based on a load decomposition result and demand response will; S2, evaluating the typical user demand response potential of the industry based on the industry user demand response characteristics and the convolution self-attention mechanism. Wherein, the S1 comprises: s11, decomposing a load sequence of an industry typical user based on an integrated empirical mode decomposition method, and extracting an index vector reflecting objective demand response characteristics of the industry typical user based on a load decomposition result; The electricity utilization types of all industries and the electricity utilization characteristics of all users in the industries have larger difference, the load change characteristics of the users in the industries are deeply analyzed, the change rule of typical load is accurately mastered, and the method is a premise for accurately evaluating the demand response potential of the typical users in the industries. The industry typical user demand response potential refers to the demand response potential corresponding to the industry typical user load, and represents the average demand response capability corresponding to the industry power utilization type. Therefore, to accurately evaluate the demand response potential of a typical consumer in the industry, it is necessary to first divide the industry electricity usage types and identify typical load curves that characterize the different electricity usage types. Considering that the normal cloud model can fully mine the local dynamic characteristics of the load curve, and the density peak value rapid clustering algorithm has the characteristics of few preset parameters, high calculation efficiency, good adaptability to data distributi