Search

CN-121984033-A - Multidimensional adjustment modeling and dynamic aggregation method considering user electricity price perception

CN121984033ACN 121984033 ACN121984033 ACN 121984033ACN-121984033-A

Abstract

The invention discloses a multi-dimensional regulation modeling and dynamic aggregation method considering user electricity price perception, which is characterized by collecting historical load data of multi-type power users, carrying out cluster analysis based on load characteristics to obtain typical load samples, calculating various regulation capability indexes for the typical load samples, carrying out normalization processing, establishing an electricity user response willingness index model based on multi-period electricity price perception, training to obtain a willingness prediction model, carrying out simulation prediction under a set electricity price period length, price and excitation strategy, carrying out normalization to obtain response willingness, synthesizing the regulation capability indexes and the response willingness to obtain a multi-dimensional regulation capability vector data set, and carrying out dynamic grouping on user regulation capability by adopting DEC to obtain an optimal aggregation structure. The invention can describe the behavior difference of users under different excitation and electricity price cycle lengths, has stronger expansibility and adaptability, and is suitable for application scenes such as virtual power plant user grouping, load scheduling, demand response optimization and the like.

Inventors

  • XU QINGSHAN
  • DU JIAO
  • WU FAN
  • PANG CHAO
  • YU JIANCHENG
  • MA SHIQIAN
  • LIU KEYAN
  • LI ZHAO
  • WANG TIANHAO

Assignees

  • 东南大学
  • 国网天津市电力公司电力科学研究院
  • 国网天津市电力公司
  • 国家电网有限公司
  • 中国电力科学研究院有限公司

Dates

Publication Date
20260505
Application Date
20251205

Claims (9)

  1. 1. A multidimensional adjustment modeling and dynamic aggregation method considering user electricity price perception is characterized by comprising the following steps: Step 1, collecting historical load data of multiple types of power users in a preset time period, and constructing a structured load data set after data cleaning; Step 2, carrying out cluster analysis on different types of user load data based on load characteristics to obtain a typical load sample; step 3, obtaining regulation capability indexes including response capacity, response time, duration, response rate and response energy; Step 4, calculating the index values of the typical load samples, and carrying out normalization processing; step 5, establishing a power user response willingness index model based on multi-period electricity price sensing, and training through the load data set constructed in the step 1 to obtain a willingness prediction model; Step 6, under the set electricity price period length, price and excitation strategy, performing simulation prediction based on a willingness degree prediction model, and performing normalization representation on the result to obtain response willingness degree; step 7, integrating the adjustment capability index and the response willingness degree, and constructing a four-dimensional vector set reflecting the user adjustment capability to obtain a multi-dimensional adjustment capability vector data set; and 8, dynamically grouping the user adjustment capability by adopting a deep embedded clustering algorithm based on the constructed multidimensional adjustment capability vector data set to obtain an optimal aggregation structure.
  2. 2. The method according to claim 1, wherein in step 1, the data cleansing includes cleansing historical load data by time alignment, missing value filling and Z-score normalization, and performing anomaly detection and holiday data screening.
  3. 3. The method according to claim 1, wherein the specific process of the step 2 is, Step 21, extracting load change characteristics of users in typical days through a sliding time window, and constructing a load characteristic vector set of each user; step 22, performing preliminary classification on all users by adopting a systematic clustering method to form a plurality of initial clusters; And step 23, performing secondary clustering in each initial cluster by adopting a fuzzy C-means clustering method to obtain a typical load sample.
  4. 4. The method of claim 1, wherein in the step 3, the response capacity refers to the maximum adjustment power which can be realized by the user in a set period of time, the response time refers to the delay time from the receiving of the adjustment command to the starting of the response by the user, the duration is the maximum time for maintaining the adjustment power by the user, the response rate is the power change rate in unit time, and the response energy is the maximum adjustment energy which can be realized; The above indices are modeled and normalized based on historical typical load data.
  5. 5. The method according to claim 1, wherein in the step 5, the following power consumer response willingness index model based on multi-period electricity price perception is established, , Wherein, the The value of the maximum preference is indicated, In response to the coefficient of sharpness, For the price of electricity purchased by the user, The electricity price at the middle point of sensing is represented, As the reference electricity price midpoint parameter, For the period adjustment factor to be chosen, The time period length of the price signal is used for reflecting the behavior threshold adjustment of the user on the power price variation of different time lengths.
  6. 6. The method according to claim 5, wherein in step 5, the method is introduced based on the power consumer response willingness index model Is the mean value, As a normal density function of the standard deviation, , The upper and lower envelopes are defined as follows, , Wherein, the An upper envelope of the uncertainty interval for the power consumer response willingness, A lower envelope of the uncertainty interval for the power consumer response willingness, For uncertain disturbance amplitude coefficients.
  7. 7. The method according to claim 1, wherein in step 6, the user's willingness to respond under different combinations of electricity prices and periodic incentives is determined The normalization process is carried out, the processing is carried out, , Wherein, the And Representing the maximum and minimum response willingness in all user samples respectively, Is the normalized response preference value.
  8. 8. The method of claim 1, wherein in step 7, the user is Four-dimensional vector of adjustment capability of (c) Is a multi-index fusion expression form, wherein each dimension is obtained by weighted linear combination of response capacity, response time, duration, response speed, response energy and response willingness, and is defined as follows, , Wherein the first Dimension by dimension The calculation formula is that, , , Wherein, the Representing a user Is used for the response capacity of the (c), Representing a user Is used for the response time of the (c) in the (c), Representing a user For a duration of time of (a), Representing a user Is used for the response rate of the (c) to the (c), Representing a user Is used for the control of the response energy of the (c), Representing a user Is a response willingness degree of (1).
  9. 9. The method according to claim 1, wherein the specific process of the step 8 is, Step 81, constructing a self-encoder network, and four-dimensional vectors of each adjustment capability Mapping to a low-dimensional latent space embedded representation , , Wherein, the For the encoder mapping function, Is a network parameter; step 82, introducing a cluster center into the latent space By minimizing the deviation between the KL divergence update sample distribution and the target distribution, , Wherein, the For the sample Belonging to a cluster center Is used for the soft allocation probability of (a), In order to achieve a distribution of the objects, To embed spatial clustering losses.

Description

Multidimensional adjustment modeling and dynamic aggregation method considering user electricity price perception Technical Field The invention belongs to the field of power system demand response and user load modeling, and particularly relates to a multidimensional adjustment modeling and dynamic aggregation method considering user electricity price perception. Background With the development of new power systems and the advancement of the "two carbon" goal, consumer side load regulation resources are gradually becoming an important component of power system flexibility. The power demand response mechanism is widely applied to multiple scenes such as virtual power plants, active power distribution networks, comprehensive energy services and the like as an important means for improving the operation efficiency of the system, promoting clean energy consumption and relieving the operation pressure of a power grid. Particularly, in the background that high uncertainty and fluctuation exist on both sides of a source load, how to efficiently identify, model and aggregate user load resources with response potential becomes one of the core problems of current demand response research and engineering practice. However, in the prior art, modeling of user response capability is mainly focused on historical feature statistics or rule-driven modeling based on load curves, and the lack of systematic quantification of multidimensional indicators of actual adjustable capability of users, such as key physical characteristics of response capacity, response time, duration, response rate, response energy and the like, are often not comprehensively considered. Meanwhile, the response behaviors of users to electricity prices and excitation mechanisms have obvious individual differences and periodical changes, and the response potential of the subjective willingness level is often simplified or even ignored in the existing model, so that the real response performance of the user under different strategies is difficult to accurately describe. In addition, the traditional aggregation method mostly adopts static clustering or a partitioning strategy based on a single feature, and is difficult to adapt to dynamic, multi-scale and multi-mode features presented by the change of user adjustment capability and behavior preference along with time, so that the flexibility and the effectiveness of the aggregation strategy in multi-service, multi-period and multi-target scenes are limited. Disclosure of Invention The invention aims to provide a multidimensional adjustment capability modeling and dynamic aggregation method considering user electricity price perception, which establishes unified multidimensional adjustment capability vector representation by fusing user physical response characteristics and behavior preference under an electricity price excitation period, and realizes dynamic identification and optimal aggregation of a user grouping structure by combining a deep embedded clustering method, thereby effectively improving user response identification precision and aggregation scheduling adaptability. In order to achieve the above object, the solution of the present invention is: a multidimensional adjustment modeling and dynamic aggregation method considering user electricity price perception comprises the following steps: Step 1, collecting historical load data of multiple types of power users in a preset time period, and constructing a structured load data set after data cleaning; Step 2, carrying out cluster analysis on different types of user load data based on load characteristics to obtain a typical load sample; step 3, obtaining regulation capability indexes including response capacity, response time, duration, response rate and response energy; Step 4, calculating the index values of the typical load samples, and carrying out normalization processing; Step 5, establishing a power user response willingness index model based on multi-period electricity price sensing, and obtaining a willingness prediction model through the load data set constructed in the step 1; Step 6, under the set electricity price period length, price and excitation strategy, performing simulation prediction based on a willingness degree prediction model, and performing normalization representation on the result to obtain response willingness degree; step 7, integrating the adjustment capability index and the response willingness degree, and constructing a four-dimensional vector set reflecting the user adjustment capability to obtain a multi-dimensional adjustment capability vector data set; and 8, dynamically grouping the user adjustment capability by adopting a deep embedded clustering algorithm based on the constructed multidimensional adjustment capability vector data set to obtain an optimal aggregation structure. In the step 1, the data cleaning includes cleaning the historical load data by time alignment, missing value filling and Z-score sta