Search

CN-121998490-A - Thermocouple comprehensive evaluation method based on self-evolution learning

CN121998490ACN 121998490 ACN121998490 ACN 121998490ACN-121998490-A

Abstract

The invention discloses a thermocouple comprehensive evaluation method based on self-evolution learning, which comprises the steps of establishing a thermocouple transformation evaluation index system and an index calculation model, calculating comprehensive weights of secondary indexes in the evaluation index system, constructing historical environment state vectors based on historical multi-source information, clustering the historical environment state vectors to obtain a plurality of scene clusters, acquiring real-time environment state vectors at a new evaluation moment, determining corresponding scene clusters, dynamically correcting the comprehensive weights through a rule base constraint and deviation-driven self-evolution mechanism for each scene cluster, acquiring secondary index quantification data of each thermocouple transformation scheme through the index calculation model, combining the comprehensive weights after correction of each secondary index, and calculating the comprehensive score of each thermocouple transformation scheme.

Inventors

  • ZHOU CHENGJIAN
  • QIN HUAN

Assignees

  • 云南电力试验研究院(集团)有限公司

Dates

Publication Date
20260508
Application Date
20260116

Claims (10)

  1. 1. The self-evolution learning-based comprehensive evaluation method for the thermal decoupling is characterized by comprising the following steps of: Establishing a thermocouple transformation evaluation index system and an index calculation model, and calculating the comprehensive weight of a secondary index in the evaluation index system; Constructing a historical environment state vector based on historical multi-source information, and clustering the historical environment state vector to obtain a plurality of scene clusters; Acquiring a real-time environment state vector at a new evaluation moment, determining a corresponding scene cluster, and dynamically correcting the comprehensive weight of each scene cluster through a rule base constraint and deviation-driven self-evolution mechanism; And acquiring secondary index quantification data of each thermocouple transformation scheme through the index calculation model, and calculating the comprehensive score of each thermocouple transformation scheme by combining the comprehensive weight corrected by each secondary index.
  2. 2. The comprehensive evaluation method of thermocouple based on self-evolution learning according to claim 1, wherein the thermocouple transformation evaluation index system comprises a plurality of primary indexes and secondary indexes, wherein the primary indexes comprise financial indexes, thermocouple transformation indexes, energy consumption indexes, environment-friendly indexes and macroscopic policy indexes, and the financial indexes comprise three secondary indexes of net present value, internal yield and dynamic recovery period; The thermal-decoupling transformation indexes comprise four secondary indexes of newly increased peak regulation capacity, newly increased maximum heating capacity, technical safety and technical applicability; The environment-friendly indexes comprise four secondary indexes of sulfur dioxide, carbon dioxide, nitrogen oxide and smoke emission reduction; The macro-policy index comprises two secondary indexes of policy support and promotion of the development of the thermal power industry.
  3. 3. The method for comprehensive evaluation of thermal decoupling based on self-evolution learning according to claim 2, wherein subjective weights of the secondary indexes are calculated by using a hierarchical analysis method, objective weights of the secondary indexes are calculated by using an entropy weight method, and comprehensive weights are obtained after multiplication and summation and normalization processing are performed on the subjective weights and the objective weights.
  4. 4. The method for comprehensive evaluation of thermal decoupling based on self-evolutionary learning of claim 3, wherein constructing a historical environmental state vector based on historical multi-source information comprises: collecting historical multi-source information, wherein the historical multi-source information comprises historical policy and market information, historical unit operation data and historical item feedback information; analyzing the historical policy file by adopting a natural language processing technology and quantifying the historical policy file into a policy intensity index, carrying out normalization processing on historical market price and historical unit operation data, and carrying out deviation statistics on a predicted value and an actual value in historical item feedback information to obtain standardized multi-source historical data; And constructing a plurality of historical environment state vectors according to a preset time dimension or a project period dimension based on the standardized multi-source historical data.
  5. 5. The method for comprehensive evaluation of thermal decoupling based on self-evolutionary learning of claim 4, wherein clustering the historical environmental state vectors to obtain a plurality of scene clusters comprises: Carrying out cluster analysis on all historical environment state vectors by adopting a K-means unsupervised clustering algorithm, presetting a value range of cluster quantity, and determining cluster groups by calculating Euclidean distance between vectors; Calculating a center vector of each cluster group, wherein the center vector is a mean vector of components of all historical environmental state vectors in the corresponding cluster group, each cluster group corresponds to one scene cluster, and each scene cluster represents a group of operation scenes with similar environmental characteristics.
  6. 6. The method for comprehensive evaluation of thermal decoupling based on self-evolution learning according to claim 5, wherein obtaining a real-time environmental state vector at a new evaluation time and determining a corresponding scene cluster comprises calculating Euclidean distance between the real-time environmental state vector and a center vector of each scene cluster, and selecting a scene cluster corresponding to the scene cluster center vector with the smallest Euclidean distance as a matching scene cluster at the evaluation time.
  7. 7. The method for comprehensive evaluation of thermal decoupling based on self-evolution learning according to claim 5, wherein dynamically correcting the comprehensive weight for each scene cluster through a rule base constraint and deviation-driven self-evolution mechanism comprises: Configuring a weight correction vector and an environment sensitivity coefficient matrix for each scene cluster, and constructing a local loss function corresponding to the scene cluster; Performing iterative updating on the weight correction vector and the environment sensitivity coefficient matrix based on the local loss function, and completing iterative updating of the weight correction vector and the environment sensitivity after the decreasing amplitude of the local loss function is smaller than a preset threshold value or reaches the preset iteration times; and normalizing the final updating result to obtain a weight correction vector and a final correction result of the environment sensitivity coefficient matrix.
  8. 8. The method for comprehensive evaluation of thermal decoupling based on self-evolution learning according to claim 7, wherein the rule base comprises a plurality of constraint rules, the constraint rules are used for limiting initial values and value ranges of weight correction vectors and environment sensitivity coefficient matrixes corresponding to each scene cluster, and the constraint rules are set according to a preset priority order.
  9. 9. The self-evolution learning-based thermal decoupling comprehensive evaluation method of claim 7, wherein constructing a local loss function corresponding to a scene cluster specifically comprises: Determining an implemented set corresponding to a target scene cluster, wherein the implemented set is a thermocouple transformation project which belongs to the scene cluster and has acquired actual operation data; Calculating a multidimensional deviation vector of each implemented embodiment, and converting the multidimensional deviation vector into a comprehensive deviation scalar, wherein the multidimensional deviation vector comprises instantaneous deviation, statistical characteristic deviation and time sequence morphology deviation; setting the project importance weight of each implemented embodiment, wherein the project importance weight is determined based on the project scale, the operation duration and the scene matching degree; A local loss function is constructed based on the integrated bias scalar and project importance weights.
  10. 10. The method for comprehensive evaluation of thermal decoupling based on self-evolution learning according to claim 1, further comprising, after implementation of the scheme, collecting actual operation data of all the implemented embodiments and data calculated by the index calculation model, calculating a corresponding global multidimensional deviation vector, and updating physical mechanism parameters, weight generation parameters and rule threshold parameters of the index calculation model by a hierarchical parameter optimization algorithm in combination with the global multidimensional deviation vector.

Description

Thermocouple comprehensive evaluation method based on self-evolution learning Technical Field The invention relates to the technical field of scheme evaluation, in particular to a thermal-decoupling comprehensive evaluation method based on self-evolution learning. Background With the advancement of "two carbon" targets, there is a need for improved flexibility in electrical power systems to absorb high proportions of renewable energy. The cogeneration unit is used as a main heat supply source in northern areas, and the peak regulation capacity is severely limited by the operation mode of 'heat and electricity setting'. The thermal decoupling transformation is a key way for improving the flexibility of the unit by decoupling the thermoelectric output through a technical means. The main stream reconstruction scheme at present comprises low-pressure cylinder near zero output, bypass heat supply, heat pump waste heat recovery, heat storage tank, electric boiler and the like. However, the conventional thermocouple flexibility modification evaluation system has the following problems that ① evaluation is static, and the conventional evaluation is independent of a static index system (such as financial indexes, technical performances and environmental benefits) and does not consider dynamic factors such as unit running states, policy adjustment, market electricity prices and the like. ② Weight solidification, wherein the traditional comprehensive evaluation method (such as AHP, entropy weight method and TOPSIS) depends on fixed weight, and cannot adapt to the differentiated requirements of different areas, unit types and operation scenes. ③ The learning capability is lacking, and the existing method cannot autonomously learn the optimized evaluation rule from the history transformation case, so that the evaluation result is lagged behind the technical development and practice feedback. ④ The targets such as economy, flexibility, environmental protection and the like are balanced, and the priority of each target is difficult to dynamically balance by the traditional method. Disclosure of Invention Aiming at the prior art, the invention aims to provide a thermocouple comprehensive evaluation method based on self-evolution learning, which mainly solves the technical problems in the background art. In order to achieve the purpose, the technical scheme of the embodiment of the invention is realized by the steps of disclosing a thermocouple comprehensive evaluation method based on self-evolution learning, wherein the comprehensive evaluation method comprises the following steps: Establishing a thermocouple transformation evaluation index system and an index calculation model, and calculating the comprehensive weight of a secondary index in the evaluation index system; Constructing a historical environment state vector based on historical multi-source information, and clustering the historical environment state vector to obtain a plurality of scene clusters; Acquiring a real-time environment state vector at a new evaluation moment, determining a corresponding scene cluster, and dynamically correcting the comprehensive weight of each scene cluster through a rule base constraint and deviation-driven self-evolution mechanism; And acquiring secondary index quantification data of each thermocouple transformation scheme through the index calculation model, and calculating the comprehensive score of each thermocouple transformation scheme by combining the comprehensive weight corrected by each secondary index. Optionally, the thermocouple transformation evaluation index system comprises a plurality of first-level indexes and second-level indexes, wherein the first-level indexes comprise financial indexes, thermocouple transformation indexes, energy consumption indexes, environment-friendly indexes and macro-policy indexes, and the financial indexes comprise three second-level indexes of net present value, internal yield and dynamic recovery period; The thermal-decoupling transformation indexes comprise four secondary indexes of newly increased peak regulation capacity, newly increased maximum heating capacity, technical safety and technical applicability; The environment-friendly indexes comprise four secondary indexes of sulfur dioxide, carbon dioxide, nitrogen oxide and smoke emission reduction; The macro-policy index comprises two secondary indexes of policy support and promotion of the development of the thermal power industry. Optionally, the subjective weight of the secondary index is calculated by adopting an analytic hierarchy process, the objective weight of the secondary index is calculated by adopting an entropy weight process, and the comprehensive weight is obtained after multiplying and summing the subjective weight and the objective weight and normalizing. Optionally, constructing the historical environmental state vector based on the historical multi-source information includes: collecting historical multi-sourc