Thermal Power Unit Energy Consumption Feature Extraction Based On Data Mining

In an increasingly tight supply of coal resources, mining the energy characteristics of coal-fired power units is an important foundation to achieve energy-saving operation of the unit. It is also an important direction of the power system to achieve energy conservation and a key to achieve the sustainable economic development of China. Therefore, basic research for thermal power unit energy saving work is of great significance. For this subject, the following researches are made to the issues mentioned above.Firstly, the paper proposes the factor analysis method to dig out the important indices of affecting the unit energy consumption, by using its function of dimension reduction. On the basis of thermal power unit’s historical operation data, this paper extracts some parameters to establish the unit energy consumption common factor analysis model, which the parameters are probably reflect the unit operation and also closely relate to unit’s coal consumption. To verify the extracted common factor model effectiveness, it checks the number of extracted common factor; while, it also considers the special factor factor variance, which it often ignored in the application of factor analysis method, to judge whether the proposed unit energy consumption analysis model appears Heywood phenomenon, further to judge the superiority and inferiority of the fitting extracted factor models quantitatively. In order to better describe the characteristics of the original variables and the extracted common factor, it uses regression mothods to calculate the common factor scores, reaching the results of dimension reduction purposes.Secondly, in the process of thermal power unit target mining,a method of weighted fuzzy C-means clustering (WFCM) algorithm was proposed to determine the monitoring parameters target-value of thermal power plant in this thesis. Considering the influence of different sample points energy consumption indicis for classification is not the same, and they have different contributions to the clustering in the process of mining. Therefore, in order to distinguish the influence degree of all kinds of datas on the target-value of unit energy consumption, this paper uses the weights to express the relative degree of the importance of various data in the target value of unit energy consumption, and presents a weighted fuzzy c-means clustering algorithm which can accomplish target-value clusteing of the unit energy consumption. The excavation of the parameters target-value is more reasonable.Lastly, to improve the objectivity and rationality of energy consumption of thermal power unit comprehensive evaluation, it is proposed to decide the weights of indices in energy consumption by analytic hierarchy process (AHP) and entropy weight method (EWM), therefore the defect of single weight determining is remedied and the decision of weights becomes more reasonable. On this basis, by using two-stage fuzzy comprehensive evaluation method to assess the energy consumption of thermal power unit, at the same time, compared with the single weight determining of analytic hierarchy process. Case analysis show that the evaluation results by the proposed method are more rational and perfective, and conforms to the reality of thermal power unit’s energy consumption level.Through the analysis real-time condition of thermal power units, the results are consistent with the theoretical knowledge.lt is verified the validity and practicability of the proposed models and analysis methods, having some certain guidance to the operation, adjustment and optimization of power plant unit.