Theoretical Research On Data Mining-based Energy-saving Diagnosis And Optimization For Large Coal-fired Power Units

It is essential to build an economic, efficient and stable energy supply system for the development of national economy. Considering the resource structure and distribution, coal is definitely the major primary energy in China and the large coal-fired power units are dominant in the whole power and energy supply system. It is of great significance to strengthen the research related to energy conservation of large coal-fired power units for the accomplishment of both industrial energy conservation and low-carbon economy.The large coal-fired power units characterize as wide thermodynamic scale, huge equipment, large flow and mass, which results in distinct nonlinear feature in energy transmission, conversion and dissipation for specific equipment, system and process. There's high-dimensional nonlinear correlation between the energy consumption in power generation and the external environment, resources and load demand. For this, several advanced data mining theories and methods were introduced in this research for the energy-consumption analysis and energy-saving diagnosis of large coal-fired power units. Large volume of historian and online real-time monitoring operation data were processed by data mining method to find the potential and useful patterns as well as the knowledge relevant to the energy-consumption features of power units. Based on this, the advanced energy-saving diagnosis and optimization methodology was proposed in this research, which can reflect the complicated and diverse operational conditions and constraints. It is a comparatively complete and new methodology of energy-saving diagnosis and performance optimization covering the hybrid of data processing, complex thermal power system modeling, energy-saving decision making, realizable optimal target determining, off-line energy-consumption analyzing, online energy-consumption diagnosing and optimizing etc..Considering the characteristics of large volume, multiple dimension, hybrid category, highly nonlinear and coupling in the operation data of large coal-fired power units, fuzzy rough sets (FRS) theory was introduced in data processing and feature selection. By means of FRS-based attribute reduction algorithm, the key energy-consumption variables and corresponding significant degree were determined, which is important for the modeling of energy-consumption characteristics of power units.The modeling theory and methodology for complex thermal power system was developed by means of a historian operation data-driven model. By taking account of the dependence between the input features and decision index, an improved support vector regression (SD-SVR) modeling algorithm was proposed to model the energy-consumption characteristics of large coal-fired power units. The resultant model is convenient and accurate to illustrate the correlation between the energy consumption and external operation constraints, equipment features and control conditions of power units.The data mining-based dynamic determination method of realization optimizing targets was proposed, which is adaptive to the similar or comparable operation conditions. For this purpose, based on the FRS decision table reduction algorithm, the knowledge base of performance optimization and energy-saving diagnosis was built for large coal-fired power units. The proposed method is fast, adaptive, recurrent and automatically adjustable for the determination of optimizing targets in different operation conditions of power units.The theory and methodology of data mining-based energy-saving diagnosis and optimization was presented in this work. By categorizing the key energy-consumption variables, the features of specific target values, controllable energy losses in power generation were specified. Based on this, the energy-saving diagnosis model was proposed to determine the energy losses and distribution for the given operation conditions, external constraints and equipment features. A new concept of contrastive optimization and diagnosis was proposed to improve the energy-saving diagnosis for the power units of the same kind and breakthrough the bottlenect of high-level energy consumption resulted from the improper operation of power units. The experimental and simulating platform of such data mining-based energy-saving diagnosis and performance optimization was built and the conceptual models were proposed related to online operation optimization and energy-saving diagnosis.