Study On Operation Optimization Of Thermal Power Units Based On Big Data Mining Technology

Due to the unique resources structure of our country,thermal power will continue to be dominant in the whole power system in the future for a long period of time.Considering the requirements to construct green electrical ecology and build a safe and efficient energy system,it is of ggreat significance to study the operation optimization,energy saving and comprehensive efficiency evaluation of thermal power plant,in order to improve the level of energy efficiency of thermal power generation.With the development of information integration in power system,large quantity operation data has been accumulated in thermal power plant.How to discover and utilize the hidden knowledge of the massive data to promote the operation optimization and green power development further has been a research focus in the field of thermal power generation.To solve this problem,we introduced big data technology into the research of operation optimization,and established a analysis platform of big data in power plant to achieve the parallel improvement of mining algorithms.Some strong association rules were obtained to determine the target values of operation parameters based on the new parallel algorithms.These target values were used to guide the operation optimization and build a comprehensive energy efficiency evaluation system.Firstly,the basic characteristics of big data in power plant were clarified through analyzing the definition and "3V3E" characteristics of big data in electric power industry.According to.the analysis process of big data in power plant,we deepened the connotation as a analysis hierarchy,and concretized the form as a analysis chain.Due to the lack of big data processing techniques in electric power field,a analysis platform of big data in power plant was proposed to construct,while the two data processing techniques of batch processing and steam processing were introduced.Therefore,value mining and knowledge acquisition can be achieved through the information interaction of big data flow and analysis platform.Secondly,data quality of a ultra supercritical coal-fired unit and a combined cycle unit were analysed,while,the data quality improvement strategies which are data detection and preprocessing were presented.In the process of data detection,the method of variance threshold judgment was used in steady data extraction and the moving average method was used in data synchronization.In the data pretreatment process,a improved BP neural network algorithm based on dynamic adjustment strategy was adopted to forecast and supply the missing values.Meanwhile,a optimized parallel K-Means clustering algorithm based on MapReduce architecture of the big data analysis platform was obtained to realize the efficient data discretization.Then,a association rule algorithm was selected to determine the target values of operation parameters.For the sake of enhancing the massive data processing abilities of the association rule algorithm,two advanced strategies were adopted in this thesis.On one hand,the attribute reduction concept was introduced to reduce the original data set size.On the other hand,a parallel realization of the association rule algorithm was presented combined with the MapReduce architecture.Accordingly,the improved efficient parallel algorithm was employed to extract some strong association rules to determine the operating target values,which were selected as the reference index of the operation optimization.Furthermore,the economic benefit according to the fuel saving was rough estimated.Finally,current approaches to evaluate the energy efficiency were discussed and the disadvantages were summarized in this thseis.Combined with the target values obtained by big data technology,a comprehensive evaluation system of energy efficiency based on the distribution of gas sensitivity and improved principal component analysis was proposed.In the comprehensive evaluation system,real-time effect of the energy consumption caused by operation parameters were quantized by solving the multi-stage distribution column of gas consumption sensitivity;while,a reasonable comprehensive assessment of the units was achieved according to the improved principal component analysis algorithm based on logarithmic change strategy and entropy method.Furthermore,a performance monitoring and evaluation system for combined cycle was developed to realize visualization of the comprehensive evaluation system.