Research On Target Value Of Unit Optimal Operation Based On Data Clustering

Optimized operation plays an important role in ensuring the safety and economy of the unit operation.The method for determining the target value of operation is an important research content.The historical operation data of the unit can provide a large number of modeling samples for the optimal operation of the unit.However,there are problemsj such as mixed distribution,difficult feature extraction and uneven distribution in the operational data.Therefore,it is necessary to research and apply appropriate data mining methods in a targeted manner.This paper takes a 660 MW unit as the research object,and carries out research work around the method of determining the target value of optimization based on data clustering.The main research contents include:Firstly,a data discretization method is proposed.The ordered clustering algorithm is used to cluster the ascending order data samples to obtain the best discretization interval.Taking the coal quality data discretization as an example,coal quality data is effectively classified.Combining the proposed discretization method with rough set theory,a method of attribute importance calculation is proposed,which can solve the problem of difficult feature extraction of thermal data.A downsampling method is also proposed based on the discretization method,and data samples with uniform distribution can be obtained.Secondly,in order to obtain the predicted load of the unit load required to determine the target value of the operating parameters,a load forecasting method based on cluster analysis and HMF is proposed.The load series of several hours before the predicted time point is matched with the historical load data of the same segment,and the future load is forecasted by the load trend of the most similar day.Compared with the traditional ARMA prediction model,the load prediction model established in this paper has higher prediction accuracy.Furthermore,aiming at the problem that the thermal data has many different cases and the clustering efficiency is low,an improved density peak clustering algorithm which can automatically determine the clustering center is proposed.The algorithm is used to cluster the controllable characteristic parameters of the desulfurization system,and the different operating conditions are divided.The economical evaluation of each working condition is carried out by using the sum of desulfurization variable cost and environmental protection tax as the index.The target operating conditions of the desulfurization system and the target operating parameters are obtained.The target value of the SO2 outlet concentration of the desulfurization system is obtained by taking the lowest average cost under each boundary condition as the index.In addition,a feature weighted clustering algorithm is proposed for the different degree of tightness of the distribution of various characteristic parameters of thermal data and the dynamic increase of data.By dividing the newly added data points into existing clusters,the cluster center can be updated under incremental data.The feature weighted clustering algorithm is used to determine the coal consumption target value,and the coal consumption target value of each working condition is obtained,and the coal consumption target value can be updated in real time.Finally,the unit target value optimization running software is designed,which can guide the field operation in real time.