Research On Prediction Of Boiler State Parameter Based On Big Data

With the continuous progress of China’s biomass power generation technology,more and more operating data have been accumulated in biomass power plants.These massive operating data often have valuable development and utilization value.Through the use of big data theory,these massive data can be transformed into information that reflects the combustion state of biomass boilers,and this information can provide reference and guidance for the optimization of on-site boiler operation.This paper takes a 30MW biomass direct-fired unit in Anhui Province as an example.First,a certain amount of historical operation data is extracted from the onsite DCS main engine,and the corresponding boiler thermal efficiency is calculated by combining the operation data with the fuel daily in the plant.Pauta criterion was used to eliminate abnormal data from the original data;In order to improve the establishment speed of the forecast model and avoid the phenomenon of overfitting,the characteristic variables of the input parameters are screened by combining the BP neural network and the average value influence method;then the K-means clustering algorithm is used to process the data.Finally,the data is divided into two categories.Combined with the first-order derivative method to compensate the lag of the unit load,the unit load forecast error has dropped by 25.78%after compensation;two random forest algorithms and BP neural network algorithms,which are widely used in big data analysis in various industries,are used to analyze the two The prediction models for unit load,boiler thermal efficiency,and exhaust NOx content were established under similar working conditions,and the optimal values of the hyperparameters in each model were selected.Most of the final forecast models have an average relative error of less than 1.5%.Combined with the BP neural network model with better forecasting performance,an optimization method based on ant colony algorithm(ACO)for primary and secondary wind baffle control parameters is proposed,and this method is used to optimize and verify random samples.From the results,it can be seen that the unit load and exhaust NOx When the content remains basically unchanged,the thermal efficiency of the biomass boiler has increased by 0.05%-0.25%compared with the predicted value before optimization,and increased by 0.05%-0.5%compared with the actual value.Based on the big data theory,this paper proposes a method for predicting and optimizing combustion parameters of biomass boilers.Compared with traditional methods,it has the advantages of high precision and low cost,and provides a reference for the establishment of intelligent control systems for biomass boilers.