Optimization for Boiler Based on Data Mining and Multi-Condition Combustion Model

To facilitate the online calculation and optimization of boiler efficiency, a 130 t/h Circulating Fluidized Bed boiler in a power plant is taken as the research object. Firstly, the historical data is classified into six different working conditions by K-means algorithm. Then, the model parameters of least squares support vector machine are optimized by the sparrow search algorithm to construct the boiler combustion model. The parallel optimization strategy of online optimization and offline optimization is proposed. An optimal case base with a sample size of 300 is established on the basis of the combustion model. In order to filter the most similar cases to the current working condition in the optimal case base, a comprehensive similarity is proposed on the basis of weighted Euclidean distance and weighted gray relation. While ensuring the stability of boiler operation, the condition variables for the current working condition are inputted for case matching, and the optimal values of operating variables and the flue gas oxygen content are outputted. As indicated by the results, the root mean square error and coefficient of determination of the model are 0.1721% and 0.9533, respectively, while the modeling time consumption is 1674 s. According to the optimization calculation, the boiler efficiency is improved by 1.16% on average. Thus, the system is deemed suitable for online optimization.