Optimizing Boiler Efficiency by Data Mining Teciques: A Case Study

In a fertilizer plant, the steam boiler is the most important component. In order to keep the plant operating in the effective mode, the boiler efficiency must be observed continuously by several operators. When the trend of the boiler efficiency is going down, they may adjust the controlling parameters of the boiler to increase its efficiency. Since manual operation usually leads to unexpectedly mistakes and hurts the efficiency of the system, we build an information system that plays the role of the operators in observing the boiler and adjusting the controlling parameters to stabilize the boiler efficiency. In this paper, we first introduce the architecture of the information system. We then present how to apply K-means and Fuzzy C-means algorithms to derive a knowledge base from the historical operational data of the boiler. Next, recurrent fuzzy neural network is employed to build a boiler simulator for evaluating which tuple of input values is the best optimal and then automatically adjusting controlling inputs of the boiler by the optimal values. In order to prove the effectiveness of our system, we deployed it at Phu My Fertilizer Plant equipped with MARCHI boiler having capacity of 76-84 ton/h. We found that our system have improved the boiler efficiency about 0.28-1.12% in average and brought benefit about 57.000 USD/year to the Phu My Fertilizer Plant.