REAL TIME PULVERISED COAL FLOW SOFT SENSOR FOR THERMAL POWER PLANTS USING EVOLUTIONARY COMPUTATION TECHNIQUES

Pulverised coal preparation system (Coal mills) is the heart of coal-fired power plants. The complex nature of a milling process, together with the complex interactions between coal quality and mill conditions, would lead to immense difficulties for obtaining an effective mathematical model of the milling process. In this paper, vertical spindle coal mills (bowl mill) that are widely used in coal-fired power plants, is considered for the model development and its pulverised fuel flow rate is computed using the model. For the steady state coal mill model development, plant measurements such as air-flow rate, differential pressure across mill etc., are considered as inputs/outputs. The mathematical model is derived from analysis of energy, heat and mass balances. An Evolutionary computation technique is adopted to identify the unknown model parameters using on-line plant data. Validation results indicate that this model is accurate enough to represent the whole process of steady state coal mill dynamics. This coal mill model is being implemented on-line in a 210 MW thermal power plant and the results obtained are compared with plant data. The model is found accurate and robust that will work better in power plants for system monitoring. Therefore, the model can be used for online monitoring, fault detection, and control to improve the efficiency of combustion.