A Novel Fractional Order Intelligent Approach: Validation and Control of Industrial Deaerator

The deaerator is an essential part of the steam network for both static and dynamic operation of the boiler systems. If not controlled properly, the variation in the load of power plants can also produce very dangerous scenarios such as over-pressurization, water hammer, and turbine damage. Thus, this work proposes to model an industrial deaerator from the perspective of improving its dynamic performance in the operational load range. A model of the deaerator system is developed and then validated in both open loop and closed loop with respect to the operating characteristics and long-term dynamic data from the industrial deaerator. Further, a novel analytically developed locally blended fractional order recurrent fuzzy neural network proportional-integral-derivative (FORFNN-PID) control strategy is implemented. Its integer-order counterpart (IORFNN-PID), fuzzy PID (FPID), and PID structures are also designed for performance comparison. The design parameters of the controllers are optimally tuned by the metaheuristic Mayfly algorithm (MA) to minimize the mean squared error (MSE). The results obtained indicate the superiority of the proposed FORFNN-PID technique with the fastest response and minimum MSE (287.26 sec & 0.0701), as compared to IORFNN-PID (308.93 sec & 0.0712), FPID (572.05 sec & 0.0715), and PID (1345.4 sec & 0.0745). The adaptive intelligent FORFNN-PID controller demonstrates excellent control even during adverse industrial scenarios such as sudden load variations and disturbance in the system.