Neuro-Fuzzy and Particle Swarm Optimization based Model for the Steam and Cooling sections of a Cogeneration and Cooling Plant

Developing a first principle nonlinear model for a thermal system that is already in operation is a very difficult task attributed to missing design parameters. This paper considers nonlinear modeling of subunits of a Cogeneration and Cooling Plant (CCP)-Heat Recovery Steam Generator (HRSG), Steam Header (SH) and Steam Absorption Chiller (SAC). Neuro-fuzzy approach trained by a sequence of optimization algorithms-Particle Swarm Optimization (PSO) followed by Back-Propagation (BP)-is used to develop models for the steam drum pressure, steam drum water level, steam flow rate and chilled water supply temperature. It includes the calculation of model confidence intervals (CI) based on the assumption that model and measurement errors are normally distributed and independent. Real operation data collected from Universiti Teknologi PETRONAS CCP is used to train and validate the models. Varying the probability in reading the percentage value of t-distribution for fixed degrees of freedom, a test is also performed on the capacity of the models for fault detection. The results show that the technique can be used to develop a substitute model for the three units, with the confidence level decided by the user.