Implementation of a multi-agent system for optimized multiobjective power plant control

• Main Authors: Head, J D, Gomes, J R, Williams, C S, Lee, K Y
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial neural networks; Feedforward neural networks; Mathematical model; Message systems; Monitoring; multi-agent systems; multi-objective optimization; Multiagent systems; Power generation; Power plant control; reference governor
• DOI: 10.1109/NAPS.2010.5619588

Solving the problem of optimally controlling a power plant based on multiple objectives, such as minimizing pollution, maximizing equipment life, etc., and coordinating each of the involved tasks that must be performed in distributed environments is a very complicated feat, which involves many individual computationally intensive tasks. For the presented method of design, these tasks include calculating feasible control valve operating ranges based on unit load demand, parameter optimization, training neural networks, monitoring and managing real-time input/output data, and task delegation, among others. It is because each of these tasks requires such computational overhead and these systems need to be coordinated among distributed environments that dividing them up into multiple agents is necessary. The presented method of design of the multi-agent system is a continuation of research to develop a multi-agent system to implement a technique for computing optimal multiobjective power plant controls in Matlab.