System Optimization Analysis Program (SOAP)

The relationships between the operating parameters of a fossil power plant and its optimum performance are highly nonlinear and complex.  Computer modeling based on the basic physical principles typically fails to optimize the multiple variables of plant functions.  The artificial neural network (ANN) technology is very effective in solving nonlinear problems and is capable of performing system optimization tasks.  ANN uses a self-adaptive learning process that can condition the data and develop the patterns of plant processes.  It mimics human brain functions and can learn on its own as it processes a large number of variables.  Special features of ANN technology include:

• high speed response to parallel input signals
• adaptive mapping capabilities
• ability to perform classification and function modeling
• ability to predict future behavior
• ability to learn system performance characteristics from systematic tests or running data
• ability to perform performance tradeoffs among many different plant objectives

SOAP is a PC-based computer program with a fast-learning algorithm developed by FGS to be used for system optimization.  Boiler performance and combustion optimization are typical examples that demonstrate the effectiveness of employing this software for the reduction of NO_x and preservation or improvement of system performance.  Pre- and post-processes were also developed to determine the critical operating factors and to promote fast system learning which reduces time and cost of optimization processes.  SOAP is an online system and can be operated in either closed or open-loop mode.  Some examples can be found in Power Plant Performance Optimization.