Use of neural networks for modeling and predicting boiler operating performance

Abstract

The need for high boiler operating performance requires the application of improved techniques for rational use of energy resources. Presented analysis is led by finding possibilities how energy resources can be used wisely to secure more efficient final energy supply, to save money and limit environmental impacts. But the biggest problem and challenge are related to the stochastic nature and variety of influencing factors in energy transformation process. In that context, reliable prediction of energy output variation for expected influencing factors is very valuable because it allows planning and carrying out necessary measures to reach mentioned goals. The paper proposes and presents one method for modeling, assessing and predicting efficiency of boilers based on measured operating performance. The method implies the use of neural network approach to analyze and predict boiler efficiency. Neural network calculation reveals opportunities for efficiency enhancement and makes good insight into influencing factors onto boiler operating performance. The analysis is based on energy surveys for randomly selected 65 boilers in the Province of Vojvodina carried out at over 50 sites covering the representative range of industrial, public and commercial users of steam and hot water. Collected experimental measurements amount to 65 × 5 out of which 300 are used as training data for neural networks and the rest (25) is used for testing and validation. The sample formed in such a manner covers approximately 25% of all boilers in the Province of Vojvodina and provides reliability and relevance of obtained results which can assist in recommending and specifying opportunities for reducing irrational energy use and related costs. Creating database combined with soft computing assistance enables wide range of possibilities for identifying and assessing factors of influence and making critical evaluation of used practices on the supply side as a source of identified inefficiency.