Analysis of electrical energy consumption and neural network estimation and forecasting of loads in a paper mill: Dissertation

The present research was divided into two subareas: an examination of the factors affecting electrical energy consumption of the papermaking line and estimating and forecasting loads in the papermaking line. During recent years, events such as the deregulation of power markets and general energy-saving agreements have increased the need for more efficient electrical energy management and for exact knowledge and more accurate forecasting of electrical energy consumption, even at the process line level. For efficient and continuous savings and to develop more accurate forecasting concepts, information on the components of loads and their dynamics is a basic requirement. Previous load research has been mainly carried out at the power distribution and power station levels. The present project initiated a detailed study of loads in industrial processes, beginning with the papermaking environment. Factors affecting electrical energy consumption of the papermaking line were classified and studied in four papermaking lines of different age and product grade. A more detailed study of loads was carried out mainly in the fine paper and newsprint lines. Loads were classified according to their effects on the dynamics of line power consumption and their relationship with the main elements of the production plan: production amount and grade (basis weight). A new and accurate neural network-based method for estimating the motor power basis of current measurement has been presented. A new, accurate, knowledge-based cascaded neural network method for forecasting electric loads in a paper mill has been developed. The new method is adaptive, self-learning, easy to adapt to a fully automatic mode and able to utilize expert knowledge.