Padmavathy Kankanala, Sanjoy Das, and Anil Pahwa
Committee Machines, Neural Network, Annealing, Outages
Electric reliability is an important concern for the utility companies. Weather related outages have a significant impact on it. There are many regression based models to estimate outages from weather factors in overhead distribution system. This paper proposes the use of committee machines composed of multiple neural networks to estimate outages. A major challenge for using a committee machine is to properly combine predictions from multiple networks, since the performance of individual networks is input dependent due to mapping misrepresentation. This paper presents a new method in which the individual network predictions are combined dynamically. The error minimization is performed using the mean field annealing theory. Results obtained for the study area in Kansas are compared with observed outages to evaluate the performance of the model for estimating these outages. The results are also compared with previously studied regression and neural network models to determine an appropriate model to represent effects of wind and lightning on outages.
Important Links:
Go Back
