Load interval forecasting methods based on an ensemble of Extreme Learning Machines

In this paper, three individual indices, as well as a new comprehensive index, are introduced to evaluate prediction intervals. Then, two practical methods, namely, Interval Extension Method and Optimal Scalar Method are proposed to build the prediction intervals based on an ensemble of Extreme Learning Machines. Case studies on hour-ahead load interval forecasting with respect to Chicago Metro Area are conducted, results of which illustrate that the proposed two methods are able to build high-quality prediction intervals and Optimal Scalar Method tends to have better performance.