Simple Exponential Smoothing for Forecasting the Numbers of Pole-Mounted Transformer Failures

Pole-mounted transformers form the vital link between the electric utility company and the final consumer of electricity. In this paper, the authors use the simple exponential smoothing (SES) forecasting technique to forecast the quarterly numbers of pole-mounted transformer failures. The value of the smoothing constant,
$\alpha$
, to be used in the forecast, is found by first formulating a non-linear programming problem (NLP) problem to minimize the forecast root mean square error (RMSE). Then, a software code that implements exhaustive search algorithm to solve the NLP is written in Python. The SES forecast is compiled and compared to the multiplicative and additive decomposition forecasting methods. Various measures of error, including mean absolute deviation (MAD), mean absolute percentage error (MAPE), mean squared error (MSE), and RMSE. It was found that, irrespective of the measure of error considered, the SES forecast was always outperformed by either of the decomposition methods.