GM(1,1) based improved seasonal index model for monthly electricity consumption forecasting

Accurate monthly electric load forecasting plays an important role in the development of a smart grid. Although many algorithms have been proposed for load forecasting, these algorithms suffer from poor prediction accuracy or high computational complexity; moreover, they require a large amount of data for training. To address these challenges, in this study, a new seasonal-index model based on grey model optimisation is proposed. First, three trend equations, i.e., linear, exponential, and logarithmic equations, are adopted in the proposed model to fit the development law of electricity consumption. We then use a grey model to predict the seasonal index (SI). Finally, the forecast is obtained by combining the trend forecast value with the SI. To completely evaluate the performance of the proposed algorithm, monthly electric load data from some typical industries of a city in southern China were used in the test. On comparing with some benchmark models, the results indicate that, if the electric-load data of the industries shows the characteristics of trends and (or) seasonality, the proposed algorithm can significantly improve the forecast accuracy even with a small monthly electric load dataset, otherwise, its performance will degrade in a certain level. In particular, we can obtain a mean absolute percentage error (MADE) of 3.6%, 2.6% and 2.9% for three typical datasets, respectively, and each with only 54 data points. Moreover, the proposed algorithm has the lowest execution complexity, which confirms its advantages over the benchmark schemes. (c) 2022 Elsevier Ltd. All rights reserved.