A hybrid system for forecasting 24-h power load profile for Polish electric grid

作者：Brodowski Stanislaw; Bielecki Andrzej; Filocha Maciej

来源：Applied Soft Computing, 2017, 58: 527-539.

DOI：10.1016/j.asoc.2017.04.053

• Machine learning
• Power load forecasting
• Combined model
• Hierarchical Estimator
• Hierarchical Approximator

We present a method of forecasting 24-h power load profile in state-wide power system in Poland. The presented method is based on a hybrid artificial intelligence system. It employs actual temperature forecasts prepared by Interdisciplinary Centre for Mathematical and Computational Modelling of Warsaw University. The machine learning part of the system consists of 24 instances of Hierarchical Estimator: a machine learning method that divides the problem into non-exclusive subproblems with the help of fuzzy clustering and combines results of fairly simple neural networks trained on those subproblems into one, possibly more accurate solution. The presented system also includes a part responsible for dealing with days that have distinct power load patterns, such as additional state holidays. That latter part uses 30 (or 33) appropriately arranged linear regressions.
The proposed approach was tested on historical load data from Poland and a few other countries. The achieved MAPE varied from 1.08% to 2.26% in dependence on the country. Such errors are among the lowest achieved by the published methods.

• 出版日期2017-9