A Comparison of Different Load Forecasting Methods Considering Combined and Multiple Strategies

The concept of load forecasting is one of the most fundamental studies in power grids due to its tremendous impact on the operation and expansion of electricity networks. Due to the global approved contracts and restrictions on green production, the adjustment of energy production to demand, is no longer a priority and has become a necessity. With the advent of new meta-heuristic algorithms and machine learning methods, these studies have also undergone many changes. In this paper, a comparison of various load forecasting methods has been done through a case study by considering various constraints, limitations and combined and multiple ways. These methods, which are mainly based on processing a specific pattern, and then predicting, have different categories and can be classified from different perspectives. These algorithms individually have acceptable efficiency and are compared with different evaluation criteria such as MAPE, MSE, etc. However, each algorithm has its drawbacks and may fail in certain circumstances and it is not possible to implement a specific algorithm in all states. Therefore, hybrid algorithms are used to mitigate these shortcomings and challenges, some of which are introduced as open window study topics in the conclusion section.

Keywords:

Load Forecasting , Machine Learning , Artificial Neural Network , Time Series , Smart Grid