Improving Short-term Output Power Forecasting Using Topological Data Analysis and Machine Learning

Forecasting, also referred to as now-casting, has been deployed in various areas such as weather, economics, power generation, and load forecasting. Different techniques are used for short-term forecasting. These techniques include traditional methods such as auto-regressive integrated moving average (ARIMA), and artificial intelligence (machine learning) techniques such as artificial neural networks and support vector machines. In this paper, we report the improvement in the power utility's short-term output power generation using a combination of machine learning techniques and topological data analysis (TDA). The coefficient of determination is used as a metric to assess the performance of various machine learning schemes used. From the results obtained, it can be observed that this coefficient of determination improves in cases where TDA is used as a pipeline to machine learning schemes. These results underline the significance of the approach reported in this paper.