Comparing Implicit vs Explicit Regime Identification in Machine Learning Approaches to Short-Range Solar Power Forecasting

In systems with a high penetration of solar power, energy grid systems operators require accurate solar power forecasts to effectively balance the supply and demand of electricity due to the inherent variability in solar power production due to clouds and aerosols that attenuate the irradiance reaching the solar panels. The objective of a solar power forecasting system is to predict the evolution of the attenuation of irradiance from clouds and aerosols. However, the data are generally too coarse resolution for providing information on each cloud in the forecasting domain. Therefore, a solar power forecasting system must find the actual relationships between the predictors in the available data such as surface weather and irradiance observations and the solar power production without modeling each individual cloud’s growth and evolution. Although the surface weather and irradiance data may not be able to model each individual cloud, it is possible to statistically classify cloud regimes based on similar patterns in the data. We use a multi-step machine learning method to build regime-dependent artificial neural networks (ANNs) where we first classify the cloud regime with k-means clustering, and then apply separate ANNs for each cloud regime. We compare this method to a model tree that uses a “separate-and-conquer” algorithm to search for a rule that explains part of the training instances, separates these instances, builds a regression model on these instances, and continues this process until no instances remain. The prediction for a given instance is a weighted average of the multivariate linear regression equations at each node in the tree down to the final node where the instance was assigned. The goal of this study is to compare and contrast the two methodologies for predicting solar power based on either the explicit or implicit separation into regimes for short-range (15-min to 180-min) solar power prediction.