Machine Learning in Solar Energy Cost Optimisation

By shifting the maximum demand from peak rate periods to mid-peak or off-peak rate tariffs, customers can lower their monthly electricity costs by several hundred to several thousand ringgit. With the help of machine learning using historical data to predict load consumption and solar energy generation for the next 7 days with an interval of 30 minutes, users can recognise when they should charge and discharge their energy storage to avoid maximum demand at peak rate especially on rainy days. This research uses all the available machine learning models in an automated machine learning library for time series prediction to identify the well performing models for weighted ensembles. In addition, the historical data is sorted by month and calculated with the help of a tariff library. This allows different users to decide which tariff scheme they can use to reduce their electricity bill when using solar energy and energy storage.