A Machine Learning Approach to Dairy Farm Energy Disaggregation

The growing number of behind-the-meter solar panel installations introduce potential safety and cost concerns. One critical issue is reverse power flow, posing risks to maintenance personnel. Given that behind-the-meter solar panel installations operate independently of the grid and lack monitoring capabilities, there is a pressing need to develop methods for effectively monitoring their energy generation. An effective method can be the disaggregation of net load to solar panel generation and load consumption. Disaggregation provides utilities with valuable insights into customer behavior by revealing detailed energy consumption and generation patterns. This understanding allows utilities to customize services more effectively, improving customer satisfaction. Additionally, disaggregation aids utilities in planning and optimizing operations, enabling informed decisions on infrastructure investments and grid management strategies. In this study, the Gradient Boost Regression method is applied to disaggregate the net load of a dairy farm. Synthetic data from an Agent-based model of a dairy farm's electricity consumption is combined with data from the System Advisor Model for solar panel electricity generation. The experimental findings demonstrate that the proposed algorithm, when employed for disaggregating the net load of dairy farms equipped with solar panels, exhibits notable levels of accuracy and reliability.