Shaving the peaks through statistical learning: smart use of solar energy and storage solutions

摘要

This paper demonstrates how big-data, statistical learning and simulation of local energy production and storage, can contribute to reduce costs and shift energy consumption from the main power line to locally produced solar energy and battery storage during peak hours. This is demonstrated by using more than 5 million of hourly energy meters readings from 600 Norwegian grocery and large hardware stores. Many of the Norwegian grid operators use fixed peak-load tariffs, thus shaving the peaks will result in decreased energy costs. Our aim is to find the largest peaks; where the most potential for cost reductions can be found. To isolate the stores with the largest variation from hour-to-hour we suggest using the coefficient of variation (CV); we demonstrate this by calculating CV for 600 stores and use the results to rank and identify stores with both large variation and little variation in energy consumption. Further, three of these stores are used in solar photovoltaic (PV) production and energy storage simulations. The simulations will highlight the cost savings between stores with different CV values. Results suggest that by using such methodology, we can reduce total energy costs, and at the same time lower energy loads through peak shaving and phase shift.