Residential battery sizing model using net meter energy data clustering

The high upfront costs of batteries have limited the investment in retrofit residential energy storage systems for solar customers. Battery size is one of the most important factors that impact the financial return since it determines the major operational capabilities of the solar-coupled storage system. To select the optimal battery size for a photovoltaic solar customer, it is important to perform an analysis taking account of the customer's on-site generation and consumption characteristics. However, in most cases there are insufficient pre-existing data of the required quality making it difficult to perform such analysis. In this paper, we propose a model that can achieve satisfactory battery sizing results with a limited amount of net meter electricity data. The model uses K-means clustering on customer net meter electricity data to discover important information to extrapolate limited input net/gross meter energy data and uses this in a techno-economic simulation model to determine the optimal battery size. The approach is validated using a set of 262 solar households, two tariff structures (flat and Time-of Use) and a naive forecasting method as a comparison to the proposed model. The results indicate that the proposed model outperforms the alternative baseline model and can work with as little as one month of net meter energy data for both of the evaluated tariff structures. On average, the model results in 0.10 normalised root mean squared error in yearly battery savings and net present values, 0.07 normalised root mean squared error in annual electricity costs and a r-squared value of 0.717 in finding the optimal size of batteries. Moreover, this study reveals a linear correlation between the used clustering validity index (Davies-Bouldin Index), and errors in estimated annual battery savings which indicates that this index can be used as a metric for the developed battery sizing approach. With the ongoing rollouts of net meters, the proposed model can address the data shortage issue for both gross and net meter households and assist end users, installers and utilities with their battery sizing analysis.