Determining Energy Production and Consumption Signatures Using Unsupervised Clustering

The selection of the peak power of a photovoltaic system to meet the energy demand of a building is a key task in the energy transformation. This article presents an algorithm for assessing the correctness of the selection of a photovoltaic system with a peak power of 50 kWp for the needs of a university administration building. This is made possible due to the use of an advanced photovoltaic inverter, which is a device of the Internet of Things and the smart metering system. At the beginning of the review, the authors employed the naked eye measurement data of the time series related to the power production by the photovoltaic system and its consumption by the university building. Then, traditional statistical analyses were performed, characterizing the generated power divided into self-consumption power and that fed into the power grid. The analysis of the total consumed power was performed with the division into the power produced by the photovoltaic system and that taken from the power grid. The analyses conducted were subjected to expert validation aimed at explaining the nature of the behavior of the power generation and consumption systems. The main goal of this article is to determine the signatures of the power generated by the photovoltaic system and consumed by the administration building. As a result of unsupervised clustering, the power generation and consumption space were divided into states. These were then termed based on their nature and their usefulness in managing the power produced and consumed. Presentation of clustering results in the form of heatmaps allows for localization of specific states at specific times of the day. This leads to their better understanding and quantification. The signatures of power generated by the photovoltaic system and consumed by the university building confirmed the possibility of using an energy storage system. The presented computational algorithm is the basis for determining the correctness of the photovoltaic system selection for the current energy needs of the building. It can be the basis for further analysis related to the prediction of both the power generated by Renewable Energy Sources and the energy consumed by diverse types of buildings.