A Neural Network-based Methodology for Non-Intrusive Energy Audit of Telecom Sites

Energy Signatures (ES) are simple multi-purpose energy audit techniques. For instance, ES are employed in i) the determination of the Balance Point (BP) of a building, ii) the ranking of heating or cooling systems efficiency of a building, iii) the provision of building diagnostic information, and iv) the estimation of potential savings and strategies for more energy efficient buildings. In this paper, we propose an innovative energy audit tool based on a Neural Network (NN) for determining ES from aggregated electric load profile of industrial sites. The energy audit methodology defines and applies an innovative Key Performance Indicator (KPI), called Temperature Unstandardised Beta Weight $(\beta_{T\mathrm{e}mp}^{*})$, to take into account not only the thermal behaviour of the building, but also the cooling system efficiency and the electrical base load. This energy audit has been applied on a real-case electric consumption pattern dataset of around sixty Central Offices from a Telecommunication service provider in Italy. The useful outputs from the proposed methodology, together with its simplicity, effectiveness and applicability, are intended to support diffused understanding of buildings thermal behaviour with the perspective of enhancing energy efficiency and consumption reduction.