Facilities Management (FM) companies can use load monitoring of electrical appliances (assets) to track energy consumption and predictive maintenance. Reliable algorithms are needed to automatically identify or verify appliances through their energy signatures to improve efficiencies during installation and inspection tasks. Most approaches rely on Voltage-Current (V-I) trajectory. These features are extracted from steady-state current and voltage signals. However, these methods often assume signals are uniformly sampled. In real-world conditions, this assumption does not always hold, leading to misclassified steady-state events when signals are noisy. This paper introduces a novel feature extraction and classification pipeline to ensure the validity of detected steady-state events. The approach measures the approximate entropy of current signals and their correlation with voltage to extract denoised features for appliance type classification. The proposed pipeline is evaluated on a large-scale real-world operational dataset spanning multiple appliance categories. We demonstrate that the extracted denoised features significantly improve the performance of Machine Learning (ML) models used for appliance type classification. Finally, we present a deployment framework for FM settings, enabling digital cataloguing of appliances informing businesses on sustainable choices for appliance requirements.