Effective identification of distributed energy resources using smart meter net‐demand data

Abstract

International policies and targets to globally reduce carbon dioxide emissions have contributed to increasing penetration of distributed energy resources (DER) in lowvoltage distribution networks. The growth of technologies such as rooftop photovoltaic (PV) systems and electric vehicles (EV) has, to date, not been rigorously monitored and record keeping is deficient. Non-intrusive load monitoring (NILM) methods contribute to the effective integration of clean technologies within existing distribution networks. In this study, a novel NILM method is developed for the identification of DER electrical signatures from smart meter net-demand data. Electrical profiles of EV and PV systems are allocated within aggregated measurements including conventional electrical appliances. Data from several households in the United States are used to train and test classification and regression models. The usage of conventional machine learning techniques provides the proposed algorithm with fast processing times and low system complexity, key factors needed to differentiate highly variable DER power profiles from other loads. The results confirm the effectiveness of the proposed methodology to individually classify DER with performance metrics of 96% for EV and 99% for PV. This demonstrates the potential of the proposed method as an embedded function of smart meters to increase observability in distribution networks. K E Y W O R D Sdemand-side management, distributed energy resources, non-intrusive load monitoring, smart grids, supervised machine learning methodsThis is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Impact Assessment of NILM Methods for an Enhanced Observability of Low Voltage Distribution Networks

“…Offline models contribute to DER identification in electric systems where there is not a direct measurement of independent circuits containing a DER. As it was demonstrated in [3], a NILM model was trained with some aggregated household loads and the test was done with data from a different house. However, to increase performance metrics of the general NILM methods, a large, annotated, and balanced dataset with disaggregated loads is required for the purpose of training the model.…”

“…In this regard, powerful computational sources could be used to handle large datasets to create accurate models and to improve computational times before the algorithms are loaded to the meters. Subsequently, smart meters equipped with trained NILM models could identify DER electrical patterns in real time considering the scoring times achieved in scientific studies (in the order of milliseconds in several NILM methods [3]), which represent faster rates than the typical high resolution temporal sampling from smart meters from 15 to 60 minutes [13], [15]. Therefore, DER identification data could be attached to the periodic information transmitted from each household to utility companies with low impact on data storage requirements or transmission data rates.…”

“…The literature offers multiple studies covering the technical performance of non-intrusive load monitoring (NILM) methods for the detection of DER from aggregated power measurements. For example, research studies at both residential [3]- [5] and LV distribution levels [6], [7] propose machine learning algorithms (e.g., k-nearest neighbour or artificial neural networks) to technically evaluate for classification and regression approaches that recognise electric vehicle (EV) and photovoltaic (PV) electric power signatures among other electric appliances. Table I summarizes the technical performance of these models for classification and regression.…”

A hybrid data-driven online solar energy disaggregation system from the grid supply point

“…This method treats energy disaggregation as a classification problem by finding the boundary between the demand load and the PV generation. Based on the results in [19], KNN and RF regression models are proposed by combining Principal Component Analysis (PCA) techniques [20]. PCA is a technique for reducing the dimensionality of input features while minimizing information loss.…”

“…From Table 2, following models in the literature are employed as the benchmark models in this work: the conventional model-based method proposed by [27], upscaling method introduced in [16], machine learning-based methods, including KNN + PCA [20], SVM + PCA [19], RF + PCA [20], GBRT [22], deep learning-based methods, i.e., MLP [23], CNN [24], LSTM [25], and GRU [25].…”