Modelling the Disaggregated Demand for Electricity in Residential Buildings Using Artificial Neural Networks (Deep Learning Approach)

Abstract

The paper addresses the issue of modelling the demand for electricity in residential buildings with the use of artificial neural networks (ANNs). Real data for six houses in Switzerland fitted with measurement meters was used in the research. Their original frequency of 1 Hz (one-second readings) was re-sampled to a frequency of 1/600 Hz, which corresponds to a period of ten minutes. Out-of-sample forecasts verified the ability of ANNs to disaggregate electricity usage for specific applications (electricity receivers). Four categories of electricity consumption were distinguished: (i) fridge, (ii) washing machine, (iii) personal computer, and (iv) freezer. Both standard ANNs with multilayer perceptron architecture and newer types of networks based on deep learning were used. The simulations included over 10,000 ANNs with different architecture (number of neurons and structure of their connections), type and number of input variables, formulas of activation functions, training algorithms, and other parameters. The research confirmed the possibility of using ANNs to model the disaggregation of electricity consumption based on low frequency data, and suggested ways to build highly optimised models.

Non-Intrusive Load Monitoring by Voltage–Current Trajectory Enabled Transfer Learning

“…Numerous researchers are actively working on developing unique appliance signature databases based on various techniques, including graph signal reconstruction [223], mixed-integer nonlinear programming [224], generative adversarial networks (GANs) [135], admittance-based algorithms [173], Karhunen-Loève Expansion [225], and K-means clustering [215]. With the availability of established load signature databases, pattern recognition-based NILM primarily relies on techniques such as graph signal processing (GSP) [226], [227], dynamic time warping (DTW) [175], [228]- [232], soft dynamic time warping (sDTW) [233], Hungarian matching [234], global alignment kernel (GAK) [233], all common subsequences (ACS) [233], and minimum variance matching (MVM) [235] for pattern matching. In today's digital age, patterns are ubiquitous, manifesting themselves statistically through algorithms or in physical forms.…”

Recent trends of smart nonintrusive load monitoring in buildings: A review, open challenges, and future directions

“…Other ML models : Other ML models, such as the conventional classification algorithms, are still receiving the attention of the NILM research community. A plethora of research studies, R&D NILM projects, and even commercial solutions are developed based on ANN, 137 extreme learning machine (ELM), 138 multilayer perceptron (MLP), 139 radial basis function neural network (RBFNN), 140 KNN, 83 SVM, 141 ensemble bagging tree (EBT), 7 and RF 142 . Additionally, many NILM studies are built upon using clustering techniques 143 .…”

Benefits of three-phase metering for load disaggregation

“…A comparison of performance of four different NILM algorithms on this dataset is presented in [3], where it was concluded that to achieve adequate results, a supervised approach was required. An Artificial Neural Networks (ANNs)based approach for disaggregation of the ECO dataset is proposed in [10], where data were resampled to 10 min intervals. Again, only the aggregate signal of the phases was used as an input.…”