Anomaly detection and visualization of school electricity consumption data

Abstract

Abstract-Anomaly detection has been widely used in a variety of research and application domains, such as network intrusion detection, insurance/credit card fraud detection, health-care informatics, industrial damage detection, image processing and novel topic detection in text mining. In this paper, we focus on remote facilities management that identifies anomalous events in buildings by detecting anomalies in building energy data. We have investigated five models to detect anomalies in the school electricity consumption data. Furthermore, we propose a hybrid model which combines polynomial regression and Gaussian distribution. Based on this model, we have developed a data detection and visualization system for a facilities management company to detect anomalous events in school electricity facilities. The system is tested and evaluated by the facilities managers of the company. According to the result of the evaluation, it reduces the effort required by facilities managers to identify anomalous events in school electricity facilities.

A New Anomaly Detection System for School Electricity Consumption Data

“…Our system is used to reduce this tedious and time-consuming eyeballing of the data. Parts of the results of the model investigation and the system development contributed to anomaly detection of school electricity consumption data have appeared in [8]. However, our research in this paper significantly differs from the previous work in the following aspects:…”

A Novel Approach for Detecting Anomalous Energy Consumption Based on Micro-Moments and Deep Neural Networks

“…[27] Difference between real and Neural networks-ARIMA Supervised predicted consumption Ieracitano et al [34] Statistical features Autoencoder base LSTM Unsupervised Ramchandran et al [36] Raw images and detected edges Convolutional DNN-based Unsupervised autoencoder and LSTM Janetzko et al [46] Power spectrum Clustering and visual analytics Unsupervised Ma el al. [47] Standard deviation of temporal FCD-POD-LSE Unsupervised coefficients Cui and Wang [48] Power consumption time series Polynomial regression and Semi-supervised Gaussian distribution Buzau et al [52] Historic and non-sequential power data DNN-based LSTM and MLP Supervised Lin and Claridge [54] Power consumption deviation between measured Unsupervised and temperature and simulated consumption Araya et al [55] Contextual/behavioral features Detecting incomplete data in Unsupervised sliding window Liu et al [56] Power and temperature Lambda architecture Supervised Current paper Micro-moment consumption and DNN and rule-based algorithm Supervised occupancy data Fig. 1 Block diagram of the proposed system for detecting abnormal energy consumption…”

“…Thus, they suggest integrating other intelligent algorithms. In [48], abnormal events in a school building are appointed by capturing anomalous patterns occurred in electrical consumption. Accordingly, Cui and Wang have explored a hybrid system that merges the polynomial regression and Gaussian model to detect abnormal consumption observations.…”

Detecção de outliers no consumo de energia elétrica

“…O aumento na demanda de energia elétrica está relacionada a sua utilização nos mais diversos setores, o que traz benefícios econômicos e melhoria da qualidade de vida, por outro lado também implica em esgotamento dos recursos utilizados para a produção de energia, quando utilizadas fonte não renováveis, e em impacto ao meio ambiente [1][5] [11]. Tem-se estimulado o uso eficiente da energia elétrica, porém o modelo tradicional do sistema elétrico e do registro de informações de consumo, não propicia aos consumidores informações suficientes que permitam decisões mais efetivas a respeito desse consumo.…”

“…Analisar os dados de consumo de energia elétrica permite a detecção de defeitos, entendimento do padrão de consumo e oportunidades para melhoria da eficiência energética [8]. Neste contexto, a identificação de consumos que diferem dos normalmente observadosé uma oportunidade para detectar os que não são autorizados, provocando diminuição de custos [11].…”