Evaluating Anomaly Detection Algorithms in Power Consumption Data

The quality of data is an important aspect when performing data scientific tasks.Having a clean ground truth dataset is critical to be able to derive analytical resultsfrom experiments. In this thesis, an automated method of checking the correctness of new data against adefined ground truth dataset is evaluated. With the use of machine learningalgorithms, anomaly detection was applied to separate normal and abnormalmeasurements of power consumption data, collected as time series from real worldhousehold appliances. Due to high variance in the energy data, the problem ofdetecting anomalies was solved with generative models, using the reconstructionerror as anomaly score. Extensive experiments were performed using a range ofparameters with three different models, simple regression with Multilayer Perceptron(MLP), Long Short-Term Memory (LSTM) and Dilated Causal Convolutional NeuralNetwork (DC-CNN). The results of the experiments show promising performance of using generativemodels for anomaly detection. All three models managed to learn the general powersignature of the household appliance measurements, with a varying degree of successrelative to the complexity of the signature. Out of the three models, the experimentsidentified the DC-CNN to be the best performing. Compared with the other models,the DC-CNN had both a higher success rate at classifying anomalous sequences aswell as a faster computational speed. Finally, this thesis concludes that fine-tuning the parameters of the models to thespecific task is required to achieve good performance. Finding a good combination ofmodel and parameter values is especially important in the case of handlingmeasurements from household appliances, due to the complexity of the data