To this day, hyperparameter tuning remains a cumbersome task in Non-Intrusive Load Monitoring (NILM) research, as researchers and practitioners are forced to invest a considerable amount of time in this task. This paper proposes adaptive weighted recurrence graph blocks (AWRG) for appliance feature representation in event-based NILM. An AWRG block can be combined with traditional deep neural network architectures such as Convolutional Neural Networks for appliance recognition. Our approach transforms one cycle per activation current into an weighted recurrence graph and treats the associated hyper-parameters as learn-able parameters. We evaluate our technique on two energy datasets, the industrial dataset LILACD and the residential PLAID dataset. The outcome of our experiments shows that transforming current waveforms into weighted recurrence graphs provides a better feature representation and thus, improved classification results. It is concluded that our approach can guarantee uniqueness of appliance features, leading to enhanced generalisation abilities when compared to the widely researched V-I image features. Furthermore, we show that the initialisation parameters of the AWRG’s have a significant impact on the performance and training convergence.

Research on smart grid technologies is expected to result in effective climate change mitigation. Non-Intrusive Load Monitoring (NILM) is seen as a key technique for enabling innovative smart-grid services. By breaking down the energy consumption of households and industrial facilities into its components, NILM techniques provide information on present appliances and can be applied to perform diagnostics. As with related Machine Learning problems, research and development requires a sufficient amount of data to train and validate new approaches. As a viable alternative to collecting datasets in buildings during expensive and time-consuming measurement campaigns, the idea of generating synthetic datasets for NILM gain momentum recently. With SynD, we present a synthetic energy dataset with focus on residential buildings. We release 180 days of synthetic power data on aggregate level (i.e. mains) and individual appliances. SynD is the result of a custom simulation process that relies on power traces of real household appliances. In addition, we present several case studies that demonstrate similarity of our dataset and four real-world energy datasets.

Non-Intrusive Load Monitoring (NILM) comprises of a set of techniques that provide insights into the energy consumption of households and industrial facilities. Latest contributions show significant improvements in terms of accuracy and generalisation abilities. Despite all progress made concerning disaggregation techniques, performance evaluation and comparability remains an open research question. The lack of standardisation and consensus on evaluation procedures makes reproducibility and comparability extremely difficult. In this paper, we draw attention to comparability in NILM with a focus on highlighting the considerable differences amongst common energy datasets used to test the performance of algorithms. We divide discussion on comparability into data aspects, performance metrics, and give a close view on evaluation processes. Detailed information on pre-processing as well as data cleaning methods, the importance of unified performance reporting, and the need for complexity measures in load disaggregation are found to be the most urgent issues in NILM-related research. In addition, our evaluation suggests that datasets should be chosen carefully. We conclude by formulating suggestions for future work to enhance comparability.

Non-Intrusive Load Monitoring (nilm) techniques are increasingly becoming a key instrument for identifying the power consumption of individual appliances based on a single metering point. Particularly Deep learning models are gaining interest in this regard. However, the challenges brought by NILM datasets and the non-availability of common experimental guidelines tend to compromise comparison, research transparency and replicability. The limited adoption of efficient research instruments and a lack of best practices guidelines contribute in huge part to this problem, where no features, encouraging standardised formats for benchmarking and results sharing, are offered. To address these issues, we first present a brief overview of recent best practices for Deep Learning (DL) and highlight how deep NILM research can benefit from these practices. Furthermore, we suggest a novel open-source toolkit leveraging these practices: Deep-NILMTK. The proposed toolkit offers a common testing bed for NILM algorithms independently of the underlying deep learning framework with a modular NILM pipeline that can easily be customised. Furthermore, Deep-NILMTK introduces the concept of Experiment Templating to offer pre-designed experiments allowing to enhance research efficiency. Leveraging this concept and DL best practices, we present a case-study of creating an online NILM benchmark repository ^{1 1}https://github.com/BHafsa/DNN-NILM-benchmark considering eight of the most popular deep NILM algorithms. All sources relative to the tool are made publicly available on Github ^{2 2}https://github.com/BHafsa/deep-nilmtk-v1 along with the corresponding documentation.