The NILM (Non-Intrusive Load Monitoring) field shows significant diversity in approaches and features, complicating performance evaluation and comparability. This work introduces a methodology to parameterize, optimize, and evaluate NILM algorithms, leveraging a general framework. Temporary appliance measurements and a consistent event/state definition enable an extended appliance database and reliable evaluation grounds. Focusing on event-based algorithms and data from a commercial building, the methodology proved effective, optimizing two event-detection algorithms for optimal appliance operation.

Access the paper here.