A Parallel Evolutionary Computing-Embodied Artificial Neural Network Applied to Non-Intrusive Load Monitoring for...

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Energy Systems

Background:

• Non-intrusive load monitoring (NILM) enables appliance identification from aggregated electrical signals without individual meters.
• Accurate NILM is crucial for effective Demand-Side Management (DSM) in smart homes.
• Training Artificial Neural Networks (ANNs) for NILM can be computationally intensive due to large datasets.

Purpose of the Study:

• To develop and evaluate a parallel Genetic Algorithm (GA)-embodied Artificial Neural Network (ANN) for NILM.
• To improve the efficiency and reduce the computational cost of training ANNs for NILM.
• To apply the developed method to a Home Energy Management System (HEMS) in a real residential setting for DSM.

Main Methods:

• Integration of a parallel GA with an ANN for load disaggregation.
• Utilizing evolutionary computing (GA) to optimize ANN training.
• Employing a divide-and-conquer strategy for parallel computation to accelerate ANN model evolution.

Main Results:

• Demonstrated the feasibility and effectiveness of the parallel GA-embodied ANN for NILM.
• Significantly reduced the execution time for training ANN models using parallel computing.
• Successful application in a real residential Home Energy Management System (HEMS).

Conclusions:

• The parallel GA-embodied ANN is a viable and efficient approach for NILM in HEMS.
• This method offers a cost-effective solution for appliance identification and DSM.
• Parallel computing significantly enhances the training process for NILM applications.