Load identification optimization deployment method based on knowledge distillation

The amount of residential customers participating in grid demand-side response has been rising in recent years as new power systems have been developed. The secret to home users participating in demand-side response is load identification technology. The majority of approaches now only have theoretical research applications and no actual deployment. Methods for deploying targeted optimization are lacking. So, using knowledge distillation, this research develops a load detection optimization deployment approach. This article performs knowledge distillation on MobileNetV2. Build a model that is lighter while making sure that the performance impact is essentially same. Through an investigation of calculation examples using independently gathered laboratory data sets, the efficiency of the suggested method was demonstrated. In order to confirm the viability, the storage size in the embedded system was lastly verified.