Modulation training method of prediction model for smart grid FastADR power limitation of building air‐conditioners

Abstract

Fast automated demand response (FastADR) of building air-conditioners is a future smart grid demand-side technology. Neural network models will be useful to predict the power limitation result of the FastADR. However, neural networks require training data, which are collected by the experimental air-conditioning operations. In this letter, we propose a new training data collection method in which the FastADR-like signal is modulated into the normal air-conditioning operations.

Estimation on Averaging Effect of Large-scale Aggregation of DR Transient Responses of Building Multi Air-conditioning Facilities

“…Multitype building air-conditioners are usually controlled so as to maintain room temperature at a set level; therefore, it is difficult to acquire time-series response data related to FastADR power curtailment commands from normal operation data. Thus we used our previously developed method of FastADR power curtailment control [18][19][20][21] to remotely control actual multitype building air-conditioning equipment, and to collect time-series response data. Certain amount of response data is needed to evaluate responsivity of multitype building air-conditioners, which requires medium-to long-term data collection.…”

Dynamic SMOTE training of neural networks used in real‐time pricing control for building air‐conditioners

“…In this research, a combination of P L ( m ) in present and P L ( m + 1) in next frame is grouped to five patterns. Figure shows the distribution of 1330 data observed from an actual building in august, 2016 . In this figure, it can be found that the number of samples of the ‘LoHi’ pattern is significantly less than that of other patterns.…”

Estimation on averaging effect of large‐scale aggregation of DR transient responses of building multi air‐conditioning facilities

“…Multitype building air‐conditioners are usually controlled so as to maintain room temperature at a set level; therefore, it is difficult to acquire time‐series response data related to FastADR power curtailment commands from normal operation data. Thus we used our previously developed method of FastADR power curtailment control to remotely control actual multitype building air‐conditioning equipment, and to collect time‐series response data.…”