We propose a novel method to model daily peak electricity demand using temperature and additional hourly and daily weather covariates, such as humidity and wind speed. Rather than enter into the temperature response function additively, the additional covariates may flexibly impact the demand response to temperature. Such flexibility allows differential responses to the actual temperature based on the heat index and wind chill factor, for example. Most notably, we find that ignoring humidity substantially underestimates the effect of high temperatures, while ignoring the effect of cloud cover overestimates the effect of low temperatures. Time of day also matters: a demand response to the same temperature may be different at different times of day. Moreover, accounting for weather-related covariates improves the model’s ability to explain daily peak demand.

我们提出了一种使用温度和额外的每小时和每日天气协变量（例如湿度和风速）来模拟每日峰值电力需求的新方法。附加协变量可以灵活地影响对温度的需求响应，而不是附加地进入温度响应函数。例如，这种灵活性允许基于热指数和风寒因子对实际温度做出不同的响应。最值得注意的是，我们发现忽略湿度大大低估了高温的影响，而忽略云层的影响则高估了低温的影响。一天中的时间也很重要：对相同温度的需求响应在一天中的不同时间可能不同。而且，