Crowdsourced Electricity Demand Forecast

We propose a new approach to forecasting the demand for a commodity in which the commodity supplier asks each consumer to forecast its own demand for the commodity and in return gives a monetary reward that is proportional to the accuracy of the forecast. Such an approach might be applicable when consumer demand for a perishable commodity is uncertain and forecast error leads to waste for producers or suppliers. We apply this approach to the forecasting of residential electricity demand over 24 hours, i.e., short-term load forecasting (STLF). Accurate STLF is vital to meeting the large daily fluctuations in the demand for electricity in a reliable and economical way. Improved STLF accuracy might reduce the variable costs incurred by power system operators and energy retailers through more precise generation scheduling and energy purchasing. We propose a new method to model both the true demand profiles for individual residential electricity consumers, and the consumer forecasts of those demand profiles. We study the error associated with the resulting forecast of aggregate demand as a function of the model parameters.