A utility company will forecast energy usage for residential and commercial customers using 10 years of historical consumption plus features like weather, occupancy, and holidays. Using a managed forecasting service, which algorithm is most appropriate for this scenario?

CNN-QR is ideal for this scenario because it's a deep learning algorithm designed for large-scale, high-frequency time series forecasting, which aligns with 10 years of detailed energy consumption data and multiple features. It can capture complex non-linear relationships and interactions between features like weather and holidays. Its quantile regression capability allows for probabilistic forecasts, which are crucial for understanding the uncertainty in energy demand. ARIMA and ETS are traditional statistical methods better suited for simpler, univariate time series without many exogenous variables. Prophet is good for time series with strong seasonality and trend but may not handle the complexity and scale as effectively as CNN-QR for this multivariate, long-term forecasting task.