Overcoming the challenge of complexity: a new data analytics framework for power and water demand forecasting

A major challenge in managing critical infrastructure such as power grids and water supply systems is to continually balance generation and demand. A reliable forecast is essential to optimize production and maintain grid stability. However, dynamics of power generation and consumption in modern grids are becoming increasingly difficult to predict due to dependencies on meteorological and socio-economic factors. Here we present a new data analysis framework designed to overcome such complexity. The forecast itself is generated using nonlinear time series analysis combined with machine learning. However, to reduce complexity, the forecast is made either by strictly univariate analysis or after filtering by causal interference analysis. The method thus provides good forecasts even for complex, high-dimensional situations in which classic methods usually fail. We illustrate the performance of the method using real data from the most important use cases of load and renewable energy forecasting (see https://24insight.zonos.de/ for a live demo).