A Novel Method of Irrigation Detection and Estimation of the Effects of Productive Electricity Demands on Energy System Planning

Abstract:

In many parts of Sub-Saharan Africa, growth of irrigated agriculture, particularly higher value horticulture, may address shortcomings in both nutrition and farmer income. Neither utilities nor the private sector is fully prepared for the scale or geospatial distribution of the associated productive electricity demand, demand which will likely influence energy system planning decisions, especially if it can be paired with low-cost generation.

To determine the distribution and impact of electricity demands for irrigation, we demonstrate a novel method of ascertaining the location of off-season vegetation growth that uses satellite imagery and gridded rainfall estimates to extract relevant vegetation endmembers. After finding the contribution of each endmember to overall vegetation timeseries, we deploy machine learning techniques to predict irrigation presence. We validate this approach using cropland data from California and demonstrate results for Ethiopia at 10m and 250m spatial resolutions. Because our approach controls for cropping seasonality using rainfall data, it is transferable to parts of the world where off-season vegetation growth is limited by the availability of water; as such, limited region-specific ground truth is required to predict the location of irrigation systems which enable off-season cropping. We compare our results to other publicly available irrigation prediction products, including the International Water Management Institute’s global irrigated area map.

Using our predictions of irrigation presence, we next evaluate the impact of corresponding agricultural electricity demands on optimal grid planning. We specifically consider local solar resources and examine if the temporal alignment of generation and demand – made possible by the flexible nature of irrigation loads – reduces the need for more expensive storage or back-up dispatchable generation. We show least-cost electrification options for a set of agricultural demand scenarios, comparing results to a baseline case without any productive loads. As all the data used and code developed for this work are open-source, we provide a methodology for irrigation detection and system planning that stakeholders and local planners can adopt at no cost.