In electricity distribution networks, one of the challenges is to identify non-technical losses. The challenge is even greater in rural distribution networks. These have large extensions and the costs for on-site inspection are higher. Irrigated rice crops have particular characteristics in terms of consumption, due to the seasonality of the crop, different irrigation modes and processes, and climatic characteristics. This article presents a methodology that aggregates the consumption recorded by the concessionaires with more relevant information to analyze how much a given consumer presents risks of non-technical losses, as well as its neighbors. This work presents the use of Artificial Intelligence in the identification of non-technical losses in rural areas with integrated irrigation systems. The technique considers robust sub-methodologies, which evaluate: meteorological data, satellite images of the region, technological mapping of crops and the calculation of the energy balance for the estimation of non-technical loss by distribution feeder. The methodology was applied in consumer units located on the western border of RS. These consumer units use the flood irrigation system for rice planting, so they have a high monthly consumption. This high consumption significantly impacts the energy utility, in the event of a PNT. The results were obtained and validated with real information from rice harvests between 2018 and 2021.