In industrial environment, plants are exposed to multiple contextual factors, which affect the way that energy is used. Actual energy models provide guidelines for energy consumption, but fail in explaining some patterns observed in practice. This paper provides a novel approach for building energy models of industrial plants, based on the influence of context variables. Multiple RLS algorithms are used for multi-model context-based estimation of energy consumption. For context identification, regression trees are used to divide the context variables into regions, where the local models are estimated. The effectiveness of the proposed approach is illustrated with some examples and validated with a practical case in a real industrial cement plant.