The abstract makes it make more sense with clearer language:

The significance of heating load (HL) accurate approximation is the primary motivation of this research to distinguish the most efficient predictive model among several neural-metaheuristic models. The proposed models are through synthesizing multi-layer perceptron network (MLP) with ant lion optimization (ALO), biogeography-based optimization (BBO), dragonfly algorithm (DA), evolutionary strategy (ES), invasive weed optimization (IWO), and league champion optimization (LCA) hybrid algorithms. Each ensemble is optimized in terms of the operating population.