Evaluation of effectiveness and resources consumption of water mist spray systems in Mediterranean areas by predictions based on LSTM Recurrent Neural Networks

Abstract

To counter the increasing urban overheating, climate adaptation solutions are proposed. Among them, water mist spray recently acquired particular attention, due to its efficiency, cost-effectiveness, and versatility. However, spray devices require a large amount of water and energy to cool even limited areas, thus their environmental costs/benefits ratio should be carefully evaluated. This study analyses cooling benefits and resources consumption of mist devices in 11 cities within 3 climate contexts, through Recurrent Neural Networks (RNNs) trained with experimental data. RNNs predict the expected time series of thermal benefits and of energy and water consumptions, also considering different design solutions of devices. Results show that when sun/wind shielding is used in the sprayed area, or the height of nozzles is limited, higher cooling results are obtained. However, energy and water consumption are extremely high if misting systems are perennially active during the day. Considering all simulated conditions, the predicted average daily energy to obtain a unitary variation of the Mediterranean Outdoor Comfort Index is 4,17 Wh/m2, while the corresponding average daily volume of water is 0,56 l*h/m2. These results confirm the need for applications managed by control logics based on the acquisition of real-time climatic data to reduce the environmental loads.