Reconstituted Data-Driven Air Conditioning Energy Consumption Prediction System Employing Occupant-Orientated Probability Model As Input and Swarm Intelligence Optimization Algorithms

Chengyu Zhang,Liangdong Ma,Han Xing,Tianyi Zhao
DOI: https://doi.org/10.1016/j.energy.2023.129799
IF: 9
2024-01-01
Energy
Abstract:With recent energy and environmental crises, the energy consumption prediction of air conditioning (AC) is essential. Notably, existing general prediction systems use the air temperature, real-time AC usage, time, oc-cupancy, and AC historical energy consumption as inputs and employ feedforward neural networks. However, such systems face certain challenges. First, the input system may be difficult to reflect the real-time interaction among the building environment, occupants, and energy system; Second, the algorithms need to be upgraded to further improve accuracy, speed up convergence, and avoid overfitting. his paper introduces a novel parameter (group occupant behavior probability, GOBP) in the input system; and uses swarm intelligence algorithms with chaotic mapping and adaptive weight adjustment as optimization with LSTM and BP as the basic algorithms. The verification results obtained from five different buildings indicate the following: (1) The novel prediction system improved prediction performance with 20.08%-78.37 % of MAPE and 3.99%-53.28 % of CV-RMSE, however, it will increase 148.70s-1065.11s of calculation time. (2) For accuracy, using GOBP will improve the 0.30%-51.87 % and 0.18%-64.77 % by using swarm intelligence optimization algorithms. (3) This paper discusses the priority of selecting suitable inputs, algorithms, and optimization to combine a whole prediction system.
What problem does this paper attempt to address?