Why does Building Energy Optimization fail and how to avoid it?

The fast growth in population is equally responsible for the increased carbon footprint we observe today. To be precise, there exist a direct proportionality relationship between population and resource utilization. Among various resources, the building sector has become extremely energy intensive; statistically being responsible for approximately 39% of the primary energy used worldwide. In this scenario, building designs that emphasize energy efficiency are therefore significant for achieving energy conservation and reducing negative environmental impact.

As shown in several important review works, a new technique known as building energy optimization (BEO) has become a very active research field credit to its ability to simulate energy consumption and automatically search for the best energy efficient design. Algorithms employed in this technique are critical and as the authors stated “they are the heat and soul”. Algorithms can be grouped into three: direct search algorithms, intelligent optimization algorithms, and hybrid algorithms.

Research has shown that these algorithms play a critical role in determining the effectiveness and efficiency of BEO techniques. Nonetheless, an effective optimization algorithm should simultaneously meet three criteria as outlined in the original work referenced here. However, it is rather unfortunate that the BEO optimization algorithm is vulnerable to being ineffective.

In this context, Southeast University researchers: Binghui Si (PhD candidate), Zhichao Tian (PhD candidate), Dr. Xing Jin, Dr. Xin Zhou and led by Professor Xing Shi studied carefully the reasons for the ineffectiveness in the popular building energy optimization technique. In particular, they focused on assessing and hopefully adequately addressing inherent problems in the stated technique, such as; firm definition of algorithm ineffectiveness, the reasons that cause an algorithm to be ineffective, and its risk of becoming ineffective. Their work is currently published in the research journal, Renewable Energy.

In brief, to investigate the causes of ineffectiveness, four commonly used optimization algorithms were selected and applied to a representative BEO problem to perform numerical experiments. Specifically, they began with a systematic definition of optimization algorithms’ ineffectiveness, describing five ineffective scenarios. Then, a reference building and a representative energy optimization problem were proposed. Four commonly used optimization algorithms, namely, discrete Armijo gradient, Hooke-Jeeves, particle swarm optimization with constriction coefficient and particle swarm optimization with inertia weight, were then tested to determine the circumstances and the causal factors under which they become ineffective.

The authors reported that the effectiveness of the discrete Armijo gradient algorithm was dependent on its control parameters and the position of the initial solution in the design space. Additionally, for the Hooke-Jeeves algorithm, the position of the initial solution in the design space was observed to be crucial to the effectiveness of the algorithm because of its poor ability to escape from local optima.

In summary, the study Professor Xing Shi and his research team focused on addressing whether different algorithm control parameter settings and different initial solutions could cause the four most common algorithms to be ineffective when used in BEO. For this purpose, four sets of different control parameter settings and four different initial solutions were randomly chosen for each test. Overall, the research presented by Southeast University scientists shed more light on the performance of algorithms in BEO and revealed their failure mechanisms. The results can be very helpful to architects, engineers, and designers who use BSO and want to avoid ineffective algorithms. Since BEO is an emerging design technique for green buildings, the green building design in general can also benefit significantly from the results.