home › Annex 37

Smart Design and Control of Energy Storage Systems

  • Annex 37
  • Currently running
  • Energy storage in energy systems



In this Annex, we investigate the present situation of smart design and control strategy of energy storage systems for both demand side and supply side. The research results will be organized as design materials and operational guidelines. Specifically, artificial intelligence that has developed significantly in recent years can be expected to make a significant contribution to the smart design and control systems. This Annex also covers the availability of artificial intelligence.


The final objective of this Annex is to address the design/integration, control, and optimization of energy storage systems with buildings, districts, and/or local utilities. In order to realize optimal control, the constraints must be properly predicted and the system must first be optimally designed. For designing the system more optimally, it is necessary to properly understand the performance of the components. Therefore, the focus here is to model components, develop design methods and advanced control strategies for effectively predicting, evaluating, and improving the performance of buildings and districts when energy storage is available.

■ Establishment of prediction method

In order to properly design and control the system, it is necessary to predict various conditions. For example, the prediction of renewable energy production and the prediction of electricity price change are boundary conditions for appropriate control. The prediction of energy demand in buildings and districts is a constraint that must be satisfied in control. Especially, the energy demand of buildings and districts varies significantly depending on the usage, composition of the energy system, change of weather, occupant behavior, and so on. There are many demand prediction models are previously proposed including IEA’s achievement. In this Annex, we will summarize those findings and examine the establishment of prediction methods that contribute to smart design and control.

■ Establishment of modeling method of component and system

To optimally design and control different energy systems depending on the building, it is necessary to construct a prediction model that reproduces system behavior. Specifically, performance prediction models of the system and its components such as heat pumps, pumps, and energy storage devices are required. Various components and systems have already been modeled in the previous IEA Annexes. However, in order to use it for actual real-time control, it is necessary to predict fast and accurately. In addition, it is necessary to reproduce the deterioration over time the change in performance of the component. The recent development of artificial intelligence enables them to be realized.

■ Establishment of optimization method for design and control

Optimization technologies are very useful to both design and control. The design act is to determine a specification that maximizes the functionality to achieve a certain purpose under limitation of resources. In addition, when there are multiple objectives and they are in a trade-off relationship with each other, finding a compromise between them is also a design act. In order to minimize costs or fossil energy consumption or maximize human comfort, how to arrange various devices including heat storage devices is a major design issue. On the other hand, control/operation is to decide which devices, when and how we should activate for a certain purpose. These are both optimization problems. Previously, the system was simple, so these problems could be handled by human designers and operators. Today, however, the system has become more and more sophisticated and complex, it is almost impossible to handle these problems with human abilities. Again here, artificial intelligence is a very useful tool. Conventionally, many optimization methods based on mathematical programming such as linear programming have been proposed. Recently, optimization methods based on metaheuristics and deep-Q learning have attracted attention as highly generic methods. Therefore, investigation of an optimization method that can efficiently solve such difficult optimization problem is important from the viewpoint of realization of optimum design and control. This Annex deals with the following subjects.

  • Classification of optimization methods (mathematical programming, metaheuristics, deep-Q learning)
  • Strong and weak points of different optimization methods: in terms of search efficiency and stability
  • Setting of objective function for optimal control problems

Operating agent(s)

  • Ryozo Ooka, PhD
  • Institute of Industrial Science, The University of Tokyo, Japan


For requests and information about Annex 37, please use the contactform below.