Design and assessment of optimized control strategies for solar heating and cooling system

Solar Combi+ systems are thermal systems used for satisfy the thermal loads for DHW, heating and cooling using solar thermal energy. They are normally complex systems made of an elevate number of components. These components are required to operate correctly and harmoniously in order to achieve high system performance. The traditional design process of these applications is based on numerical simulations, which are used for understanding the performance of the system with respect to different parameters (geometrical dimensions, thermal capacity, orientation) or boundary conditions (thermal loads or climatic data considered). At the same time, the optimization of the control strategy is developed using the same models that, in order to represent the reality, need to be validated with monitoring data or laboratory tests. The process of numerical model validation, sensitivity analysis and optimization of the control is a time consuming process. In order to explore the possibility of reducing the simulation effort, an alternative approach has been developed in this thesis based on another approach. In the beginning, a rough numerical model is created and used for the design of the controller without undergoing the model validation steps. The fine-tuning of the controller, instead, is done directly on field where the control parameters are optimized using an appropriate learning methodology. Following this new approach, time saving are realized because the optimization of the control is fast and the validation of the model components is skipped. This makes the control optimization process more high-level without requiring access to monitoring data, laboratory test or deep knowledge of the system. Moreover, the selection of the software adopted for the development of the controller (LabVIEW) allow to share the same code for real-time control implementation as well as the numerical simulation. A direct connection with the software that simulate the energy fluxes between different components (Trnsys) and the software of the controller (LabVIEW) has been developed. Two innovative control methodologies are discussed in the thesis. The first one is related to the design of the controller using the fuzzy logic (Fuzzy controller) the other one is relative to the ability of the controllers to optimize themselves from the actual data taken from the field (Q-learning). These two methods can be efficiently connected each other allowing for a robust model-free learning approach able to find the best operative point of the fuzzy controller. Following this path, the thesis here presented is composed by a the first chapter where an introduction of the solar thermal systems from the simple configuration (only for DHW demand) to the more complex systems (solar Combi+) is given. In the end of the chapter a review on the logic used in the control of such a systems is reported. In the second chapter the numerical models adopted are presented, with a focus on the boundary conditions considered and the effects on the system loads. The third chapter is devoted to present the methodology of validation developed and its application to the models used in the simulation. In the fourth chapter a Solar Combi+ system for residential application is presented with a focus on the control structure, the sensitivity analysis and the traditional control optimization process. Finally in the fifth chapter the advanced control logic methodologies are presented with an introduction on Fuzzy Logic and on Reinforcement Learning. The chapter continues with a focus on the Fuzzy logic controller and the Q-learning controller algorithm description. In the final part of the thesis this methodology is applied to a simple well known solar system to show the characteristics, the operation and the potential application on the other components of the Solar Combi+ system.