The building energy efficiency has received massive attention from the government, industry and academia, due to the mismatch between the shortage of energy resources and the growing energy demands. The building is a complex system with a variety of components. One or several components can comprise a subsystem that provides additional or enhanced functionality to the building. These subsystems reveal enormous energy efficiency opportunities in buildings, including the power quality control, smart appliance operation, energy flow balance and energy efficiency project planning. Accordingly, a hierarchical building energy efficiency framework can be identified by categorising these energy efficiency opportunities into four layers: the power electronics layer, smart appliance layer, energy flow layer and planning layer. The four layers are distinguished by different functionalities and control intervals. While the first three layers involve excessively studied engineering fields, the energy efficiency planning is nevertheless less well understood, due to the lack of a systematic approach to model, evaluate and optimize the planning of building energy efficiency projects. As a result, the energy efficiency project planning has received increasing attentions from the researchers in the recent years. The retrofitting of existing buildings is one of the most important types of building energy efficiency projects, as the existing buildings account for a large portion of final energy consumptions in the world. The retrofitting planning aims at maximizing the energy and economy performances with limited budget and manpower. Therefore, the retrofitting planning is a kind of investment decision to make best use of the investment. However, such an investment decision is difficult due to the interactions of the multiple layers in building energy efficiency framework. Furthermore, the retrofitting investment decisions suffer significant risks from the failures of retrofitted items during operation. According to measurement and verification principles, failures of retrofitted items result in the decrease of the energy savings, which are the major concern of a retrofitting project. Although the deteriorated energy savings can be restored by applying maintenance actions, the economy performances receive further impacts from the maintenance costs. In summary, the investment decision of a retrofitting project can be very complex, manifesting multiple time scales and significant dynamics when simultaneously taking into account the retrofitting and maintenance planning. In order to address the investment decision complexity, a control system framework is proposed, where the dynamics of aggregated performances can be addressed and optimized. A necessary simplification is adopted where the retrofitted items are categorised into several groups. Each group consists of items that are considered to be homogeneous ones, i.e., with the same inherent energy and reliability performances, the same operating schedules and similar operational environment. Thereafter, the aggregate energy savings can be computed by the individual item savings and the item group populations. In this way, the control system modelling at management level can be obtained. The state variables are the item group populations, and the control inputs are the maintenance intensities, i.e., the count of the restored items from one group at a specific instant. Such instant is called maintenance instant, i.e., a time point at which the maintenance actions are scheduled to take place. The statistical laws of the item group population decay comprise the system dynamics. The measured outputs are the aggregate energy and economy performances. Thereafter, the retrofitting and maintenance planning are cast into an optimal control problem. A finite decision horizon, namely the sustainability period is defined, based on which the control objectives are obtained, i.e., maximising the aggregate energy savings and financial benefits. A series of constraints are accordingly introduced, e.g., the targeted energy saving limit, budget limit and payback period limit, etc. The influences of uncertainty factors are taken into account to be random noises on the state variables and measured outputs. Consequently, the control approaches can be introduced to address the retrofitting and maintenance planning. A model predictive control approach with a differential evolution algorithm based numerical solver is employed for the controller design in most of the illustrative studies. The control system framework allows development and expansion by selecting different state variables and control inputs. Given that the selective control inputs involve a broad field of maintenance engineering, a number of maintenance categories comprise the alternatives of control inputs. The introduction of different maintenance categories provides more options to decision makers. Thereafter, the complexity of performance dynamics can be addressed, and the utility of limited capitals and manpower can be improved. Following this idea, a series of extensive studies are conducted and illustrated after the elaboration of the control system framework. Firstly, a control system modelling with coupled state variables is proposed to address the interacting energy effects between different categories of retrofitted item groups. Secondly, the energy saving deterioration of retrofitted items before malfunctions is modelled by a multi-state system approach, which incorporates two different maintenance categories into the planning. Thirdly, the collaborative optimisation of the maintenance intensities and instants is proposed, where additional energy efficiency opportunities are identified. Finally, the robustness of control system performances when different grouping methods are applied is investigated. These extensive studies will be introduced in respective chapters in this dissertation.