"Introduction Advanced solar houses This article presents predictive control strategies appropriate for advanced solar homes. Design techniques and technologies available today@ such as passive solar design@ building-integrated solar energy technologies@ and energy storage devices@ make it possible to build homes whose energy needs are satisfied mainly or completely with solar radiation received on the exterior surfaces of the house. A solar house is defined in this article as one with the following key features for the case of a cold sunny climate: ? First and foremost@ integrated design-the conception of the house sub-systems (lighting@ HVAC@ appliances) forms a coherent plan in which there is complementarity of functions. This includes the building integration of the renewable energy systems@ which play a role as components of the building envelope (walls or roofs). ? Passive solar techniques-this key component of the overall design strategy includes highperformance windows oriented toward the equator@ increased thermal mass for storage of solar heat gains@ higher insulation levels@ and improved air-tightness. ? Active solar energy technologies-the house design generally includes a building-integrated photovoltaic (BIPV) or BIPV/thermal (BIPV/T) system@ a solar thermal collector@ or similar systems. ? Typically@ a grid-tied installation-whenever the photovoltaic (PV) generation exceeds the internal loads of the house@ the surplus is sent to the grid; if the consumption of the house exceeds the generated power (e.g.@ at nights or under cloudy conditions)@ energy is purchased from the utility grid. ? Motorized shading devices on windows@ electrochromic windows or similar technologies that enable partial control of the solar heat gains entering the space. ? Energy storage devices@ such as thermal energy storage (TES) or batteries. ? A centralized supervisory control system that enables the implementation of advanced control strategies for energy management. Appropriate control strategies are essential for the successful operation of high-performance buildings (Torcellini et al. 2004). For over two decades@ optimal control has been investigated as a tool for the management of passive and active TES capacity in buildings (Braun 1990@ 2003; Morris et al. 1994; Henze et al. 1997@ 2004). Optimal control algorithms use estimates of future loads to select a sequence of control operations to optimize an objective function (typically@ energy@ peak load@ or cost). These investigations have addressed mostly optimal control of cooling capacity storage (ice or chilled water) in commercial buildings (Henze 1995). Predictive control has also been applied to the case of solar buildings (Kummert et al. 1996@ 2001; Chen 2001). Several building simulation software tools@ such as ESP-r (ESRU 2010) and EnergyPlus (EERE 2010)@ achieve accurate representations of buildings through careful integration of detailed models of physical phenomena into a single@ comprehensive tool. Such a model provides a reliable representation of the building's response to external impulses and its HVAC system@ which is particularly useful for research purposes. Although a fullscale building simulation tool can be used for the testing and design of advanced control strategies@ this application can be quite cumbersome. The need for complex building and HVAC models has been identified as a hurdle for the deployment of control strategies (Wang and Ma 2008). For example@ anticipatory control strategies used to select setpoint trajectories (Coffey et al. 2006) require estimating the effect of an action (such as turning on a piece of equipment or changing the position of a valve) based on expected loads. Predictive control calculations imply performing building energy simulations at regular intervals with a moving time horizon. It has been found that simplified building models can successfully be used for control applications (Athienitis et al. 1990; Kummert et al. 1996; Fraisse et al. 2002; K??ampf and Robinson 2007). These simplified models have commonly been based on thermal network representations with a limited number of thermal resistances and capacitances (Fraisse et al. 2002). In a recent large-scale project on predictive control@ a simplified model of a room@ later validated with a more detailed model@ was used in the development of control strategies (Gyalistras 2010). A simpler model has several advantages (such as ease of implementation@ insight into physical phenomena@ computational efficiency); however@ deciding a priori the right complexity level for a given application is a difficult task. The selection of the appropriate level of model complexity for an advanced control strategy is always a crucial decision (Kummert et al. 2006; Gyalistras 2010). The difficulty lies on deciding which details can be neglected without jeopardizing the validity of the conclusions. Simplified physical models@ or gray-box models@ derived from either more detailed building simulation models or from actual measurements@ have been proposed as a way to facilitate the implementation of control strategies (Wang and Ma 2008). This approach has the advantage of providing a link between design and control. Control strategies can (and should) be created during the design of the building by applying the same models with adjusted levels of resolution. This article investigates the application of predictive control strategies in a solar house. The control strategies are applied at two different control levels: (a) for the control of a high-mass radiant floor heating (RFH) system in a room exposed to solar gains and (b) for controlling the charging of a TES tank with a solar-source heat pump. Anticipatory strategies for distributed and isolated TES Appropriate passive solar design includes fenestration with the right dimensions and orientation to collect solar gains@ as well as adequate levels of thermal mass to receive and store these heat gains while avoiding overheating. By gradually releasing the collected heat into the space@ it is possible to maintain adequate comfort levels over a period of about 24 h with the solar gains obtained during one sunny day (Candanedo and Athienitis 2010). It is possible to extend this period of space heating autonomy by using renewable energy technologies to capture thermal energy@ which can then be stored in isolated TES devices. Therefore@ controlling the heating energy used in a solar house consists of coordinated strategies for the control of the ""distributed"" energy storage of the building's thermal mass and the ""isolated"" TES devices charged with solar energy. This article addresses both kinds of thermal storage. A model predictive control (MPC) algorithm is used in this investigation for the control of the heat delivery rate of an RFH system and@ consequently@ for the management of distributed TES in the floor's thermal mass. MPC is the collective name given to a group of algorithms originally developed for control of chemical engineering processes@ often characterized by long time constants. When significant thermal mass is used in a building@ as in the case of a passive solar building or in so-called thermally activated building systems (TABS) (Gwerder et al. 2008)@ time constants of the order of many hours or days are also present. Traditional control strategies (e.g.@ ON/OFF or proportional-integralderivative control (PID) control)@ which trigger a control operation when the response deviates from the reference@ are inadequate for systems with long time constants@ since the effects of control actions are only perceived after a relatively long period. By using a model of the system and estimates of future loads@ MPC algorithms can take anticipatory action rather than correctivemeasures and@ thus@ better control the system's output and track the desired set-point. Therefore@ MPC is suitable for the control of slow-responding systems (Oldewurtel et al. 2010). This investigation also includes a dynamic programming algorithm used for the control of an isolated TES device (a hot water tank). BIPV systems@ BIPV/T systems@ solar thermal collector(s)@ and solar-source heat pumps installed in a solar house can produce a significant amount of energy during sunny days while having poor production during cloudy periods. It is common to observe amismatch between the power production and consumption profiles in a solar home (Voss et al. 2010). Interaction with the utility grid (i.e.@ ""grid-tied"" scheme) and energy storage devices contribute in dealing with the energy mismatch problem. A naive control approachwould be trying to keep TES devices charged at all times. However@ solar thermal systems perform more efficiently when TES devices are maintained at lower temperatures (Duffie and Beckman 2006). For example@ if a TES tank is used as the energy sink of a solar-assisted heat pump@ maintaining the tank at a low temperature can significantly improve the coefficient of performance (COP) of the heat pump. In places where electricity is extensively used for residential space heating-as in many Canadian provinces (Snider 2006)-heating loads can have amajor impact on peak electric loads. Solar-heated TES devices@ together with appropriate control strategies@ can be employed to reduce electrical peak loads for buildings heated with electrically operated heat pumps."