Local Operation Overview

A typical residential house has one or more zones and an HVAC system operating each such zone. Each zone is assigned a schedule – one for heating and one for cooling, that allows a home owner to specify time intervals, range of temperature for each interval, comfort level desired for each interval, type of algorithm(learning/not learning), and collaboration algorithm, and the user can alter the schedules and assignment of these schedules for each zone. Each zone operates for heating or cooling in following modes:

  1. Probe: Allows the system to build a thermal profile of the zone. This profile enables the system to take basic decisions for runtime operation.
  2. Normal: Control the house temperature based on currently available house profile information and any previously learned information, perform additional learning to refine the system performance. The runtime operation is further divided into two phases:
    1. Transit: The system makes decision on how to move between two interval-specifications and performs necessary operations to meet requirements across those interval-specifications.
    2. Sustain: Once the system reaches the required temperature for an interval-specification, the system performs necessary operations to retain that temperature for the interval-specification.
Probe Mode

In probe mode, the system builds an initial house thermal profile by monitoring unit temperature changes for each stage. It enables the system to determine rate at which zone cools down or warms up and is useful to estimate expected time to reach a target temperature using a specific HVAC stage. Once the profiling is completed, the zone moves to other learning modes.

Probe Mode Algorithm
  1. Find maximum temperature \( T_{max} \) and minimum temperature \( T_{min} \) across all interval- specifications in all schedules.
  2. Calculate the range of temperature between the temperature extremes:
    • \( T_{diff} = T_{max} - T_{min} \)
    • If \( T_{diff} < 10 \), \( T_{max} = T_{max} + \left( 10 - \frac{T_{diff}} {2} \right) \) and \( T_{min} = T_{min} + \left( 10 - \frac{T_{diff}} {2}\right) \) . This ensures that the effective \( T_{diff} \) is atleast 10 degrees.
  3. Active Probe: For a cooling system mode, measure current temperature and run HVAC with most powerful stage cooling until \( T_{min} \) is reached. Record the time it takes to change each degree along with the operating condition at start of each such change including the inside and outside humidity, temperature, heating or cooling, stage of heating or cooling, and baffle.
  4. Passive Probe: For a cooling system mode, with HVAC system stopped, record time it takes to reach Tmax. After waiting for 8 hours, if system does not reach \( T_{max} \) abort the attempt and move to next stage. Such anomaly is encountered if user's schedules have unrealistic large values.
  5. Repeat the last 2 steps using lower stages (2 and 1).
  6. Repeat steps 3 – 5 with a baffle, if applicable.

For a heating system mode, the algorithm is the same except that \( T_{max} \) and \( T_{min} \) are swapped as in a heating system, the temperature rises when HVAC system is active.

Transit Mode

A transit operation for a zone enables transiting a zone from its current temperature to the target temperature of the zone. Learning to optimize HVAC stage and prepare time is crucial as a transition started too early or with higher stage results in wasted energy whereas a transition started too late or with lower stage fails to meet schedule. Moreover, changing the target-temperature by one degree saves around 3% energy and therefore, it is vital to also learn to optimize the target temperature for an interval.

Here are some important terms before we discuss the target temperature formulas:

  • Current interval-specification: Interval-specification that defines the operation of an ongoing (i.e. current) interval.
  • Upcoming interval-specification: Interval-specification that defines the operation of the next interval-specification to which the system must transit.
  • Current operating conditions: The actual conditions as qualified by various factors that affect the HVAC system such as current inside temperature TC and humidity, outside temperature and humidity.
  • Prepare-time: Time before the start of upcoming interval-specification when the system must start an operation to meet the upcoming interval-specification's strict requirement.
  • Target Temperature (TU): Temperature that the thermostat attempts to sustain for a given interval specification.
  • Start Time (MU): Start time is the time when the target temperature must be met for an interval specification.

Consider \( TU \) = Target temperature, \( T_{max} \) = Maximum interval specification temperature, \( T_{min} \) = Minimum interval specification temperature, \( CL \) = Percentage comfort level user setting - 100% for maximum comfort and 0% for minimum comfort (maximum savings).

Interval Target Temperature

For a cooling system, the initial target temperature that the system seeks is given by:

\[ TU = T_{min} + 1 + \left[(T_{max} - T_{min} - 2) \times \frac{100 - CL}{100}\right] \]
resulting in \( TU = T_{min} + 1 \) at 100% comfort-level and \( TU = T_{min} - 1 \) at 0% comfort-level

For a heating system, the initial target temperature the system seeks is given by:

\[ TU = T_{max} - 1 - \left[(T_{max} - T_{min} - 2) \times \frac{100 - CL}{100}\right] \]
resulting in \( TU = T_{max} - 1 \) at 100% comfort-level and \( TU = T_{min} + 1 \) at 0% comfort-level
Transit Mode Algorithm
  1. Determine the target temperature \( TU \) for the upcoming interval and current operating conditions. Also determine a proposed optimal prepare-time to reach the target temperature for the current conditions. This step requires machine learning as there are numerous factors that control an optimal target temperature and the required prepare-time for the current conditions.
  2. Compute the change in temperature \( \delta T = | TU - TC | \) required to reach the target temperature and based on the allowed prepare-time to reach this temperature, start the appropriate stage of HVAC system at the appropriate time.
  3. Monitor the change in temperature and actual time it took to transit to the target temperature in the current conditions and report it as feedback to fine-tune learning.

The system has to learn to seek an optimal target temperature starting at an optimal time before the upcoming interval that best conserves energy while operating the system for these temperature and comfort level settings.

Sustain Mode

Sustain operation is activated when the zone successfully transits to its target temperature. Here the challenge is how to maintain target temperature. The thermostat is determining how to replenish the leaked energy from the zone to the surroundings. In sustain mode, the system minimizes energy and maximizes comfort by determines an optimum sustain range of temperature. Narrow range sustain most comfortable set temperature but causes frequent start and stops of HVAC system, thereby wasting energy. So system learns to obtain widest possible sustain range and to compensate for any loss of comfort, it bumps target temperature such that optimal comfort can be achieved without incurring significant energy cost.

Sustain Mode Algorithm

Sustain operation requires a sustain temperature range threshold \(D_{max}\), \(D_{min}\) and HVAC stage that must be used to sustain the temperature between this sustain temperature range. Unlike conventional thermostats, the idea here is not to maintain a fixed temperature but to maintain temperature between a range of temperatures that are considered acceptable for the user-defined comfort level.

Determining the sustain temperature range however is quite tricky, forms the basis of energy conservation versus comfort tradeoff and requires machine learning to be effective. An exemplary analysis is performed to learn these effective sustain operation parameters for various conditions, some of which are described below.

  1. In absence of other previously learned information, compute the target target temperature \( TU \) based on current interval's temperature and comfort level.
  2. Starting with an initial sustain temperature range threshold \([D_{max} - D_{min}]\) progressively expand this range until it remains within the bounds of the interval's allowed temperature range \([T_{max} - T_{min}]\).
  3. Bump up the comfort by offseting the sustain temperature range \([D_{max} - D_{min}]\) towards \(T_{max}\) for heating and towards \(T_{min}\) for cooling modes.