Building automation: heating and cooling at once, missing sensors

Heating and cooling run at the same time

What happens. In one building the heating circuit warms the air that the cooling circuit has just cooled, or two zones fight each other. Each control loop does its job, and nobody looks at the whole. It shows up on the energy bill and in the complaints about cold and hot rooms.

What it costs. Energy for heat and cold that cancel out, wear on valves and machines that switch too often, and comfort that is still not good.

Why the usual fix fails. Adding dead bands between heating and cooling mode makes rooms drift. Setting every zone to a conservative setpoint wastes energy. A fix per room fails when the next season changes the load.

How a fixed-price remote diagnosis checks it. From trend logs of valve positions, flow temperatures, room temperatures and outdoor conditions I look at where simultaneous heating and cooling happens and why. The fixed-price remote diagnosis tells you whether a coordinated predictive control of heating and cooling is realistic with the building’s existing points.

Background: Economic predictive control

Room temperature sensors are misplaced or missing

What happens. The room sensor sits in a draught, behind a cupboard or next to the sun, or there is none at all. The control follows that value, and the rest of the room is warmer or colder than the display says.

What it costs. Comfort complaints, overheating that nobody sees, and energy spent to correct for one wrong measurement.

Why the usual fix fails. Moving or adding sensors costs wiring and time per room. Raising the setpoint to keep people quiet raises the bill. A complaint-driven fix by the caretaker changes things in one room and not in the next.

How a fixed-price remote diagnosis checks it. From existing sensor trends, valve and flow data and the floor plan I look at how well the room state can be computed from what is there. The fixed-price remote diagnosis tells you in which rooms a computed room temperature is good enough and where a sensor is really needed.

Background: Computed measurements

Old buildings have no heat meters, so there are no data

What happens. In older buildings the heat flows to the zones are not metered. Energy managers have a total from the boiler or the district heating connection, and nothing per circuit. Without that, a fault or a waste cannot be placed.

What it costs. Energy problems that stay invisible, retrofit measures that cannot be checked afterwards, and a meter investment that is hard to justify for each circuit.

Why the usual fix fails. Installing heat meters in every circuit is expensive and needs a shutdown. Estimating from the yearly total gives no profile. Clamp-on meters give a few days and no more.

How a fixed-price remote diagnosis checks it. From temperature and valve position data, which exist in most automation systems, I look at how well the heat flow per circuit can be computed, compared with a meter on a sample circuit. The fixed-price remote diagnosis tells you the accuracy to expect.

Background: Computed measurements

Hydraulic balancing is done once and then forgotten

What happens. When a building is balanced, each circuit gets the flow it needs. Then rooms are renovated, valves are turned, pumps are exchanged and use changes. The balance is lost without notice, and some circuits are starved while others get too much.

What it costs. Cold rooms at the end of the pipe, pumps that run harder than needed, return temperatures that are too high and reduce the efficiency of boilers and heat pumps.

Why the usual fix fails. A new manual balancing costs a visit per valve and holds until the next change. Raising the pump pressure fixes the cold rooms and costs power. Complaints are the only signal.

How a fixed-price remote diagnosis checks it. From temperature differences, valve positions and pump data I look at which circuits are starved or oversupplied. The fixed-price remote diagnosis tells you whether a continuous estimate of the balance can guide the setting of valves and pumps.

Background: Computing unmeasured states for control

Predictive control of many buildings needs a model per building, few sensors

What happens. Predictive control of heating and cooling needs a model of each building. A building portfolio has hundreds of them, each with different walls, systems and few sensors. Building the models by hand takes weeks per building.

What it costs. An approach that works in a pilot and stops being profitable at the tenth building, and engineers who spend their time on models, not on results.

Why the usual fix fails. Standard models from building types are too far from the real building. Detailed models need drawings and measurements that are missing. More sensors in every building would pay for the work itself.

How a fixed-price remote diagnosis checks it. From a few weeks of trend data of a sample building I look at how well a simple model can be identified automatically and how much control benefit it supports. The fixed-price remote diagnosis tells you whether automated identification and control can scale across buildings.

Background: Industrial process predictive control