Building management and electrical power monitoring systems

Two monitoring systems do most of the watching at Stonebridge: the building management system for everything mechanical and environmental, and the electrical power monitoring system for everything electrical. They look similar on screen, with lists of points, graphics, alarms and trends, but they collect different data, at different speeds, for different reasons. This lesson explains how each one gets its data, what it stores, and what a technician can and cannot learn from it.

Points: the basic unit of monitoring

Every reading or status a monitoring system collects is called a point. A point has a name, a value, a unit, a time stamp and a quality flag that says whether the value is fresh and trusted. 'CRAH-A07 supply air temperature' is a point; so are 'CRAH-A07 fan status', 'UPS-B2 output kW' and 'Leak cable D3-04 alarm'. A large site has tens of thousands of points: YUL-2's BMS holds about 38,000 and its EPMS about 21,000.

Points come in a few kinds, and the kind matters when reading a screen:

• Analog inputs: measured values that vary, such as temperature, pressure, flow, current or kW.
• Binary or digital inputs: two-state statuses, such as on/off, open/closed, alarm/normal.
• Outputs: values the system sends to equipment, such as a CRAH's supply setpoint or a valve's commanded position. A point that the BMS commands is often different from the point that reports what actually happened: 'valve command 100 %' and 'valve position feedback 12 %' disagreeing is exactly how the stuck actuator in lesson 1 shows up.
• Calculated points: values the software works out from others, such as a hall's total IT load or a CRAH's delta-T.

Each point's name follows a naming convention, so that anyone can tell from the name what and where it is. Stonebridge uses site, building, system, equipment and point, for example YUL2.A.CRAH.A07.SAT for the supply air temperature of CRAH-A07 in Hall A (YUL2 on the other campus). Naming conventions matter more than they seem: a point named 'TEMP-3' in a list of 38,000 is no help at three in the morning.

How the BMS gets its data

The BMS is a network of controllers. Each piece of mechanical equipment has its own controller: a CRAH's onboard controller, a CDU's controller, a chiller's panel. Field controllers in each room collect the loose sensors: temperature and humidity sensors on rack doors, differential pressure sensors under the floor, leak detection cables. The controllers talk to BMS servers in the operations room over a dedicated network, using one of a few standard protocols:

• BACnet, the building automation protocol defined by ASHRAE and ISO, used by most building controllers. BACnet devices describe their own points as objects (analog input, binary value and so on), which makes them easy to add to a BMS.
• Modbus, an older and simpler protocol widely used by industrial and electrical equipment. A Modbus device holds its values in numbered registers; the BMS must be told which register means what, from the device's register map.
• SNMP, the network management protocol, used by IT-style equipment such as rack PDUs, some UPS network cards and environmental monitors.

Many devices speak only one of these, so sites use gateways that translate, for example from a CRAH's Modbus registers to BACnet objects. Each extra translation is another place for a mistake: a register mapped to the wrong point, a value scaled wrongly (a temperature in tenths of a degree read as whole degrees), or a gateway that freezes and keeps reporting its last value.

Polling, trending and stale data

Most monitoring systems poll: they ask each device for its values at a fixed interval, or the device reports a change. A typical BMS poll interval is a few seconds to a minute. Values are stored as trends, with a time stamp, so that any point can be graphed over hours or months. Stonebridge keeps most trends at one-minute resolution for 13 months and hourly averages for seven years, long enough to compare one summer with the next.

Polling has two consequences a technician must understand. First, the screen is never quite live: a value may be up to one poll interval old, and events shorter than the interval can be missed entirely. Second, if a device stops answering, the system may keep showing its last value. Good systems flag this with a quality marker, a grey value or a 'stale' or 'comm fail' alarm, but a technician should always check the time stamp on a value that seems too steady. A temperature that has not changed by a single tenth of a degree for six hours is more likely a frozen gateway than a perfectly stable room.

# Spot stale points: a value that has not changed for too long is suspect
from datetime import datetime, timedelta
now = datetime(2027, 3, 9, 3, 0)
points = [  # (point, last value, time of last change)
    ("YUL2.A.CRAH.A07.SAT", 22.1, now - timedelta(minutes=3)),
    ("YUL2.A.RACK.A17-04.INLET", 23.4, now - timedelta(minutes=12)),
    ("YUL2.D.CDU.D2B.SUPPLY", 30.0, now - timedelta(hours=6, minutes=40)),
    ("YUL2.B.PDU.B05.KW", 212.6, now - timedelta(seconds=40)),
]
limit = timedelta(hours=1)
for name, value, changed in points:
    age = now - changed
    flag = "STALE? check comms" if age > limit else "ok"
    print(f"{name:28s} {value:7.1f}  unchanged for {str(age):>8s}  {flag}")

YUL2.A.CRAH.A07.SAT             22.1  unchanged for  0:03:00  ok
YUL2.A.RACK.A17-04.INLET        23.4  unchanged for  0:12:00  ok
YUL2.D.CDU.D2B.SUPPLY           30.0  unchanged for  6:40:00  STALE? check comms
YUL2.B.PDU.B05.KW              212.6  unchanged for  0:00:40  ok

The CDU point is the one to check. In this case it was real: the gateway serving CDU-D2B's controller had locked up after a firmware update, and the BMS was still showing the last value it received. The CDU itself was fine, and its own panel in the gallery showed correct, varying readings. The fix was a gateway restart by the controls contractor, ticketed as a change.

What the BMS does beyond watching

The BMS also controls. It can command CRAH setpoints, start and stop equipment, sequence chillers and dry coolers, and run automatic routines such as starting a standby CRAH when another fails. Much of the cooling plant's automatic behaviour described in DCT-230 lives in the BMS or in the equipment controllers it supervises. Because the BMS can control, access to it is split: the desk staff can view everything and acknowledge alarms; only named controls engineers and critical facilities staff can change setpoints, schedules or programs, and every change is logged with the user name.

A technician should know where control lives for each system. If the BMS server fails, the equipment controllers keep running on their last settings, which is why sites put the critical logic in the controllers and not only in the central software. The desk loses visibility, not cooling; but losing visibility is itself an abnormal state, and the desk falls back to more frequent walking rounds until it is restored.

The electrical power monitoring system

The EPMS is built from power meters and the electronic trip units of breakers, UPS and generator controllers, and transfer switch controllers. Meters are installed at every important point in the power chain: the utility incoming lines, each switchboard main and feeder, each UPS input and output, each floor PDU and RPP, and often each branch circuit. They measure voltage, current, real and apparent power, power factor, energy, frequency and harmonics.

The EPMS differs from the BMS in two important ways. It is faster and more detailed: good power meters capture events lasting a few milliseconds, and record waveforms of the voltage and current before, during and after a disturbance. And it records the state of every breaker, which is how the one-line overview on the desk wall shows which paths are carrying load.

The EPMS is what made the DCT-220 storm night reconstructable afterwards. The utility line loss at 02:41:07 was recorded by the incoming meters to the millisecond, along with the voltage sag that preceded it, the UPS transfer to battery, the generator breakers closing and the frequency swings on generator. The incident report's timeline came straight from the EPMS event log.

Power quality records

Power meters with power quality functions record events against limits: sags (voltage dips), swells, interruptions, transients and harmonic distortion. Each event is stored with its start time, duration, depth and a waveform. A technician on the desk reads these to answer simple questions: did a sag reach the equipment, how deep and how long was it, and did the UPS go to battery? DCT-220 lesson 2 covered how sags are judged against an equipment tolerance curve.

The EPMS also matters for billing and capacity. The energy meters on each customer's branch circuits are the source of the power figures in monthly reports and the readings that capacity planning uses (lesson 6). A meter that has drifted or been wired with a current transformer the wrong way round gives wrong figures for months unless someone notices. A negative kW reading on a load that cannot export power is the classic sign of a reversed current transformer.

import math
# Checking EPMS readings for a three-phase feeder: do V, I and kW agree?
v_ll = 600          # line-to-line volts
amps = [412, 405, 418]
pf = 0.97
kw_reported = 410.5
kw_calc = math.sqrt(3) * v_ll * (sum(amps) / 3) * pf / 1000
print(f"Calculated from volts, amps and PF: {kw_calc:6.1f} kW")
print(f"Reported by the meter            : {kw_reported:6.1f} kW")
diff = (kw_reported - kw_calc) / kw_calc * 100
print(f"Difference: {diff:+.1f} %  ->", "consistent" if abs(diff) < 5 else "check the meter or its current transformers")

Calculated from volts, amps and PF:  415.0 kW
Reported by the meter            :  410.5 kW
Difference: -1.1 %  -> consistent

Integration with DCIM and tickets

Neither the BMS nor the EPMS knows which customer owns a rack, or what a point means in business terms. That is the job of DCIM, which reads selected points from both systems and attaches them to assets: the rack's inlet temperature, its branch circuit loads, the CRAHs that cool its row. When Corvid's team asks what the inlet temperatures were in row 3 overnight, the desk answers from DCIM, which takes its data from the BMS. Some alarms also create tickets automatically; others are turned into tickets by the desk. Lessons 5 and 7 return to both links.