Two chillers, a secondary loop and a fouled tower

Workbook exercise plant-chiller-multi · PNNL re-tuning chapter 8 · about 50 minutes

Goal

Judge a plant with more than one of everything. Find two chillers running when one would do, a secondary loop that runs warm while both chillers hold their setpoint, a chiller with a fouled condenser, and two fouled cooling towers: one pinned at full fan, one whose controller hides the fouling in its fan. Then look at a plant a little off everywhere that passes every check.

Learn more

Read these first (PNNL, free):

• Central Utility Plant Cooling Control, the re-tuning guide: chiller staging, the primary-secondary loop and the cooling tower are all in its questions.
• Chapter 8: Central Utility Plant of the re-tuning training.

Datasets and licence

• synthetic-chw-plant-multi: a synthetic primary-secondary chilled-water plant that CAMBER generates on your computer: two equal chillers on constant primary flow, a decoupler, a variable-flow secondary loop and one cooling tower. Nothing is downloaded and no real building is behind it. It holds the outdoor dry-bulb and wet-bulb; each chiller's run status, power, leaving and entering water, and the condenser and evaporator approaches its controller reports; the secondary loop's supply, return, setpoint, flow, pump speed and pump power; and the tower's leaving and return water and fan speed. The data are every 15 minutes for three summer weeks, with a heat wave in the third. Licence Apache-2.0 (CAMBER's own; the data may be shared). About 1.7 MB. See synthetic datasets for how CAMBER generates it.

The same weather and the same building drive seven plants: well_run, over_staged, reverse_mixing, fouled_condenser, tower_full_fan, tower_controlled and near_miss. Each plant is four pieces of equipment in the facility ds-synthetic-chw-plant-multi: CH1__<plant> and CH2__<plant> (the chillers), SECLOOP__<plant> (the secondary loop) and TOWER__<plant>. The plant names say where to look; what is wrong, and which rule sees it, is the exercise.

Setup

In the lab

1. Run camber lab and open the URL it prints (http://127.0.0.1:8765/lab?token=...).
2. Tick synthetic-chw-plant-multi (its kind reads synthetic) and press Fetch & ingest. CAMBER generates the file instead of downloading it, then ingests it like any other dataset.
3. When the job is done, the row links to trends (the trend viewer) and to report, the audit report of the dataset's template at the trend's own 15-minute resolution. The fleet steps below use the command line.

On the command line

camber datasets fetch synthetic-chw-plant-multi
camber datasets ingest synthetic-chw-plant-multi --store lab_store
camber datasets config synthetic-chw-plant-multi --exercise plant-chiller-multi --store lab_store --out multi.json
camber run multi.json --out multi_out
camber datasets score synthetic-chw-plant-multi --store lab_store --findings multi_out/findings.json

The exercise's config (--exercise plant-chiller-multi) runs four rules at their defaults on the 15-minute trend: chiller_staging_fleet (over the whole facility, on purpose), chw_supply_tracking, chiller_approach_fouling and cooling_tower_approach. Its drift section adds the tower drift family with a declared reference: cooling_tower_fan_effort_drift scores each tower's fan effort against TOWER__well_run, fitted in memory on each run and never stored.

Three variants run chiller_staging_fleet on one plant's two chillers only: plant-chiller-multi--well-run, plant-chiller-multi--over-staged and plant-chiller-multi--near-miss. For the over-staged plant:

camber datasets config synthetic-chw-plant-multi --exercise plant-chiller-multi--over-staged --store lab_store --out fleet_over.json
camber run fleet_over.json --out fleet_over_out

Steps

1. Look before you judge. In the trend viewer, open CH1__well_run and CH2__well_run and plot both chillers' power over a week. When does the second chiller start? Do the same for the over-staged plant. Then plot SECLOOP__reverse_mixing's supply temperature with its setpoint, and CH1__reverse_mixing's leaving water.
2. Run the exercise's config (the commands above) and read the chiller_staging_fleet line camber run prints. How many chillers does it count, and where does its finding sit?
3. One plant at a time. Run the three fleet variants and read pct_overstaged, n_multi_samples (and multi_hours), median_running_count, rep_capacity_kw and capacity_source in each findings.json.
4. The secondary loop. Read chw_supply_tracking for every SECLOOP__<plant> and for the reverse-mixing plant's two chillers: above_pct, median_excess_above_f, chwst_median_f, deltaT_median_f and run_source.
5. The condensers. Read chiller_approach_fouling for each chiller: cond_approach_f, evap_approach_f and worst_approach_ratio.
6. The towers. Read cooling_tower_approach for TOWER__well_run, TOWER__tower_full_fan and TOWER__tower_controlled: approach_median_f, pct_hours_high_approach, n_operating and n_low_effort_excluded. Then read cooling_tower_fan_effort_drift for every tower: fan_effort_drift_pct, the summary and the caveats. Read cooling_tower_approach_drift too: tower_approach_drift_f and tower_approach_load_source.
7. Score. Run camber datasets score and read the overall rates and each rule's.

Questions

1. Over the whole facility, how many chillers does chiller_staging_fleet count, what does it leave out, and what does it conclude? Why is pooling seven plants on one clock still wrong, and what fixes it?
2. On one plant at a time, which plant over-stages? What capacity does the rule assume for a chiller, where does that figure come from, why does it read low on that plant, and what would you declare instead?
3. Which plant's secondary loop misses its setpoint? Are its chillers at fault? What is the loop delta-T, and what is the water in the decoupler doing?
4. Which chiller's condenser is fouled, and how far is its approach from design?
5. Which tower does cooling_tower_approach flag, and which does it pass? Both are fouled: what does the controlled tower do instead of widening its approach, and which rule sees that?
6. cooling_tower_fan_effort_drift passes the over-staged plant's tower, which a range-only match would have flagged. Why would two chillers' condenser water make a clean tower look harder-working, and which point prevents that?
7. What does the label score say for each rule, and for the plant as a whole? Which fault can no label score count, and why?
8. The near-miss plant passes every check. What is a little off in it, and what would you watch?

What CAMBER shows

• Findings. camber run prints one line per finding: one per rule and equipment, plus one <fleet> line for chiller_staging_fleet.
• Fleet. chiller_staging_fleet takes every chiller that carries a power point and puts them on one clock. A chilled-water loop (CHW_LOOP, SECLOOP) is left out (n_loops_excluded): its power is its pumps'. A sample is over-staged when two or more run and their total power would fit in one machine fewer at 90 % (redundancy_ceiling) of a chiller's capacity. That capacity (rep_capacity_kw) is the declared full-load power (capacity_kw) when given, else the median, across the chillers, of each one's own 95th percentile running power (capacity_source). It warns when 20 % of the multi-chiller samples are over-staged and faults at 50 %. The counts are samples (n_multi_samples), with their hours beside them (multi_hours). Its finding sits on <fleet>, not on a chiller.
• Supply tracking. chw_supply_tracking reports the share of running time with the supply more than 3 °F above its setpoint (above_pct): a warn at 10 %, a fault at 25 %. "Running" is the chiller's status when mapped, else its power (run_source). The secondary loop is a loop, so it is gated on its pumps: their status when mapped, else the loop flow (here, run_source is flow), else the pumps' power.
• Approach. chiller_approach_fouling compares each chiller's median condenser and evaporator approach with 5 and 4 °F (worst_approach_ratio): a warn at 1.5 times design, a fault at 2.
• Tower. cooling_tower_approach judges the approach (leaving water minus wet-bulb) only in samples with the fan at 90 % or more (n_operating; the rest are n_low_effort_excluded), against a 7 °F design: a warn when the median is 1.3 times design, a fault at 1.7. cooling_tower_fan_effort_drift reads the fan speed against the reference tower at matched condenser load and wet-bulb: a warn at 5 percentage points, a fault at 10. With a condenser flow (cw_flow) the load is the heat rejected, else the range, which assumes a constant condenser flow. These towers trend it (CW_GPM), so the load is the heat rejected (load_source cw_heat_rejection). cooling_tower_approach_drift reads the approach against the reference at matched load the same way (tower_approach_load_source).
• Score. camber datasets score prints the true- and false-positive rates, with 95 % intervals, for the entry's declared detectors and for the plant as a whole.

From this exercise's synthetic plants: each tower's approach (condenser water leaving minus wet-bulb) against its fan speed, every running sample. cooling_tower_approach judges only the samples right of the dashed 90 % line, and counts those above the dotted line (design 7 °F plus 3 °F) as high.

Caveats

• The data are synthetic: CAMBER's own model of one plant, built to show these faults. They show what the rules see, not how real plants behave; never cite them as evidence about a building.
• The approaches are the chiller controller's own refrigerant-side readings, generated directly. A real chiller reports them from its refrigerant pressures and temperatures, and not every controller trends them.
• In the heat wave, the badly fouled tower sends condenser water near 99 °F. A real chiller would trip on high head pressure before that.
• A 15-minute trend averages a chiller's start into its first sample. The fleet rule's counts are of 15-minute samples (n_multi_samples); multi_hours gives their hours.

Going further

• Compare with the cooling tower exercise: its simulated tower is controlled to the wet-bulb plus 8 °F, so its fouling shows only as fan effort, and cooling_tower_approach scores no fouled run there.
• Edit fleet_over.json and set chiller_staging_fleet's redundancy_ceiling lower and higher. How does the over-staged share move, and what would you set it from on a real plant? Then declare capacity_kw (about 138 kW, a chiller's draw at full load here) and read the verdict again.
• Run camber rules params chw_supply_tracking and camber rules params cooling_tower_approach to see where each default comes from.