Pressure and night flow watch for a water distribution network

At a glance

Item	Typical answer
Problem	Leakage and bursts in district metered areas are found weeks late; pressure-reducing valves fail without anyone noticing.
What is connected	DMA inlet flow meters, PRV inlet and outlet pressure, critical-point and average-zone-point pressure loggers.
Who uses it	Leakage and network teams, the control room, PRV maintenance crews, the network analyst.
Typical first scope	Three to five DMAs with their inlet meters, PRVs and two or three pressure loggers each.
Time to first screen	Typically within the first week for powered inlet chambers; LoRaWAN loggers follow once radio coverage is confirmed, often two to four weeks.

The situation

Most distribution networks are divided into district metered areas (DMAs) with one or a few metered inlets, many of them behind pressure-reducing valves. On paper, the leakage team watches inflow and the minimum night flow of every DMA. In practice, the inlet meters are often read by data loggers that are downloaded by hand or upload once a day, the pressure points are logged during campaigns only, and PRV problems show up as customer complaints.

The result is slow detection. A burst that does not surface can run for weeks before the monthly night-flow review catches the step change. A PRV whose pilot sticks open pushes the whole DMA to inlet pressure, raising the burst rate across the area. A failing pilot on the other side leaves the highest streets with low pressure at peak hours. Each of these is visible in the data within hours, if the data arrives within hours.

This blueprint sets up continuous inflow and pressure monitoring on Hermetry, computes minimum night flow per DMA every night, and alarms on the patterns that matter: night flow steps, low pressure at the critical point, and a PRV that is no longer reducing.

Goals and measures

• Minimum night flow (MNF) per DMA = the lowest 15-minute average of net inflow between 02:00 and 04:00 local time. Net inflow = sum of inlet flows − sum of outlet flows to neighbouring DMAs.
• Night flow as a share of daily flow = MNF ÷ average inflow of the same day. A rising share points to leakage growing faster than demand.
• Estimated leakage at night = MNF − legitimate night use, where legitimate night use is an allowance per property and per non-household customer that the leakage team sets as an attribute of the DMA.
• Step change = today's MNF − the median MNF of the previous seven nights. A step above an agreed volume (for example 2 m³/h in a DMA of 2,000 properties) starts a leak search.
• Average zone night pressure (AZNP) = the average of the average-zone-point logger between 02:00 and 04:00. It tracks the effect of pressure management.
• PRV performance = share of time the outlet pressure stays within ±0.3 bar of its setpoint, and the inlet-to-outlet differential.

Hermetry keeps one-minute, hourly and daily aggregates of each key (average, minimum, maximum, sum, count and last) beside the raw readings, so charts can show hourly averages or daily minimums without extra tables. The 15-minute average that MNF needs is best computed in a flow, described under device setup.

Architecture

Two kinds of site make up the field side. Powered chambers and PRV kiosks get a licensed gateway running Hermetry Edge; unpowered points get battery LoRaWAN loggers.

Component	Where	Connects how	Notes
Licensed 4G gateway with Edge	Powered inlet chambers, PRV kiosks	Reads the flow meter over Modbus RTU (RS485) or its pulse output on a digital input; pressure transmitters on analog inputs; MQTT with TLS to Hermetry	One outbound connection, factory identity, claimed into your organisation.
Licensed NB-IoT or Cat-M1 gateway	Chambers with weak 4G coverage	Low-power cellular	Check model capabilities on the licensed devices list.
LoRaWAN pressure logger	Critical points, average zone points, hydrants	LoRaWAN, EU868, OTAA, class A	Battery powered; joins Hermetry's own network server.
LoRaWAN flow pulse logger	Unpowered meter chambers	LoRaWAN, counts the meter's pulse output	Sends a running total; flow is derived from the increase.
LoRaWAN gateway	Reservoirs, pumping stations, tall buildings	LoRa Basics Station to Hermetry; uplink over the site's network or a licensed router	Gateway health is stored as telemetry, so an offline gateway raises a normal alarm.
Hermetry	EU, US or Türkiye region, or on premises		Storage, network server, alarms, rules, flows, dashboards.

Each organisation gets its own LoRaWAN network server address; the Basics Station files Hermetry produces for a gateway carry it, so nobody types it. Semtech UDP packet forwarders are not supported, so choose gateways with Basics Station. Logger frames are decoded by a codec from the library, chosen automatically by model. If a logger's keys do not match your naming, save a copy of the library codec, adjust it, check it on the test bench with a real frame in hex and publish it: your copy then overrides the library codec for that model in your organisation.

Plan radio coverage before buying loggers. Underground chambers and hydrant pits attenuate heavily; a short survey with one gateway and two test loggers at the worst locations saves a second visit. The gateway drawer lists the devices it heard recently, and each device's LoRaWAN tab shows the latest reception and which gateways heard it.

Point list

Per DMA with one metered inlet behind a PRV, one critical point and one average zone point. Keys are examples.

Point	Telemetry key	Unit	Source	Interval	Notes
Inlet flow	inflow_m3h	m³/h	Electromagnetic meter, Modbus input register 30001	60 s	Signed, so reverse flow shows as negative.
Inlet totaliser	inflow_total_m3	m³	Meter totaliser, 30003	15 min	Used to check the integrated flow.
Inlet flow, 15-minute average	inflow_avg15_m3h	m³/h	Computed in a flow	15 min	Basis for MNF.
Outlet to neighbouring DMA	outflow_m3h	m³/h	Second meter, Modbus	60 s	Only where the DMA exports water.
PRV inlet pressure	prv_in_bar	bar	4-20 mA transmitter, analog input 1	60 s
PRV outlet pressure	prv_out_bar	bar	4-20 mA transmitter, analog input 2	60 s
PRV outlet setpoint	prv_setpoint_bar	bar	Server attribute on the device	on change	Set by the network team, used by alarm rules.
Chamber flooded	chamber_flood	bool	Float switch, digital input	on change
Kiosk door	door_open	bool	Reed contact, digital input	on change
Gateway supply voltage	supply_v	V	Gateway internal	15 min	Shows failing chamber power.
Critical point pressure	cp_pressure_bar	bar	LoRaWAN pressure logger	15 min	Highest or most remote customers.
Critical point minimum	cp_pressure_min_bar	bar	Logger, minimum since last uplink	15 min	Only if the logger reports it; catches short dips.
Critical point maximum	cp_pressure_max_bar	bar	Logger, maximum since last uplink	15 min	Only if reported; hints at transients.
Average zone point pressure	azp_pressure_bar	bar	LoRaWAN pressure logger	15 min	Basis for AZNP.
Logger battery	battery_pct	%	Logger	daily	Key name depends on the codec.
Pulse logger total	pulse_total_m3	m³	LoRaWAN pulse logger	15 min	Scaled by litres per pulse in the codec or a flow.
LoRaWAN gateway status	gateway health keys		Hermetry network server	per report	Stored as telemetry of the gateway device.

Asset model

In Sites & Assets, model the network as the leakage team thinks about it:

• Supply system (free kind, for example Upper zone system) with the source reservoir.
• DMA (Zone) with attributes properties, mains_length_km, night_use_allowance_m3h, mnf_step_limit_m3h and target_aznp_bar. Give each DMA a location and a time zone so time windows use local time.
• Inlet chamber and PRV kiosk (Site or Room) holding the gateway.
• Critical point and Average zone point (Area) holding a logger each.

Place each device on the asset it measures. Add Feeds relations from the inlet to the DMA and from a DMA to the neighbouring DMA it exports to; dashboards can follow relations to collect the inlets and outlets of a DMA for the net-inflow chart. The tree also scopes rules: a rule whose home is a DMA may only name devices placed under it, so the Rules card of a DMA shows everything its rules touch.

Device setup

Profiles

Create a DMA inlet gateway profile (MQTT, TLS required, Recent message online rule with a 60 s expected interval) and use the built-in LoRaWAN profile, or copies of it, for loggers: one per logger model, so each model keeps its own codec, telemetry keys and alarm rules. On the LoRaWAN tab of a profile set ADR, de-duplication, the codec and, where the supplier uses one, a shared AppKey. Battery loggers use class A.

Provisioning

Licensed gateways are claimed into the organisation by scanning their code. LoRaWAN loggers are added with their DevEUI, JoinEUI and AppKey from the supplier's sheet; the AppKey is stored encrypted and never shown again, so keep the sheet. LoRaWAN gateways are added with their EUI; Hermetry shows the gateway's token and Basics Station files once, so put them on the gateway immediately. Place each gateway on the gateway map by dragging its pin.

Settings, intervals and deadbands

Gateway reading intervals, Modbus addresses and analog input scaling are managed with remote device settings on licensed gateways: per device from its Settings tab, or for all inlet gateways with a baseline and a bulk job. A job runs in waves (1 %, 10 %, then the rest by default) and pauses itself if fewer than 95 % of the answering devices apply the change. Logger intervals are configured with the supplier's tools before installation.

Do not put a deadband on flow: MNF needs the low values exactly. A small deadband on pressure (for example 0.02 bar) is acceptable for powered sites if the gateway supports it.

The 15-minute average

Give the inlet gateway profile its own root flow, copied from the default one: it saves telemetry and runs the alarm rules as before, and a second branch passes inflow_m3h through a Statistics block (average over 15 minutes, one result per full window), renames the result to inflow_avg15_m3h and saves it as telemetry of the same device. Test the flow in the test runner before deploying; every deploy is a version you can roll back.

Time stamps and outages

MNF depends on correct time stamps. Gateways should send ts only with a synchronised clock; LoRaWAN readings are stamped on reception. Licensed gateways can buffer readings while the cellular link is down and send them later with their original time stamps; buffer size depends on the model. Late readings are stored and evaluated by alarm rules as a backfill, so a pressure dip during an outage still appears in the alarm history.

Firmware

Inlet gateway firmware is rolled out in waves with success gates and a soak time between waves. Gateways on mobile data wait for a firmware download until metered downloads are allowed in the OTA settings.

Dashboards

Board	Audience	Widgets
Network map	Control room, leakage team	Map of DMAs and loggers coloured by status; table of DMAs with last MNF, step change and open alarms; stat cards for DMAs above their step limit.
DMA page	Leakage analyst	Line chart of net inflow over 7 days with the previous week as a comparison series; bar chart of MNF per night for 90 days; gauges for critical point and average zone pressure; state card for chamber flood and door. Opened from the map with the DMA as a parameter.
Night flow league	Leakage manager	Table of all DMAs sorted by estimated leakage per property, downloadable to Excel; bar chart of the ten largest steps of the week.
PRV health	PRV crews	Line chart of inlet, outlet and setpoint per PRV; heatmap of outlet pressure by hour of day over 30 days, which shows pilot drift; table of PRVs outside their band.
Logger fleet	Instrument technicians	Device table with battery and last seen; alarm table of offline and low-battery alarms.

Key filters keep the boards useful at scale, for example a filter that shows only DMAs whose last inflow_avg15_m3h exceeds a threshold the viewer can change. Parameter values live in the page address, so an analyst can send a colleague the exact view of one DMA and week.

SCADA screens

A distribution network is drawn as a schematic, not as a plant. Two drawings usually suffice:

• Supply system schematic: source reservoir, trunk mains, DMA inlets with their meters and PRVs, and DMA boundaries as zones. Meter symbols print flow; PRV symbols print outlet pressure as the main value and inlet pressure as the second reading.
• PRV kiosk drawing: isolating valves, strainer, PRV, bypass, inlet and outlet pressure transmitters and the meter, reused for every kiosk of the same arrangement through the profile's Views button.

Valve, meter, reservoir and hydrant symbols come from the core set and the Water pack, which covers water, wastewater and city networks; see the SCADA library. Symbol rules colour a PRV amber when outlet pressure leaves its band and red when outlet is within 0.2 bar of inlet, and make a DMA zone label blink when its night-flow alarm is open.

Rules and alarms

Threshold alarms sit on the profiles; time-of-day logic sits in DMA Rules cards, because profile alarm rules compare values but do not know the time.

Alarm	Condition	Delay and hysteresis	Severity	Notified	Escalation
Critical point low pressure	cp_pressure_bar < 1.5 (Major), < 1.0 (Critical)	Debounce 2 samples; hysteresis 0.2 bar	Major / Critical	Control room, push; Critical also SMS to network duty engineer	Reminder every 30 min while unacknowledged
PRV not reducing	prv_out_bar > attribute prv_setpoint_bar + 0.5	Debounce 600 s; hysteresis 0.2 bar	Major	PRV crew, e-mail and push	Reminder every 60 min
PRV outlet low	prv_out_bar < attribute prv_setpoint_bar − 0.7	Debounce 600 s	Major	PRV crew	Reminder every 60 min
Night flow step	Rule: when inflow_avg15_m3h stays above the DMA limit for 60 min, only if between 02:00 and 04:00, then raise alarm	Held by the rule's 60 min	Minor	Leakage team, e-mail	None; reviewed next morning
Burst suspected	Flow: inflow rises by more than an agreed amount within 15 minutes	Debounce in the flow	Major	Control room, push	Reminder every 15 min
Chamber flooded	chamber_flood == true	Debounce 60 s	Major	Control room
Logger silent	Inactivity rule, 2 h		Warning	Instrument technicians, in-app
Logger battery low	battery_pct < 15	Debounce 2 samples	Info	Instrument technicians, in-app
LoRaWAN gateway offline	Inactivity on the gateway device, 30 min		Major	Control room, push	Reminder every 30 min

Per-DMA limits. Use device attributes for every limit that differs between sites (prv_setpoint_bar, a critical point's minimum). One rule then serves all DMAs, and changing a setpoint is an attribute edit, not a rule edit. Profile alarm rules compare a key with a number or a device attribute; the time-of-day part of the night-flow alarm comes from the DMA's Rules card.

Acknowledge and clear. Alarms open, are acknowledged by whoever takes them and resolve themselves when the condition clears; a severity rise notifies again. Leakage alarms are slow by nature, so give them a runbook text: which logger to check first and how to raise a leak search.

Planned work. When a valve crew shuts a DMA boundary or changes a PRV, mute the affected devices for the duration. Their alarms are still recorded and marked muted, but nobody is paged.

Floods of alarms. A trunk main burst drops pressure in several DMAs at once. Notification of the same change is limited to once per five minutes, and above 20 alarm notifications a minute the rest go into a five-minute digest. Keep logger alarms at lower severity than the inlet and PRV alarms, so the cause stands out.

Automation and integrations

• Burst flow: on the inlet gateway's root flow, a Delta block computes the change in flow since the previous reading per device, and a comparison raises the Burst suspected alarm through the Raise alarm block when the change exceeds the DMA's limit. Alarms raised by flows follow the normal lifecycle.
• Leak search tickets: a webhook on alarm.created, filtered on the night-flow alarm type in your receiver, opens a leak search job in the work management system. Webhooks are signed with HMAC and can use the CloudEvents format.
• Hydraulic model: analysts export flow and pressure history from tables to Excel, or read it through the REST API with an API key, to calibrate the model.
• Pressure management: remote setpoint changes on PRV controllers are outside this design; PRVs keep their own pilots or controllers. If an electronic PRV controller is fitted, its setpoint can be written as a setting where the gateway model supports it, after the network team agrees the procedure.

Customer billing, the GIS and the work management system remain the systems of record. Hermetry provides the live measurements and the events that trigger work.

Roles and access

Who	Access	Typical actions
Leakage analyst	User with the IoT viewer group	Read every board and history, export data.
Control room	User with the IoT operator group	Acknowledge alarms; muting needs an administrator.
PRV crew	User, limited by policy to PRV devices of their area	Acknowledge PRV alarms, see their kiosks.
Instrument technician	Admin for devices and profiles	Add loggers, manage codecs, change settings, run firmware waves.
Network manager	Admin	Alarm rules, rules, dashboards, members.

Attribute-based policies can restrict a crew to the devices of its area. Setting jobs, mutes, high-impact commands and firmware actions are written to the audit log, which can be verified and exported for internal audits.

Sizing and data

As an order of magnitude for a network of 60 DMAs:

• 40 powered inlet and PRV sites, one message a minute each: 40 × 1,440 = 57,600 messages a day.
• 180 LoRaWAN loggers at one uplink every 15 minutes: 180 × 96 = 17,280 messages a day.
• Total about 75,000 device messages a day, roughly 2.3 million a month.
• Values: about 40 × 8 keys × 1,440 + 180 × 3 keys × 96 ≈ 510,000 a day, close to 190 million a year, plus one computed 15-minute average per inlet.

MNF analysis needs at least a year of nightly values, and pressure studies need the raw readings for a few months. Pick a plan retention class that covers at least a year (the 365-day class keeps 90 days of raw readings and a year of hourly and daily aggregates), and use daily aggregates for multi-year MNF trends. Treat these numbers as planning estimates and confirm storage with us for your deployment.

Rollout plan

Phase	Weeks	Work	Exit criteria
0. Survey	1-3	Select pilot DMAs; check meter outputs, chamber power and cellular signal; radio survey for LoRaWAN; agree keys and limits.	Point list per site type; coverage confirmed at the worst logger spots.
1. Pilot	4-8	Three to five DMAs: inlet gateways, PRV pressures, loggers, the averaging flow, DMA board.	Integrated inflow matches the meter totaliser within 1 %; MNF available every night for two weeks.
2. Validation	9-12	Compare MNF and pressure with the existing manual process; tune limits and debounce.	Night-flow alarms agree with the leakage team's own review; nuisance alarms understood.
3. Rollout	13-30	Remaining DMAs in groups of ten; bulk settings jobs; firmware waves.	Every DMA reporting MNF; under 2 % of loggers silent.
4. Operation	ongoing	Monthly review of limits; codec copies maintained for new logger models.	Leak searches raised from alarms, not from monthly reviews.

Commissioning checklist

1. Gateway claimed, on the inlet profile, placed on the right chamber; certificate active.
2. Flow sign checked: a known flow direction shows as positive; reverse flow test where possible.
3. Integrated inflow_m3h over 24 hours compared with the totaliser increase.
4. Pressure transmitters compared with a calibrated test gauge at the tapping.
5. Each logger seen by at least two LoRaWAN gateways where possible; latest reception and signal recorded.
6. Codec output checked: units, scaling and key names match the point list.
7. ts checked against a known event; DMA time zone set on the asset.
8. Each alarm forced once (setpoint attribute lowered temporarily, flood float lifted) and notifications received.
9. Cellular uplink interrupted for 30 minutes; buffered readings arrive afterwards with their original times.
10. The averaging branch produces one value per 15-minute window on the DMA page.

Pitfalls

• Reading the pulse output without the scaling. Litres per pulse differ between meters; write it on the device and check against the totaliser.
• Using the one-minute minimum as MNF. Single low samples understate night flow; use the 15-minute average.
• Ignoring outlets. A DMA that exports water at night shows a false leak if the outlet meter is missing.
• Loggers placed where radio cannot reach. Survey first; underground chambers need tested antenna positions.
• One profile for every logger model. Different models decode differently; keep one profile per model.
• Time-of-day in profile alarms. Profile alarm rules do not know the time; put night windows in Rules.
• Alarms on every logger. A trunk burst would page everyone; let inlet and PRV alarms lead.

What Hermetry does not replace

Hermetry measures and alarms; it does not control the network. Pressure-reducing valves keep their pilots or local controllers, and boundary valves are operated by crews. Hermetry does not compute regulatory leakage indicators that need data from other systems, such as customer meter reads or mains records; it supplies the night flows and pressures those calculations use. The hydraulic model, GIS and billing stay where they are.