A water supply network loses water in silence: through bursts nobody sees, through joints that drip for years and through meters that record less than what passes through them. IoT sensing in water networks turns that silence into data, and the data into two things an operator needs: knowing where water is being lost today and anticipating where it will be lost tomorrow. It has also stopped being optional: the EU Drinking Water Directive and Spain’s Royal Decree 3/2023 (Real Decreto 3/2023) require structural leakage to be assessed and reported, and the PERTE for digitalising the water cycle (one of Spain’s strategic projects for economic recovery and transformation) has financed the installation of sensors and platforms in consortia, associations of municipalities (mancomunidades) and local councils. This guide explains what is measured, what is calculated, what the models predict, what the legislation requires and how to approach a project, drawing on CEDESA’s experience with the Consorcio de Medio Ambiente de Badajoz (the environmental consortium of the province of Badajoz).

What a network fitted with sensors measures

The sensors installed in a water supply network come in only a few types, placed in many locations:

Flow meters at abstraction points, service reservoirs, sector inlets and large consumers, to know how much water enters and leaves each section.
Pressure sensors at critical points and at the extremities of each sector, to detect abnormal drops and to manage pressure, which is the first lever against leakage.
Level sensors in reservoirs, for the water balance and for pump control.
Quality probes: residual chlorine, turbidity, pH, conductivity and temperature in reservoirs and at points in the network, for the continuous public health monitoring required by drinking water legislation.
Smart meters with remote reading at service connections, which replace periodic readings with hourly or daily ones and make it possible to calculate the actual consumption of each sector.
Acoustic sensors (noise loggers) on the pipe, which detect the characteristic sound of a leak.
Equipment status: pumps, valves, variable-speed drives and energy consumed, for maintenance and energy efficiency.

Communications are handled by low-power, long-range networks (LoRaWAN, NB-IoT) for the scattered sensors, by conventional telemetry or fibre for the points that have a power supply, and by integration with the existing SCADA system where there is one. What matters is not the sensor but that its data reaches a platform that understands it, with a time stamp, a quality flag and a location.

What is calculated from that data

Water balance by sector and by period: water supplied against authorised consumption, which gives non-revenue water, the sum of real losses (leaks) and apparent losses (metering errors, unauthorised consumption).
Sectorisation into district metered areas: dividing the network into zones with metered inlets in order to pinpoint which sector the loss is in.
Minimum night flow: consumption between two and four in the morning is almost all leakage; how it evolves sector by sector is the most sensitive indicator of a new burst.
Standardised loss indicators, among them the Infrastructure Leakage Index (ILI) and litres per service connection per day, which allow sectors and years to be compared and are the ones the legislation asks for.
Pressure profile and its relationship with bursts, for adjusting pressure-reducing valves and pumping.
Abnormal consumption by meter: leaks inside homes, stopped meters, consumption that does not fit the pattern.
Energy efficiency of pumping, which in many networks is the largest operating cost.

What the models predict

With enough history, machine learning models add foresight:

Early leak detection: they learn each sector’s normal pattern of flow and pressure by hour, day of the week and season, and raise an alert when the sector deviates from it, before the leak surfaces.
Demand forecasting hours and days ahead, for scheduling pumping when energy is cheap and for sizing reservoirs.
Probability of a burst for each pipe section, combining material, age, diameter, pressure, failure history and soil, so that pipe renewal is prioritised on sound criteria and not simply by age.
Quality prediction: how residual chlorine evolves and the risk of non-compliance at remote points.
Detection of fraud and measurement errors in meters.

This is the approach CEDESA applied in the project for the Consorcio de Medio Ambiente de Badajoz: prediction algorithms for water supply networks, consumption optimisation, leak detection and demand forecasting using machine learning, financed by PERTE Agua under Spain’s Recovery, Transformation and Resilience Plan (PRTR), as set out on our innovation page.

What the legislation requires on leakage

Directive (EU) 2020/2184 on the quality of water intended for human consumption introduced, in Article 4, an obligation on Member States to assess water leakage levels in their territory, using a common indicator, and to take measures to reduce them. In Spain it is transposed by Royal Decree 3/2023 of 10 January, which sets the technical and health criteria for the quality of drinking water and, in its annex on structural leakage, the methodology and the timetable for operators to assess and report their losses. The practical consequence for a consortium or a local council is that it needs to measure: without sector flow meters, remote meter reading and a water balance, no assessment is possible, still less a reduction plan. Sensor deployment stops being an efficiency project and becomes the tool with which an obligation is met.

On top of that come the continuous analytical monitoring of water quality, which the same royal decree reinforces, and data protection in remote meter reading, which produces personal data on consumption per household and calls for data minimisation, information and security.

How it is funded

PERTE for digitalising the water cycle: this strategic project for economic recovery and transformation, approved in 2022 as part of the Recovery Plan, has financed – through funding calls from Spain’s Ministry for the Ecological Transition (Ministerio para la Transición Ecológica) – projects for sensor deployment, remote meter reading, data platforms and predictive models run by consortia, associations of municipalities, local councils and operators. It is worth checking which calls are still open or in delivery, and the reporting and justification obligations that come with them, which we explain in digitalising the reporting and justification of EU funds.
ERDF funds 2021–2027 managed by the autonomous communities (regions), with lines for water efficiency and digitalisation.
Tariffs and savings: the reduction in non-revenue water and in pumping energy translates into less water abstracted and treated per cubic metre billed; in many networks the project pays for itself out of that saving.
Public procurement of innovation, when the solution does not exist on the market, as we explain in how to tender a development that does not yet exist.

How to approach the project

Baseline audit: what is measured today and how well, which SCADA and billing systems exist, and what the current water balance is, even if only approximately.
Sectorisation and measurement points: designing the district metered areas and deciding where flow meters, pressure sensors and quality probes go, giving priority to the sectors with the greatest losses.
Data platform: ingestion from every source with a time stamp and quality flag, a single data model of the network (assets, sectors, sensors, meters), integration with SCADA and with the billing system, and dashboards.
Standardised indicators and alerts, with the calculation of the balance and of the leakage index ready for the assessment the legislation requires.
Predictive models once there is enough history, starting with anomaly detection by sector and demand forecasting.
Operation: procedures for responding to alerts, leak pre-location and repair campaigns, and measurement of the effect on minimum night flow.
Security and data: the platform is a system belonging to a public operator or a concession holder and processes personal consumption data; conformity with the National Security Framework (Esquema Nacional de Seguridad, ENS) and data protection by design, as we explain in what the ENS is.
Ownership and exit: the historical data and the models belong to the network operator; the tender specifications must guarantee their portability.

Frequently asked questions about IoT sensing in water networks

Which sensors are installed in a water supply network?

Flow meters at abstraction points, reservoirs and sector inlets; pressure sensors at critical points; level sensors in reservoirs; quality probes (chlorine, turbidity, pH, conductivity); smart meters with remote reading at service connections; acoustic loggers for leaks; and status and energy sensors on pumps and valves, communicating over low-power networks or integrated into the SCADA system.

How is a leak detected with sensors?

By sectorising the network, measuring the inflow to each sector and watching the minimum night flow: when it rises without any change in consumption, there is a new leak. Machine learning models refine that detection by learning each sector’s normal pattern and alerting to deviations before the leak surfaces; acoustic loggers pre-locate it on the pipe.

Is it mandatory to assess leakage in a water network?

Yes. Directive (EU) 2020/2184 obliges Member States to assess leakage levels using a common indicator and to take measures to reduce them, and in Spain Royal Decree 3/2023 sets the methodology and the timetable for operators to assess and report structural leakage. Without measurement by sector it is impossible to comply.

What does the PERTE for digitalising the water cycle fund?

Digitalisation projects by operators and public authorities in the urban water cycle: sensor deployment, remote meter reading, data platforms, predictive models and better management, through funding calls from the Ministry for the Ecological Transition under the Recovery Plan, with the reporting and justification obligations that apply to EU funds.

What personal data does remote meter reading generate?

Consumption per household at hourly or daily resolution, from which the occupants’ habits can be inferred. This is personal data: customers must be informed, resolution and access must be limited to what is necessary, the platform must be protected in accordance with the National Security Framework, and the processing must be entered in the record of processing activities.

Conclusion

Fitting a water network with sensors means measuring flow, pressure, level, quality and consumption where it is needed, using that to calculate the water balance and the leakage index the legislation requires, and using models to anticipate leaks, demand and bursts. Directive 2020/2184 and Royal Decree 3/2023 have made it an obligation, and the water PERTE has funded it. If your consortium, association of municipalities or council wants to move from monthly readings to real-time data, at CEDESA we have done it in Badajoz; tell us about your project via our contact page.