Real-time control for combined sewer overflows: an Australian case study

Wet-weather sewage management is becoming more important as Australian cities experience intense rainfall, infill development and tighter expectations around waterway health. A combined sewer overflow (CSO) occurs when a shared pipe network receives more stormwater than the downstream system can convey or treat, causing a controlled or uncontrolled discharge to a creek, river or harbour.

Australia has fewer historic combined systems than many North American and European cities, but the risk is still relevant in older urban catchments, mixed drainage networks and wastewater systems affected by inflow and infiltration. The following case study uses a representative inner-metropolitan catchment to show how a real-time control (RTC) system can reduce overflow frequency while maintaining safe, reliable operations.

The setting reflects practical conditions familiar to Australian water professionals: short, intense storms, dispersed asset ownership, environmental licence requirements and a strong preference for upgrades that deliver measurable value without major civil construction. It also shows why automation must be treated as an operational change programme rather than simply an instrumentation project.

The catchment and the operational problem

The utility in this case manages a 42-square-kilometre catchment draining towards a tidal estuary near Sydney. The network includes legacy combined pipes in older suburbs, newer separated sewers, two offline storage tanks and a wastewater treatment plant with limited wet-weather headroom. Local councils control much of the stormwater network, while the utility operates the sewerage assets and treatment facility.

During moderate rain, the system performs adequately. A high-intensity storm, however, can fill the trunk sewer within minutes. Once the upstream level reaches the overflow crest, sewage diluted by stormwater is discharged to the estuary. The utility had reliable historical records of pump starts and overflow alarms, but those data arrived too late to support active control.

The environmental regulator required a reduction in annual overflow volume and improved event reporting. Community concern was also growing because visible debris and odour occasionally appeared near a popular walking path along the foreshore. Building a larger trunk sewer was estimated to cost more than $40 million and would involve major road closures, so the utility investigated whether existing storage and conveyance capacity could be coordinated more intelligently.

Building the control strategy

The project team began with a hydraulic model calibrated against level, flow and rainfall data from 18 storm events. The model represented pipes, pump stations, storage tanks, overflow structures, treatment-plant constraints and tidal backwater conditions. Calibration was particularly important because a model that looks sound in dry-weather simulations may give poor decisions during fast-moving thunderstorms.

The selected strategy used upstream level sensors, ultrasonic flow meters, rain radar feeds and control valves at two storage facilities. A central supervisory platform received live data every minute and calculated the preferred operating state. When forecasts indicated a developing storm, the system pre-emptively emptied available storage, subject to minimum operating levels and downstream treatment capacity.

During the storm, the controller prioritised storage before the critical trunk sewer filled. It could throttle inflows, alter pump sequences and hold selected gates closed when the treatment plant was approaching its hydraulic limit. A rules-based fallback remained available if communications failed or data quality fell below an agreed threshold.

Sensors, communications and safeguards

The hardware package included redundant wet-well level sensors, radar-based rainfall data, flow measurement at key junctions and remotely operated actuators. Each critical signal had a plausibility check. For example, a sudden level change without a corresponding flow or rainfall pattern was flagged for review rather than passed directly to the control algorithm.

The communications architecture used a secure utility network with cellular backup at remote sites. The design team avoided making the entire system dependent on cloud connectivity. Local programmable logic controllers could maintain safe pump and gate sequences if the central platform became unavailable, and operators could return every asset to manual control.

Safety constraints were hard-coded into the system. Pumps could not be commanded below their minimum submergence level, storage could not be drawn down below emergency reserve, and the controller could not send more flow to the treatment plant than the approved wet-weather operating envelope allowed. Cybersecurity reviews, role-based access and change logging formed part of commissioning rather than being added later.

Working with Australian weather and regulation

Forecast uncertainty was a major design issue. A Bureau of Meteorology rainfall prediction can indicate a high-risk cell, but the most damaging rainfall may occur in a narrow band that changes direction quickly. The controller therefore used forecasts as one input, combining them with local rain gauges, radar updates and observed pipe levels.

The operating team also had to align the project with state environmental licences and incident-reporting obligations. In New South Wales, the utility needed clear evidence that each overflow reduction was real, while also demonstrating that automated decisions did not increase risks elsewhere in the system. Coordination with the local council was essential because blocked pits, construction runoff and stormwater diversions could affect the sewer model.

The language of operations mattered too. Site crews needed simple alarms and clear escalation paths rather than a dashboard filled with technical terms. The project documentation used familiar Australian workplace language: who is on call, what happens after hours, when to ring the duty manager and which failure requires a truck roll. That practicality helped adoption across engineering, treatment and field teams.

Testing before live operation

The utility introduced the system in stages. First, the RTC platform operated in “shadow mode”, receiving live data and generating recommendations without moving any equipment. Operators compared those recommendations with actual decisions during several rainfall events and identified issues such as delayed level signals and an unreliable valve position indicator.

Next came controlled trials at one storage tank. The team tested pre-release sequences during forecast rain, verified that pump starts matched the hydraulic model and checked that tank drawdown did not cause odour or ragging problems. Every trial had a stop condition, an assigned controller and a post-event review.

The utility also developed a digital twin for scenario testing. Engineers could replay historic storms, test sensor failures and examine the consequences of a forecast that overestimated or underestimated rainfall. This approach complemented broader treatment assessment work, including algae treatment performance where biological processes and changing water quality conditions require similarly careful interpretation of field data.

Results after the first wet season

After twelve months of operation, the representative catchment recorded a 38 per cent reduction in overflow volume and a 27 per cent reduction in overflow duration compared with the modelled baseline. The number of small nuisance discharges fell most noticeably because the system could empty storage between storm bursts. The largest storms still produced overflows, but the peak discharge was delayed and reduced.

The project also improved operational visibility. Operators could see storage availability, incoming rainfall and treatment-plant constraints on one screen instead of making decisions from separate control-room displays and phone calls. Maintenance teams used event records to identify a partially obstructed valve that had previously been assumed to be responding correctly.

The benefits were not limited to compliance. By reducing avoidable wet-weather inflows to the treatment plant, the utility lowered the need for emergency bypass arrangements and gained better information for future capital planning. The data also supported discussions about whether additional treated-water storage could be integrated into a wider municipal strategy, alongside a water reuse master plan.

Lessons for implementation and skills

The first lesson was that good control depends on good data. Several early failures came from instruments that were technically operational but poorly installed, fouled or exposed to conditions outside their stated range. A realistic allowance for calibration, cleaning and replacement is essential in wastewater environments.

The second lesson was that operators must help design the system. Their experience identified practical constraints that were absent from the hydraulic model, including access during storms, pump restart behaviour and the time required to confirm a blocked screen. Training included simulator sessions, wet-weather drills and MOC-aligned professional development so staff could understand both the technology and its operational consequences.

Finally, the business case had to measure more than avoided construction. The utility tracked overflow volume, duration, maintenance response, energy use, alarm quality and operator interventions. That balanced scorecard showed where RTC delivered value and where conventional works were still needed. For Australian utilities, the most effective solution may be a combination of smart control, targeted storage, source reduction and ongoing catchment renewal.

A real-time control system is most useful when it turns scattered information into timely, defensible decisions. The practical takeaway is to begin with reliable measurements, explicit safety limits and operator-led testing, then expand automation only after the catchment’s real behaviour has been demonstrated in live wet-weather conditions.