Connecting Waterpeople
Premium content

Turning an alarm flood into actionable alerts: Southern Water’s AI pump monitoring

Published in SWM Print Edition 28
Download PDF article
Download
08/01/2026

Southern Water has enhanced its handling of pump station alarms through the adoption of an AI-powered alerting system. In partnership with StormHarvester, the utility shifted from being reactive to proactive in its pump performance management, improving response efficiency, protecting the environment, and giving staff new confidence in the alerts they receive. The result is a network that is smarter, safer, and far more efficient – one alert at a time.

For years, Southern Water’s sewer network was monitored by static hi-hi alarms triggered whenever wet-well levels rose above a fixed threshold. During heavy rainfall, dozens of pump station alarms could sound at once, yet there was little insight into which sites truly needed attention. With no intelligent filtering, the operations team faced a barrage of alerts – most of them non-critical. Staff became overwhelmed and uncertain which alarms warranted action, undermining confidence in the system. Critical issues sometimes went unnoticed until a pump had already failed and an incident (like an overflow) was underway. This left Southern Water stuck in reactive mode: crews were dispatched after failures (often outside normal hours), leading to disruptive emergency callouts, higher costs, and increased operational strain. The need for a smarter, more discerning monitoring approach was clear.

For years, Southern Water’s sewer network was monitored by static hi-hi alarms triggered whenever wet-well levels rose above a fixed threshold

AI solution: smarter pump station alerts

Southern Water’s answer was to deploy StormHarvester’s Pump Station Alerting solution – a cloud-based platform that brings real-time intelligence to pump monitoring. This system continuously analyses each pump’s start/stop patterns, run times, and wet-well water levels, while correlating them with live local rainfall data to discern normal behaviour from anomalies. By leveraging machine learning, the platform learns what “normal” looks like for every station and can detect subtle deviations that signal trouble long before a pump fails. When the AI spots an anomaly, it issues a descriptive alert explaining the likely issue and context, rather than a generic high-level alarm. By filtering and interpreting the raw telemetry in this way, StormHarvester’s platform turns a flood of data into a trickle of actionable insights that operators can trust. Crucially, the alerting logic accounts for weather conditions. Under heavy rain, rising water levels are expected and do not trigger false alarms. The AI effectively “knows” when a spike in pump activity is justified by rainfall and when it is truly abnormal. This integration of hyper-local weather data prevents the nuisance alarms that previously overwhelmed staff, ensuring that when an alert comes through now, it likely signals a real issue in need of attention. As StormHarvester describes it, Pump Station Alerting uses simple pump telemetry combined with rainfall data and AI analytics to detect a wide range of potential pump performance problems early on. For Southern Water, this meant moving from hindsight to foresight in network operations – the alarm flood has been replaced by a concise stream of credible alerts.

Southern Water deployed StormHarvester’s Pump Station Alerting solution that brings real-time intelligence to pump monitoring

Implementing the new AI system across Southern Water’s vast territory (spanning Kent, Hampshire, Sussex and beyond) was an ambitious endeavour, yet the rollout was remarkably swift. Thousands of pumping stations were onboarded each week during deployment. In fact, most sites became “alert-active” within about two weeks of project kick-off. Even stations that initially needed sensor metadata updates or configuration tweaks were integrated shortly thereafter. StormHarvester’s engineers worked hand-in-hand with Southern Water’s operations staff to ensure a smooth transition. Daily check-in meetings were held during the first month to embed the technology into routine workflows and fine-tune the alert parameters in real time. This close collaboration helped build trust in the new alerts and transferred know-how to Southern Water’s team. After the intensive launch period, the cadence eased to weekly check-ins as the platform became part of day-to-day operations. By actively involving control centre operators in the process, the system was never a mysterious “black box” – it became a familiar tool that staff could understand and rely on. The rapid, well-supported rollout meant Southern Water began seeing benefits almost immediately, with minimal disruption to its existing processes.

From reactive to proactive operations

The impact of the AI alerting solution has been transformative. Southern Water has effectively moved from a reactive posture to a proactive one, using data-driven foresight to fix problems before they escalate. Instead of reacting to pump failures or sewer overflows after the fact, teams can now intervene at the first sign of a performance anomaly – often during normal working hours. Early warnings of developing issues have greatly reduced the frequency of middle-of-the-night emergencies. Machine learning-based dynamic thresholds (which adjust to weather and each station’s typical performance) are catching subtle faults that previously slipped through the cracks. Operators now receive alerts for the kinds of problems that once went unnoticed until they caused a failure. These alerts flag a wide range of underlying issues that can occur in pumping stations, enabling crews to address them proactively. For example, the system has identified issues such as:

  • Partial blockages or pump clogs (e.g. due to wet wipes, rags or debris restricting flow).
  • Worn or damaged pump impellers and components.
  • Pumps not properly seated or experiencing air lock issues.
  • Excessive buildup of debris in wet wells affecting capacity.
  • Stuck or faulty non-return valves (backflow prevention valves).
  • Tripped pump fuses or other electrical/control faults.
  • Failing level sensors providing incorrect readings.
  • Burst rising mains

By leveraging machine learning, the platform learns what “normal” looks like for every station and can detect subtle deviations

Each alert comes with context and a likely cause, which eliminates guesswork and ensures the response is targeted to the real issue at hand. After resolving initial data quirks and fine-tuning the system’s sensitivity, Southern Water’s alert “hit rate” – the percentage of alerts that lead to a verified issue – climbed dramatically. About 88% of alerts now correspond to real, actionable problems, meaning nearly nine out of ten times the alarm is worth staff attention. This high accuracy (far beyond the old hi-hi alarms) has given operators confidence that when the system pings, something is genuinely amiss. As the control centre saw the AI consistently point them to actual faults, they grew to trust the alerts and even began actively using StormHarvester’s pump station dashboard for real-time visibility. Instead of feeling overwhelmed, the team can clearly see what is happening at each site and prioritise work accordingly.

Southern Water’s leadership has noted qualitative improvements as well. Richard Martin, Head of the Operational Control Centre (Waste) at Southern Water, explained that their previous monitoring platform struggled with an “overwhelming amount of data waste” that was difficult to translate into actionable tasks. However, “By leveraging StormHarvester’s advanced machine learning and artificial intelligence capabilities, we have been able to significantly reduce this waste, enabling our teams to focus on meaningful, high-impact jobs. This shift has not only enhanced the quality and relevance of the insights we receive but also improved operational efficiency across the board,” Martin said.

He further praised the system’s user-friendly design, noting that StormHarvester’s intuitive interface and streamlined workflows have empowered staff to make faster, better-informed decisions. In short, the technology has improved both the precision of the utility’s data and the productivity of its people.

Pump Station Alerting uses pump telemetry, data, and AI analytics to detect a wide range of potential pump performance problems early on

One real-world example illustrates how these smarter alerts avert problems. In June, StormHarvester’s system detected that Pump 2 at one wastewater pumping station was taking longer than usual to drain the wet well after it started. This subtle change triggered a “pump running longer than normal” alert for the operations team. Trusting the AI, the team dispatched a maintenance crew to investigate. Upon inspection, they discovered that Pump 2’s impeller was damaged and a soft blockage – a mass of rag debris and even a brick – was partially obstructing the flow. These issues were causing the pump to struggle. The crew promptly pulled the pump, cleared the blockage, and repaired the impeller. Within a couple of days, Pump 2’s performance returned to normal, as confirmed by the telemetry on StormHarvester’s dashboard. A potentially serious pump failure or sewer overflow had been averted. What had started as a slight performance dip (invisible to legacy alarms) was caught by AI and translated into an actionable maintenance task. In the past, that minor clog might have gone unnoticed until it grew into a major incident – but now it was resolved before customers or the environment felt any impact.

By involving control centre operators, the system was never a mysterious “black box”, but a familiar tool that staff could understand

Southern Water’s journey is a compelling case of turning an “alarm flood” into a manageable stream of intelligence. By combining simple pump telemetry with weather data and AI analytics, the utility is now getting the right alert at the right time instead of an indiscriminate torrent. The results speak for themselves: alert accuracy has soared (nearly nine out of ten alerts are valid), response times are faster, environmental incidents are fewer, and staff are empowered rather than overwhelmed. This project has set a new benchmark for what proactive operations can look like in the water industry. By embracing StormHarvester’s innovative solution, Southern Water has not only solved a persistent alarm fatigue problem – it has created a smarter, safer, and more efficient way to run a wastewater network, one alert at a time.