Digital Energy Optimisation: Thames Water’s 30% Efficiency Shortcut
How digital energy optimisation at Thames Water's 350 treatment sites delivers efficiency gains comparable to major capital investments — without the capital intensity
The efficiency gap in legacy aeration infrastructure is recoverable through capital investment in variable speed drive retrofits — but the recovery is not automatic. A variable speed drive controller reduces blower output when the treatment process requires less air, and increases it when demand rises. The quality of this demand response depends on the quality of the real-time data informing the controller's decisions. A variable speed drive operating against design-basis oxygen setpoints — fixed parameters set at commissioning that do not reflect actual process conditions in real time — will underperform relative to one operating against continuous measurements of dissolved oxygen in the aeration tanks, ammonia concentrations in the influent, and biological oxygen demand calculated from real-time sensor readings. The physical retrofit creates the mechanical capability for variable-rate air delivery; the digital infrastructure creates the operational intelligence that determines how well that capability is used.
The Narrowband Internet of Things sensor network that Thames Water is deploying across its treatment estate addresses this data foundation requirement at the scale the estate demands. Narrowband Internet of Things technology provides low-power, wide-area wireless connectivity optimised for sensor-dense environments — the characteristics required to deploy continuous monitoring across treatment tanks, pump stations, blower rooms, and distribution network pressure zones without the cost and complexity of wired sensor networks. At a site the size of Beckton, processing over 3.6 million people equivalent, the number of process variables requiring continuous monitoring to support real-time energy optimisation runs to hundreds of measurement points. The Narrowband Internet of Things architecture makes this monitoring density achievable without a proportional increase in installation and maintenance cost — reducing the per-measurement-point cost to a level where comprehensive sensor coverage across 350 sites becomes a programme deliverable rather than a selective pilot.
Digital twin energy modelling converts the continuous sensor data into site-level operational decisions. A digital twin is a computational model of the treatment works that replicates the physical process relationships — how influent flow, organic loading, temperature, and biological population dynamics interact to determine aeration demand — with sufficient fidelity to produce reliable predictions of how process interventions will affect energy consumption. When the digital twin is fed real-time sensor data, it can calculate the energy-optimal aeration regime for current process conditions and compare it against the actual operating regime — identifying the gap between current practice and optimal practice, and providing the control signals to close it. At sites where the aeration control system is connected to the digital twin, the optimisation becomes continuous and automatic. At sites where the connection is not yet complete, the twin provides the analytical foundation for operational decisions made by process engineers using the model's outputs.
The 20 to 30 percent reduction applies to the process consuming 50 to 60 percent of total site energy — creating a 10 to 18 percent reduction in total site energy consumption from aeration optimisation alone. Digital twin integration maximises the realised saving by providing real-time process intelligence to the variable speed drive control systems that the physical retrofit installs.
Demand response integration extends the digital energy management programme beyond site-level efficiency to grid-level value creation. The electricity grid increasingly requires flexible demand — loads that can adjust consumption in response to grid frequency signals, price signals, or explicit demand response instructions from the grid operator. Water treatment works are well-suited to demand response participation: their process loads have flexibility that industrial processes with tighter quality constraints do not. Pump scheduling can be adjusted by hours within operational windows; aeration rates can be modulated within biological treatment constraints; tank levels provide storage that allows consumption to be shifted in time without process disruption. Demand response participation converts this operational flexibility into a revenue stream from grid services — payments for the grid balancing service provided by the flexibility — that supplements the energy cost reduction from efficiency optimisation. For Thames Water, demand response integration across 350 sites creates an aggregated flexibility volume that is material to grid balancing markets.
The data architecture required to make demand response participation reliable at grid-balancing timescales — response times of seconds to minutes — is more demanding than the data architecture required for efficiency optimisation alone. Grid frequency response requires that the site's automation system can act on a grid signal within defined response windows, without human intervention and without disrupting treatment process compliance. The Narrowband Internet of Things network provides the sensor layer; the digital twin provides the process model confirming that demand response activation is safe at current process conditions; the automation system provides the control execution. The integration of these three layers — sensor data, process model, and automated control — is the technical work that converts the efficiency potential of the digital energy management programme into the grid services revenue that makes demand response participation commercially justified.
Expert Follow-Up Questions
Why does the quality of real-time sensor data determine how much of a variable speed drive retrofit's efficiency potential is actually realised?
A variable speed drive adjusts blower output based on the demand signal it receives. If that signal is a fixed dissolved oxygen setpoint from commissioning, the drive modulates against a parameter that does not reflect actual process conditions — potentially over-aerating when biological demand is low, or under-aerating when influent loading spikes. Real-time dissolved oxygen measurement, continuous influent monitoring, and biological oxygen demand calculation give the drive a demand signal that reflects actual process conditions in the current moment. The difference between design-basis setpoint control and real-time process-responsive control is the difference between the theoretical efficiency potential of the retrofit and its realised operational saving.
What makes Narrowband Internet of Things technology suitable for sensor deployment across wastewater treatment sites?
Narrowband Internet of Things operates in licensed spectrum with low power consumption, long battery life, and strong signal penetration through building structures — characteristics that match the requirements of treatment works sensor deployment. Sensors installed in underground pump chambers, submerged in process tanks, or located in areas without mains power connectivity can transmit data reliably without wired infrastructure. The technology's low power consumption reduces battery replacement frequency, cutting operational maintenance cost for dense sensor deployments. At the scale of 350 treatment sites with hundreds of measurement points each, the cost advantage over wired or higher-power wireless alternatives is significant in aggregate.
What is a digital twin in the context of wastewater treatment energy management?
A digital twin is a computational model that replicates the physical and biological process relationships of the treatment works with sufficient fidelity to predict how changes in operating conditions affect process performance and energy consumption. In energy management applications, the twin models the relationship between influent flow, organic loading, dissolved oxygen levels, blower output, and energy consumption — allowing it to calculate the energy-optimal aeration regime for current conditions. When fed real-time sensor data, the twin continuously compares actual operation against calculated optimum and provides control signals or operational guidance to close the gap. The twin's value is proportional to the quality of its process model and the richness of its real-time data input.
How does demand response participation generate revenue from water treatment operational flexibility?
Grid operators pay for flexible demand — loads that can adjust consumption in response to frequency signals or explicit instructions — because flexibility supports grid balancing without the capital cost of dedicated energy storage. Treatment works flexibility comes from pump scheduling windows, aeration rate modulation within biological treatment constraints, and tank level storage that allows consumption to be shifted in time. Demand response aggregators or direct grid service contracts pay the treatment works for making this flexibility available and reliably delivering response when called. The revenue supplements energy cost reduction from efficiency optimisation, improving the overall return on the digital energy management infrastructure that enables reliable demand response participation.
What are the treatment process constraints on demand response participation at wastewater treatment works?
Demand response requires that consumption reductions do not compromise treatment process performance or compliance with environmental permit conditions. Aeration cannot be reduced below the biological minimum required to maintain acceptable effluent quality — doing so risks permit breach and ecological harm in receiving watercourses. Pump scheduling flexibility is constrained by storage tank capacity and the need to maintain network pressure. The digital twin's process model provides the real-time assessment of whether demand response activation is safe at current conditions — preventing activation when process margins are insufficient. This model-gated activation is the technical safeguard that allows participation in fast-response grid services without the compliance risk that unguarded consumption reductions would create.
The Nexus Solutions and Opportunities section of the Water-Energy Nexus in Thames Water report maps the digital energy optimisation pathway in detail — including the Narrowband Internet of Things sensor architecture across 350 treatment sites, the digital twin energy modelling capability and its process fidelity requirements, and the demand response integration with grid flexibility markets that converts operational flexibility into grid services revenue. The Operational Inefficiencies section quantifies the efficiency gap that the digital programme addresses and the site prioritisation logic that determines deployment sequencing before 2030.
