LONDON’S streets can be a bit of a maze, but below ground things are even more complex. Water pipes crisscross the city in all directions. Some areas used to have competing water companies, each of which built its own system. Not even Thames Water, the utility that operates the British capital’s water-supply network today, knows exactly where all the pipes run.

Moreover, the network is ageing. Only a few years ago more than half of the 10,000 miles (16,000km) of water pipes below the streets of London were over a hundred years old and often burst. It did not help that over many years Thames Water, which was privatised in 1989, failed to invest enough. By the mid-2000s London had one of the leakiest water-supply systems in the rich world. Every day nearly 900m litres of treated water were lost and 240 leaks had to be fixed.

Over the past five years, though, Thames Water has replaced 1,300 miles of cast-iron Victorian mains, those most likely to break, with plastic ones, reducing leakage to 670m litres per day. And when the firm puts in new pipes, it also installs additional wireless sensors, giving it a better view of its network. “We can now tell where we have a broken main even before customers call us,” says Bob Collington, its head of asset management.

Thames Water not only needs to know what is going on in its network, but to be able to act quickly on the information. The same is true of infrastructure operators around the world. Whether in water, power, transport or buildings, all are trying to turn their dumb infrastructures into something more like a central nervous system. That makes them pioneers of the convergence of the physical and the digital world.

Putting sensors and actuators (devices to control a mechanism) into physical infrastructures is not exactly new. Known as “supervisory control and data acquisition”, such systems have been around for decades. But many still require human intervention: workers have to be sent out to download sensor readings or to fix problems. And even if sensors and actuators are connected, different types often feed into incompatible systems, so they cannot be easily combined to automate processes.

The operations centre of Thames Water in Reading, to the west of London, is a good place to see both the old and the new—and soon the future. A big video screen shows expected precipitation over the next few hours, and workers monitor the water level of reservoirs on their own screens. But if one of the pumps fails, they may still have to make a call: not all the valves can be remotely controlled.

Thames Water is investing £100m ($158m) so it can take action remotely and automate a lot of its processes. If the project works, the system will not only automatically deal with leaks but also schedule work crews and send text messages to affected customers. Employees in the operations centre, explains Jerry White, the utility’s head of operational control, will then spend less time monitoring the network and more on making the utility’s processes more efficient.

A big chunk of this work will be analysing the data collected by all the systems and correlating them with other information. Not every unexpected spike in the water flow is the result of a leak, says Mr White. For instance, water use leaps after dark during Ramadan and at half-time during World Cup football matches.

One day soon Thames Water may even be able to send out work crews before a main actually breaks. In early 2010 the firm began using a web-based service provided by TaKaDu, an Israeli company, that acts as the network’s “eyes and ears”, in the words of Amir Peleg, its founder and boss. The firm analyses historical and online data to provide a basis for comparison, enabling its algorithms to detect things that are about to go wrong.

Similar progress is being made all over the world. The scope for preventing waste is enormous, in the water industry and elsewhere. Power utilities are well ahead, not least because they can use the grid itself to collect sensor data and control switches. Transport systems are behind, particularly roads, which often use nothing more than traffic cameras. Even buildings are getting more automated, with continuous checks on their energy use.

At the edge

For infrastructures to become truly smart, however, it is not enough to put more intelligence into the core of a network. The edge—the interface with users and devices—also has to become clever. This is the idea behind smart metering, which has made a good deal of progress in the power industry. According to Accenture, a consultancy, there are about 90 smart-grid projects around the world today. By the end of last year more than 76m smart meters had been installed worldwide. That number will almost treble by 2015, to 212m, estimates ABI Research.

Smart meters and other gear needed to make grids more intelligent will not come cheap. Morgan Stanley, an investment bank, predicts that the worldwide smart-grid market alone will grow from $20 billion last year to $100 billion in 2030. Yet the benefits also promise to be huge: power savings, reduced investment in electricity generation and lower carbon emissions.

The place to go to see the technology in action is Boulder, Colorado, home to what is considered the world’s first fully fledged “smart grid”. The local utility, Xcel Energy, did not skimp. It deployed equipment that automatically reports power cuts. It installed more than 20,000 smart meters, connected them via a fibre-optic network, launched a website to track power use and has started to offer pricing plans that encourage shifting consumption to off-peak hours. It has even equipped some households with gear that tells air-conditioning systems to turn themselves off when demand for electricity is high, a mechanism called “demand response”.

The results so far are mixed. The system has certainly helped Xcel to run its grid more efficiently. The utility now knows what is happening in its network and power cuts have become rare. Problems can be pinpointed and fixed much more quickly. But customers are not using much less power than they did before.

Yet it is early days. Some firms are already beginning to show what can be achieved with demand response. EnerNOC, an American energy middleman, for instance, pays other firms for allowing it to shut down their non-essential gear at times of peak demand, thus freeing up capacity. By mid-year some 3,300 customers, from steel plants to grocery stores, had signed up. Their combined consumption, which can be made available to other users if needed, is 4,800MW, exceeding the output of America’s largest nuclear plant.

The ultimate point of smart grids, however, is to allow dynamic pricing, with electricity charges fluctuating in response to demand. This could cut power demand by 10-15% during peak hours, estimates Ahmad Faruqui of the Brattle Group, a consultancy—more than twice the reduction likely to be achieved by just giving customers real-time information about their usage. That number could easily double again, he says, with a combination of dynamic pricing and demand response.

The main objective of smart power meters is to lower the peak load and thus enable utilities to keep down their peak generating capacity. In the water industry the economics are somewhat different, explains Stefan Helmcke, a water expert at McKinsey. Water can be easily stored and consumers have less discretion over when they use it (for instance, people cannot defer going to the toilet, which uses more water than any other activity at home), so the case for smart water meters is weaker.

Yet they are spreading all the same. Boston has long been the shining example. As early as 2004 the city’s Water and Sewer Commission had equipped almost all its customers with wireless smart meters. But it will soon be outdone by New York, which plans to install more than 800,000 of the devices at a cost of about $250m. Even Thames Water, most of whose customers have no meters of any sort, is now planning to install some of the smart kind.

Getting on board

In transport the equivalent of a smart meter is a vehicle’s on-board unit. That used to be a simple device, working like a radio-frequency identification tag when it passes under a gantry on a toll road, but it is also getting smarter. Germany’s Toll Collect system, which ensures that lorry drivers pay for using the country’s crowded motorways, relies on gadgets that are in some ways as clever as a smartphone. Among other things, they keep track of their position with the help of GPS, the satellite-based global positioning system.

Such toll systems are multiplying, particularly in big congested cities, including London and Stockholm. But it is Singapore that leads the pack. The city-state not only charges drivers for using much-travelled roads (driving on an expressway can be S$6, or $4.60); it also adjusts traffic lights to suit the flow of vehicles, uses data collected by taxis to measure average speed and is developing a parking-guidance system, noting that cars looking for somewhere to park are now a big cause of congestion.

Singapore may also become the first city to introduce real-time dynamic pricing on its roads. In 2006 the Land Transport Authority tested a traffic-prediction system built by IBM to set the tolls. And next year it plans to test a satellite-based system that does not require gantries and can charge according to how congested a road is at that particular time.

Another of the island’s infrastructure-management systems has become a model for the world: that for water. At the information centre at the southern tip of the island, next to the Marina Barrage, visitors can literally get a taste of it by picking up a bottle of “NEWater”, waste water that after extensive treatment has become potable again. But most of the treated water is fed back, via a separate distribution system, to Singapore’s factories and power plants—and then treated again.

This closed loop is part of a water-supply system in which “every drop counts,” in the words of Yap Kheng Guan, a director at the island’s Public Utilities Board (PUB). The Marina Barrage is another case in point. It was inaugurated in 2008 and acts as a tidal barrier to keep seawater out, thus turning the island’s most populated district into a water-catchment area and the harbour into a reservoir. When two other reservoirs are opened next year, more than two-thirds of Singapore’s territory will be used to catch rainwater.

The city-state’s desalination plants are also among the world’s most efficient. All this means that the island—smaller than Luxembourg and home to nearly 5m people as well as an economy nearly as big as that of Hong Kong—is able to meet more than 60% of its water needs on its own. But it wants to go even further: 50 years from now it hopes to be self-sufficient.

Sensors play a relatively small part in Singapore’s water management because the infrastructure is so new. On average there is only one leak a day. The PUB puts sensors only in a few key spots, for instance where water leaves the reservoirs. Should the system detect a dangerous contamination, that part of the network can be shut down immediately. And if heavy rainfall in central Singapore threatens to flood the city during high tide, seven huge pumps next to the Marina Barrage start to push water into the sea at 40 cubic metres per second each.

So far Singapore has no smart water meters, and at the moment there is no pressing need. Most Singaporeans live in multi-storey apartment buildings, which makes it easy to read meters. But if the PUB wants to reach its target of cutting daily domestic water use per person from 155 litres in 2008 to 147 litres by 2020 (about the same as in India, and a quarter of the figure in America, see chart 3), Singapore will have to become smarter still—and set yet another example.