Young Wharton entrepreneurs at the IGEL-Suez conference on smart cities expressed frustration in dealing with utilities. According to Claire Tram, Wharton class of 2018, “Entrepreneurs are giddy to get into this market, but utilities are a daunting problem.” Tram understands that state-regulated utilities, responsible for critical services, are risk-averse for good reasons, but she also knows investors are impatient for returns. “When you’re trying to manage investors who want to see growth in 18 months,” she explained, “it’s hard to work with such slow-moving companies.”

Utilities may move more cautiously than entrepreneurs like Tram would like, but in many ways they are taking the lead when it comes to smart cities. When Debra McCarty, another speaker at the conference, began her career with the Philadelphia Water Department (PWD) in 1982, water meters all around the city were read each month by people employed for that purpose. Or not. Sometimes, the meter readers didn’t get inside the house. “Sometimes, they did what we called ‘curb reads,’” says McCarty, who is now commissioner of PWD. Such guesswork inevitably led to some inaccurate data.

Then, in 1997, the city took a first early step towards smart technology. PWD installed 480,000 new water meters that automatically sent readings to the utility on a regular basis. “It was a significant capital investment by the department,” McCarty told conference attendees. But thanks to the new Automatic Meter Reading (AMR) system, the department no longer had to send employees out to every building, curb reads all but disappeared, accuracy improved and customers were pleased.

Now that the 20-year AMR contract has ended, PWD is taking its smart technology to a new level, creating an interactive meter-reading system known as Advanced Metering Infrastructure (AMI). The new system, which makes use of existing smart meters, offers both customers and PWD a new level of real-time information. Customers can opt in to receive messages alerting them to abnormally high water usage, often caused by a leak, well before their monthly bill has climbed. They can also use detailed data about their usage to lower their water bills and reduce their environmental footprint.

Better Service

The department benefits, too. The AMI system alerts the utility to suspected meter tampering and possible contamination resulting from reverse flow. Better information means PWD can more often help customers resolve problems over the phone without the inconvenience of a repair call. And it can use hard data, rather than customer guess work, to right-size pipes that need to be replaced or installed.

“We need to understand what already exists and think about how we are going to leverage the networks we already have.” –Ellen Hwang

While these innovations can help reduce customers’ bills, they can also lead to revenue recovery for the utility, said David Stanton, president of Suez in North America’s utility operations and federal services division. According to Stanton, data provided by the largest AMI network in North America — Suez works with 18 separate utilities — has shown that well-run companies are sending out inaccurate bills 6% to 8% of the time. “And it’s always wrong in the wrong direction,” he said. “We have not yet found one instance where we were over-billing, but we have been significantly under-billing a lot of people.”

It takes a combination of human and machine intelligence to capitalize on the data pouring in from smart systems like AMI. “These systems are automated, but they’re not autonomous,” said University of Pennsylvania Environmental Sustainability Director Dan Garofalo. “So we need to have people in position who know what they are looking at and can interpret it very quickly.” According to Garofalo, the university spent the past 10 years installing smart sensors in buildings across its 302-acre campus, while also developing the computer systems to store and access the resulting data. Only now is Penn ready to start figuring out how to take full advantage of the input, he said.

One important step is presenting the data visually, so trained staff can put it to use almost immediately. The goal is two-fold: to save the university money in the near term by catching problems early on, and to improve the accuracy of budgeting long-term by using the quickly accumulating wealth of historical data to accurately anticipate the university’s future utility costs.

“But what we’re really focused on is being able to use predictive modeling to identify peaks,” said Garofalo. Predicting energy usage for a campus full of buildings is a complex undertaking. A wide variety of data must be collected from numerous sources, including weather stations (which provide readings of outdoor temperature, humidity, wind speed and solar radiation, among other data points); metered consumption of electricity, chilled water and steam for every building; and occupancy rates at every hour of the day for all types of rooms (dorm rooms, offices, labs).

Machine Intelligence

It takes machine intelligence to make sense of this much diverse data and predict future energy peaks. But first, incomplete, incorrect, inaccurate or irrelevant data have to be identified and cleaned up. Computer scientists then have to harmonize the myriad data formats and use the resulting datasets to train machine learning systems, which continually refine their predictions as new data comes in.

The payoff is a substantial reduction in energy costs through “peak shaving,” a way of reducing energy costs by buying power during off-peak hours, when it is markedly less expensive, and storing it for use when demand is highest. Garofalo said the university has numerous ways to store energy on campus. When it comes to cooling, ice tanks are key. Alerted to future cooling demand peaks by predictive modeling, the university can over-cool when usage is low, then turn off its energy-hungry chillers and use the stored cold water when demand peaks.

Smart cities can save money by taking advantage of existing projects and technology. Steve Davis, who worked as GE’s Digital Business Transformation Leader for 10 years, urged smart city pioneers to carefully evaluate what kinds of data they need most and what they are already collecting. “You might not have to invest a lot in acquiring new data,” he said. “Just start getting the data you have connected, so you can get it to people who can analyze and draw insights from it.”

GE Power, Water & Process Technologies (recently acquired by Suez) already had automated sensors monitoring water quality in industrial cooling towers, Davis explained. The data they generated was crucial to optimizing the towers’ performance. “But you still had to be at the site to catch the data at the time the system had an issue,” he said. “Otherwise all that data just sat in the box.”

“Just start getting the data you have connected, so you can get it to people who can analyze and draw insights from it.” –Steve Davis

Once GE connected the water-monitoring sensors, however, people quickly identified fundamental changes that reduced costs. The data made it painfully clear, for instance, that energy-hungry pumps and fans were operating at the same level all the time, regardless of demand. “We learned to turn the system down at night,” said Davis. “Run only when you need to run.”

GE also used the networked data to make its service calls more responsive to customer needs. Instead of routinely showing up on Fridays, service representatives began showing up when the data showed there was an actual problem. Sometimes, thanks to predictive modeling, they could even show up before a problem had surfaced. GE was able to avoid unnecessary service calls and improve customer service at the same time, without having to install any new sensors.

It can take some detective work to identify valuable existing resources. “There are so many networks being built,” said Ellen Hwang, Philadelphia’s program manager for innovation management. “We need to understand what already exists and think about how we are going to leverage the networks we already have.” When new networks are needed, it’s equally important to coordinate the work, which often involves digging up city streets. With cross-department planning, the disruption such work entails — and the expense — can be drastically reduced, said Hwang.

Infrastructure, too, can be shared. Patrick Cairo, emeritus member of IGEL’s advisory board, noted that as networks are being built to meet the needs of utility customers and companies, “They can also serve as a backbone for other services.” McCarty suggested, for instance, that other utilities and city departments could piggyback on the infrastructure PWD is building to gather data from water meters across the city, just as cell phone base stations piggyback on everything from high rises to newsstand kiosks.

To take advantage of coordinated efforts like this, city departments that tend to work in isolation have to start communicating with each other. Rubicon Global is a technology company focused on improving the efficiency and sustainability of commercial waste. About a year ago, the company began to adapt its smart technology for use in cities. According to Michael Allegretti, Rubicon’s senior vice president for policy and strategy initiatives, one of the company’s goals was to use its technology to serve the needs of both sanitation and sustainability departments, “which are historically siloed and working towards totally different policy goals.”

“My definition of a smart city is an interconnected one.” –Michael Allegretti

Rubicon began its new municipal initiative in three cities, Atlanta, Santa Fe, N.M., and Columbus, Ohio. Equipped with Rubicon technology, garbage trucks in these three areas now collect and share data with both sanitation and sustainability departments in the cities. Sensors in the trucks gather data that helps improve the efficiency of trash collection (when is trash being picked up, how well are different routes and trucks performing), while other data (what’s in the waste, how does that vary from neighborhood to neighborhood, what are recycling contamination rates) increases the amount of waste being diverted from landfills.

Rubicon envisions ways its technology can bring together other siloed departments as well. “We want to turn the garbage truck into a roving data center,” said Allegretti. “Garbage trucks are going up and down every street in every city in the world at least once a week. That’s a huge untapped potential for governments.” As they collect garbage, the truck can also collect information on everything from downed power lines to abandoned cars, from air quality to noise levels. And this data can be updated on a weekly basis.

“My definition of a smart city is an interconnected one,” said Allegretti. “If we get all the departments playing off the same sheet of music, it would be a big step. And garbage trucks, of all things, can be the thing that collects all that information, brings it back to one place and distributes it to all the different departments so they can act on it.”