When an institution has many buildings to maintain, there will inevitably be problems that don’t manifest themselves for long periods of time. A leak in a valve or a faulty damper can go unnoticed for quite awhile, resulting in higher energy costs.

By Peter Hildbrandt

Stanford University receives energy from a variety of different sources, including Cardinal Cogen, Pacific Gas & Electric Co., and City of Palo Alto Utilities. This comes as electricity, steam, chilled water, and natural gas. As with many other institutions, commercial or residential customers, Stanford would often notice problems only when a bill showed a dramatic increase in the use for a particular building.

As a result, Stanford decided to implement Itron Inc.’s EEM Suite system to get a better handle on when increases or problems with energy use arose. EEM Suite features the ability to handle multiple sources of meter data, data collected at 15-minute intervals (both current and historical), data difficult to access from outside the controls shop, the ability to share utility data with customers, and the ability to make better use of meter data in identifying operational problems. These were all things Stanford was looking to have a system perform.

In 2003, the university purchased two modules, Energy Analyst and Data Analyst, while at the same time hiring Itron (known as Silicon Energy at that time) to develop an exception reporting tool. This implementation unites utility meter data from four sources: two generations of the Energy Management & Control System (EMCS), Supervisory Control and Data Acquisition System (SCADA), and the Utility Metering Database. To accomplish this, four different kinds of gateways are required: three Modbus, one OPC, and one batch.

Data collected now include building-level utility meter data, weather data and a few sewer main flows. Also collected is data on all chilled water and sewer, most electricity and most steam data, available in 15-minute intervals. Domestic water, some electricity and steam data are compiled on a monthly basis. At this time no natural-gas data is gathered. Some 100 buildings, most greater than 15,000 square feet in area, are now in the system. The system looks specifically for a 20% variance in usage to alert users.

E-Mails That Warn
Susan Kulakowski, campus energy manager, has done modeling based on past energy usage, using outside air temperature as a reference to compare a similar weekday with a similar outside temperature in order to determine what the expected use might be. If that number in turn is exceeded by 20%, an e-mail is automatically generated explaining what had happened—for example, that a particular building has exceeded its use in steam and that this is the percentage by which use is over the expected amount.

This e-mail reaches the zone manager responsible for that area, the engineer responsible for that building, and the utilities group in order for them to check the EMCS system for problems. For some two years now these e-mails have been generated. Not every building monitored has this feature built into it. The e-mail is reliant upon having the modeling done first.

Therefore, if an excessive-use alert arrives in a monitor’s e-mail stating a particular building is using 25% more steam than expected for yesterday, those keeping track can go through from different angles to look at the situation. Stanford’s energy management group can look at this and ask why the situation arose, and then the zone manager and zone engineer may look as well.

“Sometimes it’s difficult to determine what caused the alert, while at other times someone knows that there’s a new experiment that just got started up, adding a whole bunch of heat load or other changes in the system for that day,” says Tim Troxell, a zone engineer. “Stanford University has various types of laboratories and server rooms containing high-density computers as well as a lot of other different things that can affect the use in the system.”

From time to time it’s a scheduled event during periods when the building is normally timed to run, such as from 6 a.m. to 6 p.m., but instead the special event is running all week long each day until midnight. This may not be a problem at all, but instead simply a schedule change resulting in more energy use than a normal week. By getting an e-mail out to numerous people involved, they can all read it and reply with the answer for why something has happened.

“The Itron system has provided us with a great tool with which to actually identify problems in a timely manner. We have a utilities group sending out a monthly statement of energy use, but if you receive a statement telling you that 25% more was used this month than last month, that’s a lot of energy used in a month, which could have been fixed after only a few days instead of waiting for the bill to reach you.”

One scenario which could arise is a valve failure in pipes involved with chilled water. Chilled water could be leaking through a unit when it should be closed. In such a case, more energy is being used to chill the water along with more steam energy to warm the air back up again.

“Just a mechanical equipment failure, which does happen, can now be found the next day,” says Troxell. “There are roughly 100 buildings on the Energy Management System, and to go through and look at every single mechanical system in all those buildings would take a huge amount of time. Having the Itron system to go through and give us this report—which says ‘something doesn’t look right; go check on this building for the problem in the steam system, this one in chilled water or this one in electrical energy’—helps us to focus in on the problem in a very timely manner.”

At present SU’s roughly 600 buildings are not all on the Itron system. Some are connected into the university’s energy management system and therefore the metering is immediately available. Others are connected to the high-voltage SCADA system, so that electrical use can be monitored. “On some of them we still send someone out once a month to read the meter,” says Troxell. “In these cases we don’t know what the energy use was there yesterday or the day before.

“Sustainability is a major goal of Stanford University, and we are also involved in a number of different projects to reduce our carbon footprint, doing all the appropriate things we need to do.”

As far as the benefit to Stanford from Itron’s EEM Suite system, Troxell appreciates the timeliness of the notification of changes in the energy usage at the buildings where it’s installed. “Instead of waiting until the end of the month to discover a change in energy usage, it’s now a question of knowing there’s a problem the next day after the change in use,” says Troxell. “And if the problem persists, we get another look the next day after that.”

Troxell once received an excessive-use alert for a materials science building. Those involved immediately had an answer, replying that there had been some extra energy use at the building during the same day. On the next day, the system continued to show above-average energy use. Further investigation continued, and it was found that there had been a problem in that building with water leaking through a closed valve—causing simultaneous heating and cooling.

“There has been somewhat of a challenge in having one group implementing the programming and doing the modeling and then having this broad range of users who receive these alerts,” says Troxell.

“Sometimes we went through and understood exactly why alerts were coming up. In those cases we’d wish we stopped getting them. But getting it all right just takes a bit of fine-tuning. We used historical data to create a model. A curve was created that fit the data. Well, sometimes that curve dipped below zero, and there were times when we took the data when it never got below 45 degrees, and then we got a 35-degree day [and] I received alerts saying that our chilled water went above what was expected.

“But in reality it didn’t go above what was expected, the expected was minus-200 tons. However we can’t have minus-200 tons; the curve went the wrong way. That’s one of the drawbacks of going through a negative as the model doesn’t necessarily fit every scenario. It just takes time getting everything on the same page as far as what’s logical from our perspective and then getting the system to reflect that.”

Internship to Help Track the Details
Matthew Shapiro, a petroleum engineering graduate student at Stanford, started in June, 2007, as an intern with this project. As part of his major he does some computer modeling, so he has background in that. Shapiro's work with Itron involves taking the log data off of the site and then fitting curves to them. Those are subsequently uploaded onto the server to be used as prediction algorithms. This work is done using the Microsoft Excel program. Data he takes off of the system are to form a collection of some 400 points typically representing one year of operation. Then Shapiro uses Excel to find the equation that approximates those points as closely as possible.

It’s usually correlating, for example, chilled water use or the cooling load in a building with the outdoor air temperature. What EEM Suite does once Shapiro uploads that is to take a measurement of the outdoor air temperature and then use the algorithm to predict what the building’s use should be.

“If the building exceeds that by more than 20%, EEM will send an e-mail to me and my supervisor and the building supervisor,” says Shapiro. “This is a simple statement for those involved that this building is using more energy than it was predicted to use under those circumstances.”

Shapiro works mainly with the heating and cooling loads on campus. The main objective of his work with Itron is to set up a system that will be able to monitor buildings in a way that indicates when a building is using an unusually large amount of energy so that they’ll be alerted to such a situation in case there is a problem.

“When an alert gets sent out, it also sends out a chart of the use for that day and for that month,” adds Shapiro. “This way the person can immediately determine if this was over the course of the day that a lot of energy was used, [whether] there one particular hour where things spiked and also [whether] this is a trend for the month. In other words: Has the building been consistently underperforming?”

Shapiro’s main task in Itron is taking the histories of the steam and chilled-water use and turning them into predictions. The information is presented as bar graphs by EEM Suite and as scatter plots by Excel. “There have been a couple of glitches in the system, but they’ve been resolved very quickly by Itron support,” adds Shapiro.

“This has been an ongoing project. There have been some four or five interns before me who have helped to build this database of prediction algorithms. But it is an ongoing process. As buildings change and get retrofitted and built up, our monitoring improves. It’s a continuing process to improve the predictions.

“Many of the pieces that were in the system were entered in 2004 or 2005, and since then the building use patterns have changed. Since Itron has things set up so it only sends an alert when the building overuses energy and not when it under-uses energy, a lot of these buildings that have undergone lighting retrofits and other sorts of retrofits have actually seen a decrease in usage. Now we have some curves that are overpredicting their use. I’ve been going through and updating those to the 2007 numbers.

“I like that there’s central monitoring going on and that there are people studying this information,” adds Shapiro. “It’s really true that if you don’t have a system like this, things really do tend to slip through the cracks. You could wind up having a leaky boiler not getting noticed for years, whereas with this program, for example, we can look back through the backlogs of data and then send an e-mail to someone to ask them about it. Recently I sent an e-mail mentioning that on a specific date last spring they’d cut their usage in half and it stayed in half for quite a period. When usage recently rose I asked them about it. The cause of that drop has not been pinpointed, but it’s probably that they’d turned off some powerful piece of equipment, such as a laser.”

Shapiro does mention some important conditions on what the system is able to accomplish.

There are only certain buildings for which they can implement a prediction algorithm. The buildings for which loads are driven by such measurable factors as the outdoor air temperature (typically offices and labs that have big HVAC loads relative to the rest of their use) can be monitored through the system he uses. There are other buildings though, such as labs whose loads are sporadic and based on what equipment is being operated, whose use cannot be monitored under the current system.

“Of the buildings I've looked at, I’d say half have a pattern of use predictable enough to fit a curve with enough accuracy to give valuable feedback,” adds Shapiro. “Chilled water and steam are monitored through the prediction algorithms. Electricity is monitored very differently, though. Itron takes a moving average of electricity use over the previous similar days—the last 10 weekdays, the last two Saturdays, Sundays, or holidays—and uses that average as the standard. So what the electrical monitoring checks is that there are no sharp, sudden increases in electrical use.”

A Zone Manager’s Viewpoint
George Sandoval, Stanford zone manager, finds the Itron program to be especially user friendly. “I might not use it for a couple of months, but then I can use the data analyst software to see consumption,” says Sandoval. “Very often it helps us troubleshoot a problem.”

Something which has changed lately is that the main EMCS system controlling most of Stanford’s central building HVAC systems can now be accessed at least to see consumption of the main utilities, according to Sandoval. “But with something like chilled water, when the temperature outside goes to zero, you wouldn’t likewise expect chilled water consumption to go to zero as well. In the interior zones of the building you still have loads for lights and equipment.

“In that case you’d receive an e-mail stating that zero or negative chilled water should be used. This isn’t possible, and you’d tend to ignore those e-mails—not something you’d want to do, either. We wanted to rid ourselves of those ‘false alarms,’ as we also don’t want to receive more e-mail than we have to or you get the boy-who-cried-wolf effect.”

Another potential problem is the fact that, since the campus is set up in four different zones, it has been hard to get the e-mails sent only to the person responsible for that building.

Sandoval would get e-mails for the other zones. “The stuff you receive that isn’t even yours can become a problem,” he adds. “I’m hoping we find a way to make such blanket notifications a bit more difficult.

“Also, when we have our HVAC quarterly shutdowns for routine maintenance, the system tells you later, when you are up and running normally, that you are running higher. I think I have a feel now for what are ‘false’ alarms. It simply takes a bit of time remembering that, say, a week ago, we shut down that building for maintenance. A possible improvement might be to make the overview period a bigger one, such as a month instead of seven days as now. Power consumption really shouldn’t change that much over a given period, especially for electricity usage. That seems like an easy thing to implement.

“But the main positive is I am getting e-mails describing possible trouble, and this means I don’t have to go looking for that. This is a big time-saver for me. Itron has also helped with the time schedule programs that buildings sometimes need to go on.”

Higher consumption–exception reports for chilled water ended up showing that air handlers had had their economizer operations manually shut down. The free outside air that could have been used for cooling was not being taken advantage of in the building. “Without that exception report from Itron, it might have been a long time before we caught that,” says Sandoval. “Also, several times we caught a problem in which the hot water control valve was leaking and the chilled water valve downstream was compensating to make up for the leak. So not only were we using more steam because to the leak, but more chilled water as well.”

Sandoval and others are developing a program for smarter self-diagnostics in their EMCS system at the air-handler level. This way, certain “if-then” statements can be developed such that, as an example, if outdoor temperature reaches a certain level then the chilled water valve should not be open.

“EMCS has the instrumentation to do this, but it just doesn’t have the logic,” says Sandoval. “We’re going to go in and install the logic so that an alarm will be generated if for some reason something strange is going on. The variables controlling the air handlers are good at operating the controls but don’t really have self-diagnostics, such as asking, ‘Why are we using chilled water when we have cold air temperatures outside?’ It could be simply a damper or a valve stuck or a temperature sensor gone out of calibration. So rather than catch the problem at the building level through Itron, we’d instead be able to catch it at the air handler level and be very specific.”

Sandoval is especially appreciative of the fact that now he has much greater access to his EMCS system. The EMCS department itself used to be the only one which had access. “We, the maintenance users, didn’t really have access from our desktop,” admits Sandoval. “We would go over to their shop and ask them about things. But with Itron it now means that at least at the building-level utilities metering information is accessible to us. That’s a powerful tool.”

EMCS itself is going through computer upgrades to the Delta V system. This includes a Web-accessibility application. “I can now go in through the Web at a read-only level and see what’s going on at the air handler level, the alarm level, and the heat-exchanger level,” adds Sandoval.

“This is our main campus PLC [programmable logic controller] system. The manufacturer of our EMCS system was Fisher Rosemont [currently it is Emerson, Inc.] We’ve gone through several generations of equipment. With this newest, the Delta V, the Web accessibility is a great improvement. It puts information into the hands of more people and helps with management overall.

“In a nutshell, this is great in that we now get fewer and fewer false alarms, and with the Delta V upgrade we’ll have great access and more flexibility. The nice thing about Itron is that it gave us access to systems we would not have ordinarily had access to. It’s a gateway with a built-in firewall, to use the current computer terminology.

Certain utility data can be mapped into Itron from different systems. Yet still I have no access to tamper with that data. With Itron, the information can be shared but not tampered with. Both these factors, I would think, are things people surely get excited about. Work can get done quicker and more money saved for Stanford University—or any other institution out there—in the long run.”

South Carolina–based Peter Hildebrandt writes about science, nature, and industry.

DE - January/February 2008