Automated Water System Control

Built largely on land reclaimed from the sea, Netherlands has always had a close relationship with water. It has been an innovator in land reclamation, water management, and drinking-water purification. As one of the country’s leading drinking-water companies, PWN has played a role in a number of these innovations, currently delivering water to more than 1.

By Staff October 1, 2006


Maintaining constant production levels

Ensuring water quality

Meeting variable demands

Minimizing adjustments

Automating for savings, efficiency

Built largely on land reclaimed from the sea, Netherlands has always had a close relationship with water. It has been an innovator in land reclamation, water management, and drinking-water purification. As one of the country’s leading drinking-water companies, PWN has played a role in a number of these innovations, currently delivering water to more than 1.5 million inhabitants who consume approximately 100 billion liters of drinking water each year.

Treatment plants operate most efficiently and deliver the highest quality product when they maintain constant output levels, but PWN faces water demand that varies greatly during the day. In the past, operators at the company’s three control rooms had to adjust production levels 24/7, often hourly, to match output to this fluctuating demand.

Operators had to separate drinking water production from demand to make efficient use of the company’s clean water reservoirs and keep production levels constant. Theoretically, these actions reduce required production capacity. They also require operators to make highly accurate estimates of drinking water demand. Even small estimation errors often led to drastic adjustments in production rates.

Other factors complicating the process included varying costs of water sources caused by lease contracts and energy prices. In addition, strict quality requirements led to increasingly complex treatment processes that were sensitive to production output level changes.

Recently, PWN implemented a rule engine application that forecasts drinking water demand based on factors such as district, day of the week, and weather. The rule engine calculates production levels at each company plant to meet demand while minimizing production cost. This new system makes it possible to reduce the number of production rate changes to once a day.

PWN has also implemented a new Siemens automated control system (PCS-7) that allows the water company to control all plants from one site-independent location, maximizing operational flexibility and making it possible for the rule engine application to run each plant without manual intervention.

The new approach is so much more efficient that it has increased the demand level that existing plants can meet by more than 10%. That means PWN will be able to postpone its expected $30 million to $50 million investment in new production capacity for 10 to 20 years.

Clean water, low cost

PWN operates several hundred wells, 25 water treatment and distribution plants, and a large distribution network. Part of the company’s output consists of pumping surface water into sand dunes for storage, then pumping it back out through nearby wells. Part of the drinking water is produced from surface water by complex and highly sensitive processes such as membrane filtration or UV/H2O2 treatment.

Fewer adjustments

The new process, a response to increasing pollution levels and stricter legislation, requires more complex control strategies. A decade ago, operators worked at the company’s eight control rooms around the clock. Since then, PWN has invested in process automation equipment to reduce the number of control rooms to three.

Before, operators had only a general idea of future demand so they continuously adjusted production levels in the plants to maintain reservoir levels they were sure would avoid shortfalls. Frequent adjustments were required to maintain appropriate reservoir levels. On the average, more than 10 adjustments were made in each plant in a typical 24-hr period.

Before, operators had a general idea of the cost of different water sources but had no time to make the detailed calculations necessary to optimize costs. Moreover, sudden adjustments to flow levels through water treatment processes can affect drinking water quality.

Because of these difficulties, PWN management adopted a new process control philosophy: establish centralized, location-independent control and unstaffed control rooms.

“Because we were breaking new ground from an organizational and technological standpoint, it was important to find a consultant with the experience and expertise needed to guide us,” said Antoine Freijters, manager for operations and water technology for PWN. The company selected Vertis BV because of its experience developing control strategies and because of its strong communications skills, which helped to sell the project at all organizational levels.

“A main challenge was the tension between the increasing span of control of the operators and their increasing distance from the physical processes versus the need for short reaction times in case the process control system requires manual interference,” said Peer Kamp, PWN’s innovation manager. “One question was how to design control systems in such a way that process operators, relying on an automatic pilot 99% of the time, can step in effectively when abnormal situations occur. With the help of Vertis, we developed a new process-automation philosophy aimed at creating a uniform and crystal-clear human-machine interface format. We discussed ergonomic guidelines, developed a new alarm philosophy, set up a layered structure for process automation, and defined a set of software requirements.”

Developing an expert system

The existing process control system was replaced by a Siemens PCS 7 control system. The new system makes it possible to control all company facilities from any location, reducing the number of operators required to control the company’s water production network. Next, an expert system was developed to capture and deploy human expertise using models, rules, and procedures to forecast the next day’s demand for drinking water and determine production levels that meet that demand, typically with only one setpoint change per day and one additional adjustment, if needed.

“Our first step was to interview PWN managers, technical staff, and operators to obtain a good understanding of the issues involved in controlling their water production facilities,” said Maarten Wetterauw, project manager for Vertis. “Our goal was to document all of the issues involved and convert them to mathematical algorithms that would automatically control the operation of each plant.”

As they documented the supervisory control logic, Vertis engineers had to select a rules engine and platform for expert system development. “The supervisory logic could have been programmed from scratch in C, but then it could only have been maintained by the people who programmed it. Over time, even they would have likely forgotten how it was built,” said Wetterauw. “Together with PWN, we selected the G2 real-time rules engine from Gensym as the basis for expert system development because it makes it very easy to develop and maintain knowledge rules.

“G2 provides process engineers with an environment they can understand,” he went on, “so they can easily develop and modify their own rules. G2 enables development of the user interface at the same time as the algorithms and provides an offline environment where different control strategies and what-if scenarios can be evaluated. PWN also liked G2’s proven track record in closed-loop expert systems.”

Vertis accomplished project goals by developing three modules called Plenty Control. The modules were developed in close cooperation with PWN engineers and operators. G2 programming was done by Illyan, located in Amsterdam.

The first module uses trends over the last month, historical data for the day of the week, weather forecasts, district, and other information to forecast demand over a 24-hr period. The second calculates production rates at the company’s five sites based on assumptions of available capacity to meet this demand while minimizing costs and allowing for only one production rate change per day. The third maintains a database of actual capacity—for example, by tracking equipment down for maintenance—and checks the output of the second module to ensure that the production plan does not exceed actual capacity. The modules provide closed-loop supervisory control of the Siemens control system by using the OPC protocol and accessing historical information in an AspenTech IP21 database using the ODBC standard.

Henk van Duist, senior engineer at PWN said: “I am used to programming in Visual Basic and [Microsoft] Excel, and so built preliminary versions of the forecasting module and production module in these environments. I was very pleased to recognize the flexibility of G2’s user interface, which is very important in a dynamic environment. Although the usual struggles occurred between designers, developers, programmers, and key users, we managed to deliver and implement a very successful application. Now, Vertis and I are assisting a nearby drinking water company (DZH) with preparations for an implementation of Plenty Control at their sites.”

Ton Karels, operator and key user of the rule engine, involved in the development from the beginning, said that early on “only a few of my colleagues believed it would be possible to steer our sensitive and complex water treatment processes by an automatic pilot. However, since the full introduction of Plenty Control in May 2005, it is clear that the system works. Now that the drinking water process runs automatically, we can concentrate on optimizing the process.”

The system runs twice a day, shortly before 7 a.m. and 10 p.m. At those times, production rates are set and maintained until the next time the module runs. Typically, the rates need adjustment only once a day, at 7 a.m. About three or four times a year, the control system triggers an alarm when there’s a situation the rule engine has not prepared it to handle.

“By evening out the production peaks that were a characteristic of the previous manual methods, the expert system enables the current facilities to meet a level of demand that is 10% higher than was possible in the past,” Freijters said. “Providing a similar capacity increase by using the traditional method of building new facilities would cost $30 to $50 million. Taking interest and maintenance costs into account, we thus save over $5 million per year.”

The combination of the application and the new control system has enabled PWN to reduce the cost of staffing control rooms. PWN wen from staffing 24/7 at three control rooms to two control rooms running 8 a.m. to 5 p.m. Monday through Friday. The goal, says Freijters, is to staff two control rooms 8 a.m. to 10 a.m. Monday through Friday.

“This 95% reduction saves us roughly $1.8 million per year,” he says. “As we learn to use the system more effectively, we should be able to make more optimal decisions about sourcing production across our treatment facilities based on costs, which should yield additional cost savings.”