Resumen del autor, extraído del trabajo. Large benefits to refiners - on the order of 3-5% of the total energy bill - can be realized through optimization of the utility plant operation, improving the economic performance of these cost centers. As new limits on CO and NOx emissions go into effect in the US, major manufacturing companies are looking for ways to meet those constraints at the lowest possible cost. While capital projects to install more low-NOx equipment are always an option, many companies are looking for better ways to operate their existing utility plants within the EPA guidelines. Even when the equipment is available to operate well below the limits, operators are often faced with many economic decisions that must be made on a day-to-day basis, often without the proper tools to evaluate the tradeoffs. This is especially true as the utility market becomes more deregulated and electricity prices start changing on an hourly or minute basis. And the option to buy or sell NOx credits further complicates the picture. On-line, real-time optimization systems have been in use for many years in the process side of refining and petrochemical plants, but are just now finding their way into the utilities plants. These systems use engineering models of the process equipment, continuously "tuned" to the actual plant measurements, combined with a sophisticated optimizer to calculate optimal targets for the current on-line equipment. In addition, the system can recommend future operating policies for the operators and bidding strategies for the marketers. For example, what equipment should be put on/off line, and when, over the prediction horizon? How much profit is being lost from equipment running at less than peak performance and at what point is it justified to shut down and wash it? During the peak demand of the day, are we expecting to violate the maximum NOx limit, and if so, when should the boiler fuel be switched to gas? What has been our average total NOx emissions for the plant during the last 30 days and how can we best take advantage of available capacity? These and other questions can be answered with an on-line optimization application, ensuring that the utility plants are consistently operated at their optimum points. This paper describes an on-line utilities optimization system that has been installed at several major manufacturing facilities in the US and Europe. One of the plants on the US Gulf Coast installed it primarily to optimize their utility plants within EPA-mandated emission limits. For this facility, the system covers four separate cogeneration plants on the 5000 acre site, each with multiple gas turbines, steam turbines and boilers supplying power and 4 levels of steam to over 70 process units. The site has a total generating capacity of more than 1300MW and is subject to emissions limits that are common to many industrial facilities, covering individual equipment as well as a total site cap. This paper will address issues such as instrumentation, input validation, data reconciliation, model updating, optimization algorithms and plant constraints specific to the utilities optimization problem.