In today's marketplace how do you know when the cost of generating power becomes more than the market forces are willing to pay? The answers are increasingly complex. There are many factors, including fixed costs such as labor and debt charges, and variable costs including fuel, chemicals and major maintenance charges. When the generation facility is an independent power producer on the open market, additional factors can come into play, including transmission and transportation tariffs, contract-specific operating requirements, spinning reserve, and more. The list is long.
Sophisticated models have been developed to look at the long term interactions between expected market forces (both fuel and electricity prices) and power plant generation costs. However, when looking at the day-ahead and real-time operations, the choices are not as well defined. How does the asset manager decide if the plant should produce power tomorrow? How does the asset manager, or real-time energy trader, decide if the plant should produce more or less power during the next hour?
Real-time, dynamic dispatch tools are needed. The plant must be modeled in a way to give the person making the decisions, i.e. the energy trader, the information needed to make intelligent decisions, fast. The energy trader must know if selling that additional 25 MW will require a second gas turbine to come online at minimum load, which could increase the heat rate of the entire facility by 50% (and may also incur a starts-based maintenance charge), placing the facility in a position to be operating at a loss. The trader also needs to know that if, instead of selling a block of 25 MW, the block is increased to 75 MW, the plant configuration could adjust in such a way (for example switching from duct burners to a second combustion turbine at full load) that the overall heat rate would decrease – making the profits from the 75 MW block sold greater than what would have been realized if 100 MW were sold.
An additional concern is when to operate in full peaking mode, taking the plant off-line after each 16-hour day (or less), and when to leave the plant on-line during off-peak hours in order to offset maintenance charges associated with starts on the combustion turbines. Keeping in mind the maintenance agreements in place for the facility, the dispatch models can be used to determine the off-peak energy pricing needed to balance any maintenance charges associated with starts. If the traders are able to get better energy pricing than indicated from the model, the facility should remain on-line during the off-peak hours.
It is possible that an asset previously in a losing situation could start to reap profits by adjusting the major maintenance schedule of the equipment. In order to do this, a clear understanding of the cost-to-benefit ratio of running during off-peak hours is required. A model which can accurately predict the total cost of running during off-peak hours versus the total cost of re-starting the next day is an invaluable tool in making the determination to leave the plant on-line, even at an apparent loss.
Most new generation assets which have come online during the last few years are based on natural gas fired combustion turbines, either in simple cycle or combined cycle applications. Energy traders that have been brought up working with coal-fired boilers are learning that heat rates are not only a factor of load, but also of ambient conditions, the time of year (summer versus winter), and equipment availability and limitations when plants are configured with such accessories as inlet chillers and duct burners. In order to maximize the profitability of these assets, the people in charge of creating the dispatch schedules need to be aware of all the factors impacting facility heat rates – and be able to act on them quickly.
Models of power generation assets can be built using off-the-shelf computer applications, such as Microsoft’s Excel. This puts the information at the fingertips of those that need it most, in a form they are familiar with and can learn quickly. Performance correction curves, such as those used to model the plant during facility acceptance testing, can form the basis of expected performance (including maximum output and heat rates at various loads) for the daily forecast site conditions.
When plants have significant operating time, these expected performance curves can be compared to actual plant operating data and adjusted as needed for changes in design and equipment degradation. Curves which are updated once each year with the previous year’s operating data should be able to predict actual plant heat rates within 1-2% for the entire range of acceptable operating loads. The use of these real-time, dynamic models can increase the profits of a single 500 MW natural-gas fired facility by over $1,000,000 in a year where market fluctuations have been anything but stable.
Asset managers have been using sophisticated tools to determine the long-term dispatch assumptions of their assets for years. In today’s real-time markets, speed is sometimes more important than sophistication when maximizing profits for each asset as an individual within the fleet.
