With a vast array of competing demands for water resources and in the face of a competitive energy market, accurate short and longer-term operational weather and streamflow forecasting can significantly benefit hydropower generation.
Amy L. Sansone, 3TIER Environmental Forecast Group, WA, USA
Compared with other energy sources hydropower is perhaps unique in that the resource, water, is subject to a wide variety of competing demands including irrigation, water supply, recreation, in-stream flow requirements for fish, and flood control. To effectively operate a hydroelectric project in real-time, resource constraints – such as maximum water levels for flood control storage – must be incorporated along with knowledge of the current state of the watershed, and a prediction of the upcoming load and predicted availability. The current state of the resource is determined through a system of gauges, sensors and operator observations. And, while the resource constraints often vary seasonally they are usually known or predicted ahead of time through long-range planning and government regulations. However, predicting the upcoming load and availability is complex and uncertain and relies on real-time forecasts of weather and streamflow.
Constraints placed on the water resource by competing demands, coupled with the opening of electricity markets, have significantly increased the complexity of hydropower plant operations. While in the past plant operators simply had to operate in order to minimize cost – possibly taking into account some flood control restrictions – nowadays, open electricity markets and detailed regulatory constraints place considerably greater demands on operators. Simply operating to minimize cost with very limited or even no knowledge of future conditions could result in lost revenue and violations of resource constraints. Consequently, knowledge of future conditions in the form of real-time weather and streamflow forecasts has become an important component of modern hydroelectric operations.
Plant operators, power marketers and dispatchers rely on real-time weather and streamflow forecasts to help minimize cost, maximize profit and meet customer demand while still meeting all resource constraints. They also rely on real-time weather forecasts to help predict customer demand and potential outages caused by inclement weather. There are several different sources of real-time weather and streamflow forecasts, including in-house forecasts, government sources and private providers.
Accurate weather forecasts are the most important component of providing good streamflow forecasts. 3TIER, which was established by researchers from the University of Washington as a means of bringing the latest forecasting technologies from a research to an operational setting, generates weather forecasts with a high-resolution numerical weather prediction (NWP) model. Such models are especially important in areas with complex terrain, as they are able to more accurately capture precipitation amounts, patterns and gradients.
The in-house hydro forecasting system is completely automated and is comprised of the high-resolution numerical weather prediction model, together with a fully-distributed hydrology model and an observation assimilation system.
For hydrologic modelling, 3TIER uses the Distributed Hydrology Soil Vegetation Model (DHSVM), which was developed at the University of Washington and has been applied to watersheds worldwide for forecasting and research applications. DHSVM accounts for the spatial variability of meteorological parameters, land surface characteristics and watershed processes, which is particularly useful in watersheds with complex terrain and weather patterns. Layers of topography, vegetation, soil type, climatology and the stream network are generated from Geographical Information System (GIS) data and streamflow forecasts can be generated at any point in the stream network. The watershed conditions are updated and streamflow forecasts generated as real-time conditions change. However, even with a well-designed observational network and forecast system, differences will often exist between the simulated streamflow based on meteorological observations and the actual measured streamflow. Therefore it is necessary that streamflow observations are used to further correct the raw streamflow forecasts.
Other forecasting time scales
In addition to real-time forecasts covering periods from hours to days ahead, long-range forecasts months to years ahead and climate change predictions for future decades are also used. Long-range forecasts of available streamflow are necessary for project planning to meet in-stream flow requirements and to meet power demands during low flow periods and high peak demand periods. Knowledge of the expected power surplus or deficit months in advance can have direct economic benefits to a hydroelectric project. Given that climate change can significantly affect the timing and availability of water to a hydroelectric project depending on the location, understanding the impacts ahead of time and proactively developing mitigation strategies is a useful exercise for project planners.
In the modern electricity market and with competing demands for limited water resources, the objectives of optimizing hydroelectric plant operations are not just to minimize costs while maximizing plant efficiency. The objectives also include meeting water resource constraints in the system, meeting customer demand and maximizing profit. To meet all of these objectives on an hourly and daily basis, accurate real-time weather and streamflow forecasts are required.
Case study: Seattle City Light
Seattle City Light (SCL), one of the largest municipal utilities in Washington State, provides power to some 723 484 consumers, with approximately 89 per cent of that power generated from hydro resources. Along with meeting power demands, SCL must also meet flood control storage requirements, in-stream flow and ramping requirements for fish and recreational flow requirements. SCL uses real-time streamflow forecasts operationally in both their System Control Center and Power Marketing Group.
Figure 1: Location map of the Skagit River Hydroelectric Project.
3TIER has been providing real-time and long-range streamflow forecasting services to SCL for their Skagit River hydroelectric project since 2002. The Skagit River project is located in the North Cascades Region of Washington. It consists of three dams with powerhouses at Ross, Diablo and Gorge. Ross Reservoir is the most upstream reservoir and provides the majority of the storage capacity, with a capacity of 1298 million cubic meters as compared with 34 million cubic meters at Diablo and 6.6 million cubic meters at Gorge. The Ross Powerhouse provides 460 MW, or 24 per cent, of SCL’s total generating capacity of 1889 MW while the Diablo and Gorge powerhouses have capacities of 168 MW and 177 MW, respectively. Figure 1, above, shows the location of the Skagit River project and contributing watershed area. The watershed boundary is shown in black with the major tributaries shown in blue. The location of the dams and powerhouses are shown as the orange triangle.
The System Control Center is comprised of dispatchers responsible for meeting the competing resource requirements and remotely managing the generation and distribution of power from the Skagit project. The Power Marketing Group is comprised of long-range planners and real-time energy traders with several responsibilities including, trading power, providing long-range forecasts and target flows and providing the dispatchers with weekly, daily and hourly forecasts of load and targets for generation.
Figure 2: Weather forecast at Diablo Dam including precipitation, air temperature and wind chill forecasts.
The Power Marketing Group relies on real-time weather and streamflow forecasts to generate load predictions and targets for generation that are then supplied to the System Control Center dispatchers. The dispatcher uses this information, along with the resource constraints, as a basis for determining the generation schedule for the next 24 to 48 hours. The resource constraints consist of fish flow requirements and associated ramping constraints, flood control requirements and recreation requirements (for Ross Reservoir only). In addition to receiving predictions and targets from the Power Marketing Group, the dispatchers use real-time weather forecasts throughout the year to predict load based on air temperatures and ambient pressure and also expected power outages based on high winds. Dispatchers also use real-time streamflow forecasts to manage the reservoirs during the flood control season and during high flow events throughout the year. Figure 2 (above) shows the forecasted precipitation, air temperature and wind chill at Diablo Dam for a seven-day period in October 2006 while Figure 3 (below) shows the observed natural inflow and the forecasted inflow into Ross Reservoir for the same seven day period.
Figure 3: Streamflow forecast of Ross Reservoir inflow showing observed natural inflow as compared to forecast inflow.
Wing Cheng, senior mechanical engineer with SCL’s System Control Center, real-time says such forecasts are valuable not only in ensuring that resource constraints are not violated but in ensuring that opportunities to generate revenue are not overlooked.