Close 

A Computer Model for Evaluating Effects of Water Use Alternatives

To evaluate the effects of proposed alternatives for allocating water under competing demands, BC Hydro developed a hydro operations simulation model. The Canadian utility used the model to work with stakeholders during development of water use plans for its 32 hydro facilities.

By Paul Vassilev, Goran Sreckovic, and Kathy L. Groves

In the late 1990s, the provincial government in British Columbia initiated a water use planning program to ensure that hydro facilities are operated in a way that reflects present-day environmental, social, and economic values. The program featured reviews of all existing water licenses. These reviews emphasized participation by multiple stakeholders, representing a broad range of interests and cultural perspectives.

In preparing to review the licenses for its 32 hydro facilities, water planners at BC Hydro, the provincial electric utility, recognized the importance of developing a transparent, balanced, and consistent methodology for integrating stakeholder priorities. The solution was to develop a hydro operations simulation model capable of accommodating complex operating requirements while remaining accessible and comprehensible to a diverse group of stakeholders. The new model proved to be highly successful, permitting the evaluation of multiple alternatives for water use while enhancing communication and confidence among the stakeholder groups.

Water use planning at BC Hydro projects

The water use planning process was designed to provide opportunities for the public, First Nations, and other interested parties to identify issues and concerns with the operation of BCHydro facilities. The desired outcome for each project was a plan – and ultimately the issuance of provincial licenses and orders – that would clarify BC Hydro’s exercise of its water rights in the operation of its hydroelectric projects.

The process began with the appointment of a consultative committee representing key interests in the watershed, including government agencies, municipalities, and community members. The committee identified values and objectives. In the part of the process supported by the hydro operations model, the committee defined performance measures and related operating constraints, and then constructed operating scenarios for model analysis. Based on the model output, the consultative committee carried out a resource valuation exercise, trading off various objectives. The committee continued to reiterate the process to converge toward a water use plan.

With assistance from technical experts, stakeholders defined performance measures for each objective. A recreation performance measure, for example, could be the number of days a reservoir is within a desired range of elevations in a given season.

Choosing the right software

Allocation of water to satisfy competing stakeholder demands is an optimization problem, so an appropriate computer model to support the water use planning process would be similar to many models used for planning and optimizing hydro operation. However, the multidisciplinary nature of the water use planning process necessitated additional features. Specifically, the model inputs and outputs had to be credible, transparent, and accessible to stakeholders without technical expertise. It was essential that they be used to facilitate discussion, not be the center of discussion.

BC Hydro assembled a project team consisting of technical experts from various fields including engineering, fisheries, operations planning, and recreation planning. The team considered three software sources for the hydro operations model. These included a commercial or public-domain model, an existing in-house operations model, or a new model customized for the water use planning program. Given the shortcomings of commercial and in-house models, the utility decided to develop a new software package with the participation of independent reviewers and technically knowledgeable external stakeholders.

Developing the model

Model development began with a delineation of the modeling framework components (see Figure 1 on page 70). A graphic representation of the framework helped stakeholders understand characteristics of a reservoir system and visualize the effects that could be achieved by adjusting operating constraints.


Figure 1: This schematic representation of the modeling process explains how the model uses three types of input – streamflows, value of electricity, and constraints on project operation – to optimize the annual value of hydroelectricity and evaluate other consequences of the operating scenario, such as reservoir levels and outflows.
Click here to enlarge image

BC Hydro’s operations planning engineers recommended a linear programming approach. The application of linear programming to reservoir operation has been described in several water resources publications.1,2 The software was written in AMPL, a programming language that is often used to code complex optimization problems, and used the ILOG Version 6.6 CPLEX solver for solving the equations generated for each set of constraints. To facilitate the input of new alternatives and analysis of the results, a user- friendly graphical user interface was developed in Visual Basic.

To assess the feasibility of proposed operation scenarios, the optimal reservoir operation had to be determined under a set of constraints related to elevation and flow regimes. Components include the following:

Assembling the fixed input data

For each proposed water use plan, the team configured a project-specific hydro operations model. Certain inputs, including inflow data, value of electricity, and physical project constraints, were fixed for a given project over all operation scenarios.

BCHydro used historical daily inflow records from its operational data archive as the model’s inflow sequences. Typically, these records covered 40 to 50 years of inflows. The natural inflow data were quality-controlled based on comparisons to nearby hydrologically similar gaged basins. The use of historic daily inflows helped stakeholders relate the modeled events to known events and compare operating scenarios in a specific, familiar context.

Hard constraints included:

To accurately incorporate the time-varying value of energy into the objective function, the software was allowed to optimize the value of generation associated with a particular scenario, rather than the total megawatt-hours (MWh) generated. The appropriate generation value was a sensitive issue for stakeholders, because BC Hydro’s forecasts for the future market value of electricity were not easy for stakeholders to review or understand. To build stakeholder confidence in this component of the program input, a standardized methodology for determining the value of energy was reviewed and accepted by an interagency management committee at the start of the water use planning process. The methodology used the forecasted British Columbia/United States border monthly average heavy load hour and light load hour prices, along with incremental transmission cost estimates for the BCHydro electric system. The daily time series value of energy was determined at the start of each water use planning process and was not changed through the duration of the process.

Model calibration and quality assurance

The project team emphasized on establishing standardized procedures for developing, quality controlling, and clearly documenting the various modeling components. Standardization reduced pressure on the core team by allowing other BCHydro engineers to participate in developing input and performing analyses. Since the standardized process applied across all of the water use planning projects, key stakeholders who participated in multiple planning processes became familiar with BC Hydro’s approach and were supportive of it in the consultative committees.

After the model was configured by the original programmer, it was checked and reviewed by others and the final configuration was documented in a standardized “model configuration memo” that allowed the model to be easily reviewed by stakeholders and external reviewers. The configuration memos include graphs showing the storage-elevation and discharge-elevation rating curves, along with the specific data points entered into the model to represent these curves, as shown in Figure 2 on page 72.


Figure 2: The “model configuration memo” included graphics such as this one, documenting how input data points related to operating curves and other project characteristics.
Click here to enlarge image

The hydro operations model allows the user to choose the degree of foresight, up to 365-days. One of the challenges in model calibration was to select a length of foresight that realistically represented the operators’ ability to forecast future inflows. The appropriate model foresight was generally determined by modeling historic operating scenarios and modifying the foresight until it best represented key unplanned events, such as historic spills or excursions below desired reservoir operating levels. Once the foresight period was calibrated, it was held constant for all scenarios.

As a final check on the data inputs and model configuration, the programming team tested the model’s ability to replicate all the characteristics of historical project operation. This final step was also documented in the model configuration memos by including graphs and tables comparing the modeled results to the recorded historic operating data (see Figure 3).

Balancing competing priorities

Once the initial configuration and calibration were completed, only the “soft” constraints, which represent desirable conditions but not inviolable ones, remained to be entered. Describing soft constraints by means of penalty functions seemed to offer the best solution to the allocation of water under competing demands. Penalty functions quantify user-defined preferences for different reservoir elevation ranges and flow release ranges. The model optimizes by deriving a release regime which minimizes the net penalties in the system. Some soft constraints are met by compromising others. Thus, stakeholders needed to decide how to prioritize weightings for competing constraints.

Figure 4 on page 74 shows an example of a release penalty function. The most preferred flow releases, between 10 cubic meters per second (cms) and 15 cms, have a penalty of zero. Flows between 5 cms and 10 cms, and flows between 15 cms and 20 cms, are assigned a moderate penalty function. The highest penalties correspond to flows less than 5 cms or more than 25 cms. Penalty functions are date-dependent and can vary throughout the year.

Engaging stakeholders in the process

BC Hydro carried out studies for more than 20 generating stations and more than 30 dams. All had multiple objectives and competing demands. Each consultative committee generated operating alternatives, the number of which was proportional to basin complexity. More than 50 alternatives were analyzed for the Bridge River system, which is comprised of three storage reservoirs, four generating stations, and a diversion to an independent power producer. The Bridge system also includes two significant salmon runs, heritage issues around the reservoirs, flooding concerns, minimum flows for salmon rearing and spawning, and releases to remedy dam construction footprint issues.


Figure 3: The hydrographs generated during model calibration were included in the configuration memo to document the model’s performance for historic operating conditions. This chart is for the Cheakamus project during 1991.
Click here to enlarge image

Project operating scenarios were proposed by stakeholders during open meetings of the committees. At the start, stakeholders tended to propose scenarios that reflected single-interest objectives – for example, maximizing power generation or minimizing system regulation to restore the natural hydrograph. Later, stakeholders began to offer scenarios that incorporated tradeoffs in order to maximize the collective performance measures valued by the various interest groups.

Throughout the process, BC Hydro provided guidance to stakeholders on how to specify operating scenarios in terms that could be objectively modeled by the technical team. For example, a desired water level for fish spawning must be specified as a discharge; and a discharge at a given location needs to be clearly defined as either a hydro release or the combination of hydro releases and tributary inflows. It was also important to specify the desired degree of compliance for each constraint. For example, extreme changes to project operation might be needed to meet a target reservoir level by a specific date in every year. However, the constraint might be satisfied with moderate changes to operation if compliance is defined more flexibly – for example, by accepting a week or two’s delay in reaching the target reservoir level in wet or dry years. Other constraints, such as the requirement to keep fish eggs wetted once spawned, have zero tolerance for violation.

Learning through simulation

An interesting application of the model occurred with the flooding performance measure. In BC Hydro’s coastal systems, flooding is a significant concern. At the start of many water use planning sessions, flood control was a prominent objective for local resident stakeholders.


Figure 4: Model users input penalty functions, such as this one, describing the relative desirability of certain flow or stage ranges throughout the year. The model was designed to minimize the net penalty points while optimizing the value of power.
Click here to enlarge image

As the stakeholders worked their way through the process, however, they gained a more realistic understanding of each system’s control capability. At the 150-MW Cheakamus project on the Cheakamus River, many stakeholders saw flood mitigation as a key issue and initially assumed that project operation could be modified to eliminate flooding. After running scenarios where flood mitigation was the sole focus, the participants came to see that even very large winter drawdowns could not provide control for some of the major historic inflow events. Stakeholders recognized that complete mitigation could not be achieved, even through implementation of operation procedures that would have significant effects on the interests of other stakeholders. However, mitigation of all but the largest flood events in the record could be achieved by lesser drawdowns that were more favorable to other stakeholder interests. These smaller drawdowns were included in the final operating scenario.

The simulation procedure also revealed which performance measures were sensitive to other constraints, and which were not. At the 24-MW Puntledge project, the consultative committee defined a performance measure as “the number of days when the average daily flow recorded at 5th Street Bridge in the city of Courtney is greater than or equal to 350 cms.” Simulations showed relatively small variability in this performance measure between alternatives. The consultative committee was then able to focus on operational tradeoffs related to other objectives, including fish, recreation, and power generation objectives.

The 55-MW Buntzen Lake 1 project has as its forebay Buntzen Lake, which is heavily used for recreation. The main reservoir, Coquitlam Lake, is closed to the public because it serves as a water supply for the Vancouver area. This reservoir also provides minimum releases in the Coquitlam River for the benefit of fish. The consultative committee created a series of alternatives dubbed the STP series (for Share the Pain) requiring collective sacrifices by fisheries, water supply, and hydro generation in dry years. The selected alternative, STP6, allows river flow, hydro generation, and water supply reserves to be reduced together as needed during dry years, down to a minimum lower bound.

Keys to a successful process

BC Hydro views the water use planning operations model as a major success. Key elements in the development and application of the model in a stakeholder-driven process include the following:

The process of developing model inputs, applying the model, and interpreting the outputs resulted in acceptance of the model both within and outside of BC Hydro. Key stakeholders who became familiar with the model also became proponents of its use among the consultative committees. Within BC Hydro, the model has become a valuable operations planning tool for long term use.

Notes

  1. Labadie, J.W., Optimal Operation of Multireservoir Systems: State-of the Art Review, Journal of Water Resources Planning and Management, March/April 2004.
  2. Yeh, William W.G., Reservoir Management and Operations Models: A State-of-the-Art Review, Water Resources Research, Volume 21, No. 12, 1985, pages 1797-1818.

Reference

Province of British Columbia, Water Use Plan Guidelines, ISNB 0-7726-3731-8, Victoria, British Columbia, 1998.

The authors may be contacted at BC Hydro, 6911 Southpoint Drive, Burnaby, British Columbia V3N 4X8 Canada; (604) 528-1600; e-mail: paul.vassilev@ bchydro.com; goran.sreckovic@bchydro. com; and kathy.groves@bchydro.com.

Paul Vassilev, P.Eng., a power planning specialist at BC Hydro, was task manager for the hydro operations modeling and developed the modeling framework described in this article. Goran Sreckovic, P.Eng., PhD, is a resource planning specialist for BC Hydro and was responsible for mathematical formulation of the optimization model, running the operational scenarios, and interpreting model results. Kathy Groves, P.Eng., M.Eng., team leader in BC Hydro’s hydrotechnical engineering group, had responsibility for quality control and model implementation.


The authors may be contacted at BC Hydro, 6911 Southpoint Drive, Burnaby, British Columbia V3N 4X8 Canada; (604) 528-1600; e-mail: paul.vassilev@ bchydro.com; goran.sreckovic@bchydro. com; and kathy.groves@bchydro.com.


To access this Article, go to:
http://www.hydroworld.com/content/hydro/en/articles/hr/print/volume-27/issue-3/technical-articles/a-computer-model-for-evaluating-effects-of-water-use-alternatives.html