The Esmeralda dam impounds 758 Mm3 to supply water to the 1,000 MW Chivor hydropower plant in Colombia, commissioned in 1977. After an extreme flood in 2004, the minimum operating level was increased to protect the power plant from sediment abrasion. Thereafter, the Sustainable Sediment Management Plan was developed, including a design for three new intakes at different levels, among other measures.

The Esmeralda reservoir supplies the 1,000 MW AES Chivor hydropower project, located in the Boyacá province of Colombia, about 90 km north-east of Bogotá. The project is owned by AES Chivor & Cia., a subsidiary of AES Corp.

The earthfill dam is 237 m tall and impounds a narrow 18 km reservoir which receives inflow from the Garagoa and Somondoco rivers, with added contributions by diversions via tunnels from the Tunjita, Negro and Rucio rivers. A schematic of the reservoir layout is shown in figure 1. Water is diverted via a 7 km tunnel to the Chivor powerhouse, which discharges to an adjacent tributary, Río Langupá. Power is generated by eight vertical-axis Pelton units operating under 800 m of head and a total design flow of 160 m3/s.

Hydrology and sediment

The mountainous 2,420 km2 catchment produces a mean annual inflow of 2,471 Mm3. The seasonality of inflow as compared to turbine capacity is shown in figure 2.

When filling began in 1975, the 758 Mm3 reservoir had a capacity:inflow ratio (ratio of gross storage capacity to mean annual inflow) of 0.31. Of this total, 90 Mm3 consisted of dead storage. The rate of storage loss has averaged about 3.4 Mm3/yr (about a 0.4 per cent annual loss), which is not unusually high, but after 41 years of operation the reservoir had lost 18 per cent of its total volume and sediment beds were approaching the intake.

The delta deposits are comprised primarily of sand (d50 ~0.2 mm) which is deposited underwater during the wet season, but during drawdown is remobilised and transported deeper into the reservoir. The photograph in figure 3, taken in the upstream portion of the reservoir in Garagoa and Somondoco confluence, shows sand deposits that have been scoured and moved further downstream, exposing the gravel riverbed created by bed load transport.

Because the reservoir is drawn down to similar levels each year, the delta exhibits little vertical growth; most sediment is deposited on the face of the delta, maximising its rate of advance toward the power intakes, as shown in figure 4.

Sediment problems

The project experienced a severe sediment-related incident in 2004 when an out-of-season flood occurred with the reservoir drawn down to a low level. This inflow transported a high concentration of sediment along the length of the nearly-empty reservoir and into the intake, and may have also included sand scoured from the delta.

The power station continued operating throughout this event, as mandated by the grid power dispatcher, and was not taken offline until it became apparent that severe damage was occurring. The turbines, as well as the spherical and needle valves, were severely eroded.

Figure 5 gives an example of the damage that occurred over less than 24 hours of operation to the Pelton needle valve, compared with normal damage after approximately 10,000 hours of operation.

The plant was then taken offline for 25 days while emergency repairs were executed. This demonstrated the need for real-time sediment monitoring at the intake, and for operator flexibility to implement sediment-guided operation, overriding the previously-planned dispatch schedule and powering down the plant during a sediment event. It also underscored the need to develop a sustainable long-term strategy to enable the plant to continue operating despite the growing sedimentation problem.

Monitoring

Responding to the damage in 2004, a submersible pumping system was installed to sample water at the intake, and a basic sediment laboratory was installed at the dam for real-time sample analysis. Samples are pumped whenever significant inflow events occur while the reservoir is at a low level. The minimum operational level was raised from the 1,190 m design level to 1,205 m. In addition, dispatch arrangements were made to enable the plant to be taken offline to avoid future sediment damage.

Reservoir bathymetry was originally performed at intervals of three to four years using cross-section surveys. However, with increasing sedimentation concerns, the interval between surveys was shortened to annual measurements, and the methodology was changed to contour survey instead of cross-sections.

The project experienced a severe sediment-related incident in 2004 when an out-of-season flood occurred with the reservoir drawn down to a low level."

In two particular years, 2014 and 2016, both contour and range-line surveys were performed, with a survey difference between them of < 1 per cent in both cases. However, because the annual rate of storage loss averages 0.4 per cent, which is smaller than the margin of error of bathymetric survey measurements, these close-interval surveys showed the reservoir to be losing capacity in some years and increasing capacity in others. It was decided that a better monitoring strategy would need to be implemented to perform a bathymetric survey at two-year intervals, and to use the remaining monitoring budget to perform vibracore sampling, also at two-year intervals. This way, it would be possible to track not only the configuration of the deposits, but also to detect any change in their composition which could make them more abrasive to the turbines.

With sediment beds advancing to the area of the intake, it was desired to document the composition of the sediment in the vicinity of the intake, to determine if sands were being transported beyond the delta and into the area of the power intake. It was also desired to determine the extent of sands delivery by lateral tributaries near the dam. The first sediment sampling was performed by vibracore over a two-day period. Figure 6 shows the sediment sampling in La Esmeralda reservoir using battery-powered portable vibracore equipment. Vibracore was selected as the sampling method because multiple locations could be sampled over a short period of time (about 10 sample locations per day). Furthermore, by examining the samples as they arrived on deck, it was possible to adjust the sampling locations depending on the type of material encountered in each core. Sub-samples were collected from different depths in each core for laboratory analysis. The upstream portion of the reservoir, consisting of delta deposits exposed and scoured during seasonal drawdown, were observed and sampled by hand (see figure 7).  

In 2015, at LISST real-time laser diffraction system was installed at the powerhouse, sampling water at the draft tube. The objective of this monitoring is to better document the changing grain size and concentration of sediment passing through the turbines as sedimentation progressed, and to provide better data for correlation to turbine abrasion rates.

Sediment management strategies

Sustainable Sediment Management Plan

A Sustainable Sediment Management Plan was developed in order to define a set of actions leading to a viable long-term strategy.

Regulatory agencies in Colombia, accustomed to clear-water releases below dams, have not yet embraced the need to begin releasing sediments.

As a result, regulations currently impede some management options, but these may become more viable in future.

It was also recognised that long-term sediment management measures could be expected to become better refined over time as a result of both regulatory changes and technical considerations, as more complete monitoring data come available and the success of initial management measures are evaluated.

As such, the Sediment Management Plan was developed with three objectives:

• Define actions that could resolve the immediate sediment challenges.
• Identify the long-term strategies to achieve sustainable utilisation.
• Refine the monitoring data and studies needed to finalise and implement sustainable long-term project re-design and operation.

The first version of the Sustainable Sediment Management Plan was developed based on the available design and hydrologic data, the sediment sampling data described above, 1-D sediment transport modelling, and preliminary energy modelling to evaluate the sensitivity of power production to diminishing live pool volume.

Sediment transport modelling: The long-term evolution of reservoir sedimentation was simulated for different operational scenarios using the SRH-1D sediment transport software from the U.S. Bureau of Reclamation. Simulations up to 100 years were performed using a one-day time step. Modelling showed that under the current operational regime, the toe of the delta with coarse sand could reach the intake in as little as 20 years. However, by increasing the minimum operating level by 0.5 m/year, coarse sediments would be focused on the top of the delta instead of at the downstream delta face, thereby delaying the arrival of sand at the intakes by 40+ years.

Energy simulations: Energy modelling simulated the effect of reduced reservoir volume on annual energy production. Because inflow is relatively well-distributed throughout the year, the reduction in live storage from 500 Mm3 to 100 Mm3 would represent only a 5 per cent reduction in annual generation, if following an operating rule that maximises energy production when the reservoir is at a high level to maintain a reserve capacity to capture small flood events. However, it would have a greater impact on income because it would shift more power production away from drier months, when energy prices are higher, and into the wetter months when prices are lower. It would also complicate the planning and contracting of power dispatch since the firm power capacity would also be diminished.

The following options were analysed:

• Immediately begin raising the intake level to avoid the entrance of sediment, and continue raising the intake until it converts to run-of-river operation.
• Delay raising the intake level, passing as much fine sediment as possible through the turbines and thereby delaying arrival of the coarse delta sediment to the intake.
• Rehabilitate the river diversion on the right side of the dam and use this to flush or sluice sediment.

The Sediment Management Plan was developed, and then reviewed by a panel of four experts (two international and two local), including both sedimentation and civil design expertise. This panel was convened at the dam site to review the data and sediment management options, and finalise recommendations for sustainable management.

The alternatives examined at this site are briefly outlined below:

• Immediately increase the level of the intake to prevent the entry of fine sediment, and progressively raise the intake as required to stay above the sediment bed.

This has the disadvantage of trapping almost all the fine sediment in the reservoir, which would accelerate the arrival of coarse sediment to the dam as the delta progresses over the previously deposited fine sediment.

• Allow for the ability to increase the intake level in the future, but do not raise the level until the abrasion rate on the turbines requires it.

This alternative would pass as much fine sediment as possible through the turbines, minimising the trapping of fine sediment and thereby preserving reservoir volume for storage of the coarse sediment, which cannot be passed through the turbines.

• Progressively raise the reservoir level to retard arrival of the delta at the intake.

This alternative may be implemented independently of the other alternatives, and will focus coarse sediment deposition further upstream in the reservoir and further away from the intake, but it will require progressive raising of the minimum operating level. The benefit of this operation was already demonstrated by the 1D model simulations, plus prior experience at another reservoir.

• Perform reservoir flushing to achieve a sediment balance and sustain long-term storage capacity. Flushing could be accomplished by rehabilitating the abandoned river diversion on the right abutment.

This alternative has the disadvantage of construction difficulties (since the bypass tunnel entrance is deep underwater and buried in sediment), and there are also regulatory restrictions on flushing that would impede the rapid implementation of this measure. This measure may be considered as a future long-term alternative.

• Continuous dredging to stabilise reservoir capacity and preserve long-term storage capacity.

Dredging would be potentially feasible if it were possible to discharge sediment below the dam. However, under the current regulatory environment this was not considered to have an assured positive outcome in the regulatory process. It might be considered as a complementary measure in the future for the control of coarse sediments as they approach the intake.

These measures are in addition to the real-time monitoring and sediment-guided operation that were already being performed.

Actions taken

Given the need for immediate action to be taken to protect the power plant, and based on sediment sampling which revealed an absence of sand in the fine sediment being deposited near the intake, a combination of alternatives #2 and #3 described above were selected for immediate implementation. The minimum operating level was increased to 1,210 m, and programmed to increase thereafter by 0.5 m/yr until modified by new recommendations from a Sediment Management Plan update.

A design was prepared for the modification of the intake to include three new intakes at progressively higher levels (see figure 8). Only the lowest of the new intakes would be constructed at this time, so that it could be placed into operation immediately as dictated by the rate of sediment advance and turbine abrasion. However, the tunnelling work for all intake levels would be completed, except that the final 12 m of rock excavation would be completed when it became necessary to construct the successive intake level.

Additional sediment management activities are also being implemented to help define the elements which will support the transition to long-term sustainable operation:

• Preparing sediment balances and the initial Sediment Management Plan for immediate action, which will be updated over time.
• Monitoring sediment inflow from the Somondoco and Garagoa rivers.
• Completing a study of turbine abrasion and potential remedial actions, being undertaken through University of the Andes.
• Performing bathymetric studies and vibracore sampling of sediment deposits for granulo-metric analysis, every second year.
• Performing feasibility studies for: sediment contribution from each watershed and methods to reduce sediment yield; sediment release alternatives (flushing, sluicing, dredging) to achieve a sediment balance; analyse alternative operating rules and the value of storage volume, to better define the target long-term storage to be sustained; environmental evaluation of different sediment management alternatives.

Given the growing severity of the sedimentation problem, and with 70 per cent of all electric energy in Colombia being generated by hydropower, AES Chivor has also begun working with other energy producers, universities, engineering associations, and environmental organisations to promote legislation that focuses on achieving a long-term balance between sediment input and discharge in the interest of sustaining the existing renewable hydropower generation infrastructure.

Conclusion

The studies performed at AES Chivor showed that sustainable operation is feasible at a large 1,000 MW storage hydropower facility, which did not incorporate any consideration of long-term sediment management into its original design. The studies also point to the increasing complexity associated with the need to incorporate sediment management into project design and operation.Finally, experience at this site highlights the value of making an early start on the transition to sustainable use, and the importance of developing adequate monitoring data as the basis for identifying and implementing both short and long-term strategies.

Graphs and figures

‍

‍

‍

‍

‍

‍

‍

‍