The Cahua run-of-river plant suffered excessive abrasion of the Francis runners, which resulted in a loss of efficiency and a short repair cycle. The adaptive strategies employed at the plant focus on optimising its performance without making structural modifications. Optimisation of the intake headworks improved sediment removal from the diverted water, and Tungsten carbide coating was applied to the runners.

The Cahua project is a 43 MW run-of-river (RoR) hydropower plant built in 1967 on the Pativilca River, an Andean watershed which drains toward the Pacific Ocean in the Ancash department of Peru, about 170 km north of Lima. Owned and operated by Statkraft Peru, facilities include a fixed river weir, two radial gates in the river which also operate as sediment sluices adjacent to the intake, two radial gates and desander basins, a 9 km non-pressurised tunnel on a slope of 0.002, penstock forebay and overflow weir at the end of the tunnel, followed by the penstock and powerhouse with two 21.5 MW Francis units operating with 215 m gross head and a total design discharge of 24 m3/s. A project schematic is shown in figure 1 highlighting the two areas of hydraulic problems which were optimised by operational changes: (1) uneven flow splitting, and (2) discharge of excess flow over a weir following the desanders, resulting in hydraulic overload during high flows.

The headworks incorporate eight parallel desanders, each with a vertical gate and flow tranquiliser at the entrance, as shown in figure 2, and a fixed weir at the outlet. Cleanout is accomplished through empty flushing. The velocity in the inlet canal must be high enough to transport the sediment load, and the tranquiliser screens are used to break up this high-velocity inflow into a more uniform flow across the full cross-section of the basin. Three rows of tranquiliser screens are typically recommended, but only two are installed at Cahua.

Hydrology and sediments

The 2,974 km2 watershed of the Pativilca River tributary to the intake has a mean elevation of 3,360 m, and the river gradient averages 2.1 per cent in the vicinity of the project. There is a strong seasonal variation in precipitation and runoff, as shown in figure 3.

The river transports sediment ranging from silt to 0.5 m boulders. There are many active erosional features in this sparsely vegetated watershed, and discharge from the unstable Jelle Ragra Creek (see figure 4) about 25 km upstream of the intake has generated hyper-concentrated flows with suspended solid concentrations of 400,000 ppm. Slope failure of silt soils is the primary mechanism contributing to the extremely high sediment yield from this particular watershed. Suspended solid concentrations at the intake have exceed 25,000 mg/L.

Sediment problems

The primary sedimentation problem at this facility is the excessive abrasion of the Francis runners, resulting in loss of efficiency and a short repair cycle. The sediment load on each turbine averages about 160,000 t/year. Abrasion by coarse bed material (gravel and cobbles) also damages the sill of the radial gates in the river, which are kept partially closed to regulate water level at the intake.

The current turbines are cast from martensitic steel (13 per cent chrome and 4 per cent nickel) with a Mohs hardness in the range of 4.5-5.0, which is significantly softer than quartz, which has a Mohs hardness of 7.0. This makes the runners highly susceptible to abrasion by quartz and other hard minerals. Cahua’s original turbines were designed to operate with a solid concentration not exceeding 3 g/L. However, sediment concentration at the intake regularly exceeds this level, and the original operational cut-off level produced 15 to 30 days of downtime per year, equivalent to an average annual generating loss of 19.2 GWh.

To increase power production, the operating rule was modified to produce power until suspended sediment exceeded 10,000 mg/L, but this resulted in runner damage by sediment abrasion, plus cavitation in the abraded areas, which further accelerated metal erosion. Turbine damage occurred on the suction side of the stay vane at the runner’s inlet, at the stay vane outlet, the outer diameter, and the areas around the labyrinth seal as shown in figure 5. This damage required that personnel from Statkraft Peru refurbish both turbines each year at an annual cost of about USD 580,000, plus about 20 days of annual down time for equipment change-out. Refurbishment included welding, grinding and machining each component.

Sediment management strategies

The unstable soils and zones of earth movement in Jelle Ragra Creek contribute a heavy sediment load, and watershed management was considered as an option. However, this is not the only source of sediment, as there are also many areas of degraded land and active landslides throughout the watershed. High erosion rates were also determined to be directly related to land use and irrigation practices by small farmers. Given the difficulty of tackling this problem, and lacking definitive information on the relative importance of different sediment sources and treatment options across the 2,974 km2 watershed, the decision was made to initially focus on improving sediment handling in the headworks and hardening the turbines, while performing additional analysis of watershed management alternatives.

Headworks modifications

Both the river intake and the desanders were examined for opportunities to improve sediment removal. The sand concentration in the diverted water can be minimised by taking advantage of or creating secondary currents in front of the intake to maximise water withdrawal from the top of the water column, and avoid eddy vortices, which can lift sediment from the river bed and carry it into the intake. However, at Cahua the potential to generate secondary currents in front of the intake was very limited because the flow was considered too shallow and turbulent to develop a vertical sediment concentration gradient in the area of the intake. High concentration events occurred with shallow and highly turbulent flow, as seen in see figure 6A. This indicated very limited potential to reduce the amount of sediment diverted by modifying the river hydraulics at the intake.

When examining the desander’s performance, two areas for improvement were identified by studies conducted between 2012 and 2014. The first problem was related to hydraulic imbalance between the parallel desanders. During the high flows associated with high sediment concentration, a high-velocity jet enters the river intake, producing the irregular flow pattern at the trash rack as illustrated in figure 1 and photographed in figure 6B. Notice that the high-velocity flow is pushed against the right-hand side of the channel. This produced a higher hydraulic loading rate in the Group-1 (right-hand) desanders as compared to Group-2.

The second problem is related to the location of the side weir for discharging excess flow. At RoR intakes discharging over a fixed river weir, the water level in front of the intake will rise during floods, and the flow entering the intake will also increase. This excess inflow is spilled back to the river via a side weir, which may be located either before or after the desander. However, if this side weir is located downstream of the desander, as is the case at Cahua, when the side weir overflows the desander will be hydraulically overloaded: it will receive the design turbine flow plus the flow discharged over the side weir. Because of the side weir overflow, field measurements at Cahua showed that the desanders were regularly operating at about 133 per cent of their design flow. To avoid this hydraulic overload, the side weir should always be located upstream of the desander. At Cahua this problem needed to be addressed by operational changes which minimise side weir overflow.

The problem caused by both the hydraulic imbalance and the hydraulic overloading are most acute during high flows, when the suspended sediment concentration is highest and optimum desander performance is most critical.

Data on suspended solids had been collected over a number of years, making it possible to compare the suspended solid concentration in the river in front of the intake against depth-integrated concentration sampling at the exit of the desanders. These data were processed and displayed in different formats, as shown in figures 7 and 8, where desanding efficiency is computed based on the relative suspended sediment concentrations. Data from different years showed substantial variations in the sediment removal efficiency, suggesting that headworks sediment removal efficiency was being influenced by operational procedures.

In 2014, individual desander efficiency measurements were made for the first time, comparing the suspended solid concentration upstream and downstream of the desanders. These tests indicated that desander trapping efficiency varied from 3 per cent to 45 per cent, with an average of 20 per cent. This bay-to-bay variability reflected the problem of hydraulic imbalance between the parallel units. Of the sediment trapped in the desanders, 75 to 95 per cent was sand sized, whereas the material trapped in the penstock forebay was 82 per cent silt. The maximum diameter of the sediment carried into the tunnel varied from 0.84 mm to 2.0 mm. Mineralogical analysis indicated that the material trapped in the tunnel forebay was 78 per cent quartz.

To correct both the hydraulic imbalance and overload problems at the Cahua desanders it was decided to use the vertical gates at the entrance to each desander to control the flow rates. Streamflow current metering techniques were used to measure flow rates at the entrance to each desander, the vertical gate at each desander was adjusted to achieve the design flow of 3 m3/s, and the corresponding water level was marked on the desander wall near the gates. This mark could then be used by the operator to adjust the gate opening desander as flow conditions change.

When the operators were given improved guidelines for operating the gates to better regulate flows, a 50 per cent improvement in sediment removal efficiency was achieved without any structural modifications to the headworks. These data are shown in figure 9. Balancing flows and eliminating hydraulic overload is the easiest and most effective way to immediately increase the operating efficiency of the desander without making changes to the structure.

As an additional measure to optimise operation, future plans call for installation of real-time instrumentation in the watershed to provide plant operators with advance warning of floods with high sediment loads.

Because the desanders are relatively shallow (2.5 m) they have limited sediment storage capacity. As the desander fills with sediment, the flow velocity increases, which provides greater potential for sand to be carried over the outlet weir. To minimise this problem, during periods of heavy sediment loads the operational personnel were also instructed to ensure the desanders are flushed out before excessive amounts of sand accumulate.

Turbine hard coating

To increase the abrasion-resistance of the runner, a coating may be applied to protect the metal. Both hard and soft coating technologies exist, and a hard coating based on Tungsten carbide was implemented at Cahua. This coating is applied to all wetted surfaces by the continuous combustion HVOF (High Velocity Oxygen Fuel) thermal spray process, in which a fuel gas and oxygen are combusted in a high pressure, high temperature chamber. The resulting hot, high pressure gas is ejected through a small-diameter nozzle and accelerated down a long barrel at supersonic speeds. Tungsten carbide powder (or another coating formulation) is injected into the nozzle, where the powder mixes with gases flowing at velocities approaching 1,000 m/s. When these high velocity particles strike the substrate it bonds instantly to build up a very dense, adhesive and cohesive coating with comparatively very high bond strength, low residual stress, low porosity (typically less than 0.5 per cent), and high wear resistance. Coating thickness is typically between 0.2 and 0.3 mm.

The coating gun can be used manually or robotically, with the robot providing a more controlled application. Coating application is constrained in smaller turbines by the lack of space to manoeuvre the spray gun within the interior confines of the turbine. A bolted turbine design was used at Cahua to overcome this limitation (shown in figure 10). The use of exchangeable vanes allows for fully hard-coated (HVOF) runners, even for small sizes, since the vanes are coated when dissembled. Using a preventive maintenance approach, the runner may be disassembled, sandblasted, and recoated before the original coating is penetrated to the underlying metal. If metal erosion occurs, the runner geometry is reshaped prior to recoating. In either case the impact of coating is to increase overall turbine efficiency and reducing the time between overhaul, as conceptually illustrated in figure 11. Although there is a small initial efficiency penalty due to the higher roughness of the coatings, this is offset by operating at a significantly higher average efficiency, together with the ability to operate at higher sediment concentrations.

At Cahua it was decided to install new coated runners that can operate with sediment concentrations of up to 10 g/L. Replacement of the runners began in March 2009 using a bolted runner coated with tungsten carbide, designed and produced by Norwegian manufacturer DynaVec. The other unit was left with the original runner for comparison. Over a period of around three months (13 March to 9 June) the upgraded unit generated 13.1 GWh-worth of extra power while passing 131,000 tons of sediment with a maximum concentration of 20 g/l. When removed for inspection on 9 June, in most areas the coated runner did not have any visible erosion and coating thickness had only been eroded by a maximum of 30 per cent. However, at several sharp corners on the outer diameter the coating had been eroded and metal erosion was visible. Weak points were also observed on the leading and trailing edges of the runner vanes in the hub and shroud where the insertion slots for the runner vanes created sharp corners. The second runner was subsequently replaced based on results from the first runner, the main improvement being the application of Belzona 1321 - a ceramic-filled epoxy coating - to reinforce critical areas of wear observed in the first runner.

Lessons learned

By focusing on optimising plant operation, plant down time was reduced by seven days, to 15 days per year, increasing revenue generation, while simultaneously increasing the time between runner overhauls. Approximately a 50 per cent reduction in sediment load was achieved at essentially zero cost by the improved operation of the existing equipment. The use of new bolted runners with Tungsten carbide coating had an initial cost of USD 970,000. The overhaul and recoating cost is about USD 50,650 per turbine, and is performed every year. The increase in power production of approximately 15.4 GWh/yr paid for all project costs within the first seven years.

Graphs and figures

‍

‍

‍

‍

‍

‍

‍

‍

‍

‍

‍