Recovery of Gas from Lake Kivu – The Goats of Rwanda

Months before the story really starts I received a phone call from Philip. “Stanley,” he said, “I need somebody to do some Process Engineering for me”.

Philip and I know each other from the years when we both worked for the same large multinational engineering company. We both left that company at more or less the same time. Since then Philip has been freelancing and looking at situations where an innovative approach could lead to opportunities. In the meantime I have become part of a small company, Added Value Engineering Consultants (AVEC), specializing in conceptual process engineering, so naturally, I was interested in what Philip had to say.

Lake Kivu, Mount Nyiragongo in the background

There is a lake in Rwanda, Lake Kivu, where scientists have established that at depths of around 300m the water is partially saturated with methane and carbon dioxide. The scientists have also done some experimental work and published some data as to how much of the gas will be liberated at various pressures. There is an existing small operating plant which does bring gas-laden water to the surface and separates out the methane to use for running boilers at a nearby brewery. However, the process configuration used for the existing plant is not really economically viable for any large-scale gas production facility.

The existing facility requires compressors to generate the bubbles which start the gas laden water rising and also for compression of the produced gas. Philip’s idea is to generate the bubbles using excitation and to perform the separation at depth so the produced gas is already at pressure. He has managed to interest Murray and Roberts in his idea and they decided they would proceed with conceptual engineering for a new plant with his new technology.

Marion, another process engineer from AVEC, with whom I also worked at my previous company, and I get the job of doing the conceptual process engineering for the new facility. We start by trying to match a thermodynamic model to the experimental data and obtain a reasonable fit although there is not really sufficient data to be sure. During the course of the process design we realize that the real risk to the process is in the kinetics and not in the equilibrium conditions. Our concern is that bubbles will not be formed fast enough at our proposed conditions and therefore insufficient gas will be generated and that we will not obtain good separation of the gas and the water. We suggest and agree that a test rig is required.

The Process Flow Diagram in Figure 1 shows the proposed configuration of the test plant. Due to the uncertainties mentioned before, allowance was made for bubbles to be generated using excitation equipment, gas injection and injection of gas laden water. The depth of separation could also be varied between 60m and the surface. Sufficient engineering to produce Piping and Instrumentation Diagrams was performed, a Hazard and Operability Study was held and Murray & Roberts continued with more detailed engineering of the test plant.

Figure 1

Everything went quiet although I knew something was still happening because every now and then I would get a call for more information. Eventually I heard that the test plant had been procured and was in a container on its way to Rwanda for assembly. I would shortly be required at Lake Kivu for commissioning.

The story really starts here. It starts the way it continues, a series of incidents that individually are small but lead to many delays and difficulties. The construction team is mobilized and sent to Lake Kivu. But the container is delayed in customs for 2 weeks. The team sits twiddling its thumbs in Gisenyi, a small town on the lake. This will contribute to unreasonable time pressure on the test work. The job of constructing the test rig is bid for a fixed cost in US Dollars. Between the time of the bid and the purchasing of equipment the Rand will appreciate from ten to the Dollar to seven. This leads to an extremely tight budget.

I get the message, “We are ready to test, fly out to Kigali on Sunday. You will need to stay until the following Sunday.” I pack my bags and I’m on the plane. The Rwandan brewery, who are part sponsoring the project, are supposed to arrange for someone to meet me and take me to my hotel so I can travel to the lake the next day. But no-one is there. The airport starts to empty and I am in a strange place, can’t speak the language and don’t know who I am supposed to contact, and it’s a Sunday night. Eventually I find an airport employee who speaks a little English and he finds me a taxi driver who also speaks some English. We decide to go to the brewery. Although it is closed, the Security Staff may be able to contact someone.

The security guard calls his supervisor. The supervisor calls the Chief of Security. He drives out to the brewery even though it is a Sunday night. He calls the Managing Director who apologises profusely. The Chief of Security organizes a hotel and drives me there. The good part of this story is the extreme helpfulness of all the Rwandans that I met while I was there.

The next day we leave for Gisenyi, it is about a 3 hour trip although only 160km. My driver speaks English because he has only recently returned from Uganda where his parents were refugees. The scenery on the trip is fantastic. The driver calls it the place of a thousand hills. But people are everywhere. Every bit of land is cultivated. And the extreme poverty is evident all the time.

I arrived just before lunch to find that the barge containing the test rig was ready for towing to position. Far from being ready for commissioning, the next job was going to be to tow the barge 5 kilometers into the lake where it would operate. I also met the rest of the team who had already been there for about 20 days. Besides Philip, there was Rory, an experienced construction manager who has done extensive work in Mozambique. A young mechanical engineer, Fabrizio, from Murray and Roberts and an experienced instrument technician, Fred, who I knew from the previous company, made up the rest of the team. There were many other contributors, from the local labour hired to assist with construction, to the existing gas production plant manager who was often on the barge to see what was going on.

After lunch, using an old, borrowed and fairly dilapidated tug boat, we towed the barge to its location. After endless interesting adventures caused by a fierce wind-storm and rain, with tangled ropes and heavy weights, we finally got the first anchor down sometime after dark. A spectacular result of this late night work was that we got to see the nearby volcano glowing in the dark, something that was not visible from the shore. Then the Rwandan marines came to fetch us in a rubber duck to take us back to shore. The house we were staying at was right on the lakefront and the marines actually dropped us in our front garden.

The rig in position

The marines would actually fetch and carry us to the barge every day as well as assist with positioning of piping and fetch anything we required from the shore. They were another example of the extreme friendliness and helpfulness of everyone we met.

The next day we went out early to our barge and I got an opportunity to see what had actually been built. With 20-20 hindsight it was evident we had made mistakes. During detail engineering, in order to save costs and without clear evaluation of alternatives, several changes to the process scheme had been made and there had never been a design review of these changes. The major change implemented was the deletion of the gas and gas-laden water injection systems but several other changes also occurred as shown in figure 2. There was also no design review after the actual selection of equipment and instrumentation. This led to problems from sensitive electronic equipment and the impact of using a petrol generator to provide power directly to the equipment.

Figure 2

The second mistake was that there had been no constructability review with the construction manager as he was away in Mozambique. Large pipes had to be connected in the water which probably wasn’t necessary and several steel pipes could have been plastic to facilitate handling. That construction proceeded so well in the circumstances was testimony to the adaptability and resourcefulness of the construction team. During this second day on the water we managed to connect all the major pipes to the separator vessel. There was not much to do in the evenings, the satellite TV never worked. We used to sit outside drinking the local beer, kindly provided by the brewery and watch the fishing vessels on the lake and the hundreds of birds in the sky.

On day 3, the winch to raise and lower the separator failed. The problem was eventually isolated to the motor. We removed the motor and sent it to the brewery workshop but the day was almost completely wasted.

On day 4 we spent the morning on shore while the brewery electricians worked on the motor, identified the problem as a capacitor, scoured the area for something of a similar size and eventually jury-rigged 2 capacitors that were located in Goma, DRC, in place of the one. The electricians came back with us to the barge and after considerable effort got the motor working properly again. This whole episode highlighted the issue of start-up spares, of which there were few. This problem was to re-occur with several items of electrical equipment and was even experienced with some mechanical items like bolts, nipples and valves. Luckily the friendliness and willingness of the brewery to supply missing items and the marines to help allowed us to continue. Hindsight again showed that more onshore testing, simulating as realistic conditions as practical would have been very useful.

On the fifth day the weather was terrible. High winds and swells made working very difficult. During the course of the morning we were visited by one of the marine gunboats and this turned out to be very fortunate since 2 of our team were so ‘seasick’ that they had to be returned to shore by the marines. During this day spontaneous generation of gas occurred proving that it is possible to lift gas to the surface using the approach we had proposed. As the weather improved we continued piping up all the required connections and connecting the instrumentation.

Installing the vertical riser pipe

The pressure sensor’s power supply failed and we had no indication left of pressure in the vessel or differential pressure across it, both of which are required for control of the process and especially for start-up. We did eventually get the sensors working by connecting 2 car batteries instead of the power supply card but even this only worked temporarily before they failed again. At the end of the day we dropped the vessel to 60m, finished of all the connections and were ready for start-up the next morning.

The day of the test finally arrived, we switched on the excitation generator and it failed. We had no way of exciting the formation of bubbles. We did manage to get gas flow started by raising the level of the separator and at one stage a significant flow of gas occurred but the separator, because it was insufficiently ballasted began to rise and we had to vent the vessel which filled it with water and stopped the process. During this time the annubar water flow meter, which had been working, also stopped working so we couldn’t really tell what was going on. The gas flow meter range was too high for the flows we were achieving and calibrated to the conditions at 60m deep and the pressure regulator was too unstable to maintain a constant flow, all adding to the start-up difficulty.

Fountains of gas-laden water

Just before quitting taking the excitation generator back to shore we managed to obtain a stable, measurable gas flow with the separator at 8m depth, but that is far short of the 40 to 60m required to make the process viable. At least this showed that the excitation was necessary. The gas flowrate was also far short of the predicted flow.

At this stage the process concept looked like it could be viable but we still had a long way to go. I returned home but the rest of the team continued to try and sort out the problems. During the next few weeks the excitation equipment was extensively investigated and it was eventually found out that water had leaked into the submerged casings via the cabling. The equipment was taken back to Johannesburg to be repaired and even though it was assured that they had been pressure tested, the glands still leaked a little when they came back to Rwanda. Fortunately the construction team was able to properly waterproof them.

In the meantime, the ballast had come off the riser pipe and had to be replaced. The vessel ballast was also improved, the test rig had to be re-anchored and minor repairs to leaking piping and rearrangement of pipe configurations carried out. During all this time there was supposed to have been a visit by government ministers and investors to view the generation of gas, but this had to be postponed indefinitely.

At last, everything seemed ready. Philip, Fabrizio and Rory lowered the separator, completed the start-up procedure and switched on the excitation equipment. Gas was produced at 20m depth. When they came to take gas samples to check the composition the sample containers leaked, so they had to be sealed. But they had performed enough tests to be sure that they were producing gas at reasonable quantities. So I got the call to come out again.

This time everything went smoother. I was met at the airport and whisked of to the lake, arriving in the evening. The next day was overcast but dead calm. The sample bottles were fixed and everything was ready to go. That is not to say it all went perfectly. The electronic instrumentation still didn’t work which made the process more difficult to control and meant we couldn’t quite gather all the data we would have liked.

We also had an analyzer with us which was more useful as an explosion hazard detector. But we had learnt something from our experiences so far. The first contingency was to use a calorimeter from the brewery. When this didn’t work we arranged the use of an analyzer from the University of Goma, kindly lent to us by the Volcanology Department. We even had a back-up plan to fly samples back to South Africa for testing. We will still do this just to check our results.

Anyway, during the next few days we produced gas at depths ranging from 10m to 55m in all the various test configurations we wanted to test. Flowrates were in the range expected during the initial design and gas compositions were even better. We managed to produce gas while the brewery head was present and had some impressive displays of gas laden water jetting 8 – 10m into the air. So now we were all back in South Africa waiting for government ministers and investors to decide when they would like to view a demonstration.

If you are wondering what happened to the goats in the title, we came very close to sacrificing one on several occasions. As a post-script, the story is by no means over. After planning to return to the lake for demonstrations in February, these finally were scheduled in mid-May. In the intervening time somebody managed to cut an anchor rope despite 24-hour guards, causing the other anchor to seriously tangle the underwater piping and cabling and then break. This was the end of the road for electronic instrumentation, for the test-rig. The rig drifted 2km south before the deep underwater piping came to rest on an underwater ridge. The team was quickly dispatched to the lake and the rig was towed back into position and tied down with five anchors. If we had known the intensity of the wind and wave conditions beforehand, and if we had known that the rig would spend 6 months on the water instead of a fortnight, this would have been the design from the start. So the team spent two weeks in April getting things back in order and then waited yet another month for the demo to be arranged.

One of the big surprises observed on the lake was the effect of lake water on different metallurgies and eco-systems in the lake. Some paints blister and flake off instantly. Galvanised surfaces disappear rapidly. Algae grows on every shallow submerged surface. Since there is no oxygen below 40m, the fish are tiny with none growing longer than 40mm.

The team assembled in May for the last time, having now spent more than three of the previous six month on the lake. Much valuable experience had been gained and most of the pipe repairs, excitation system checks and the operational procedures were well drilled. Starting up the system and producing gas was down to five minutes and quite predictable. Less simple was to shut down the system where a series of up to five operations are required, taking up to 45 minutes, to make the system safely “go to sleep”.

Stanley on the rig

D-day arrived. A flotilla of navy gunboats and launches with over 60 visitors took turns in docking with the test-rig and allowing the passengers to board to observe it in action. A stiff wind had built up through the morning, typical of the dry season, causing the rig to rock. The gas flare had been lit and the metre-high flame at the end of the 6m flare-line was roaring away and could be felt as the southerly breeze pushed the heat wave towards the rig. This was just as well – because the blue flame from the methane is barely visible during the day. The visitors were duly impressed and discussions over lunch turned to how soon this project could come on-line to end the country’s chronic power shortages.

Success was hard earned on this venture. After so many setbacks in the early months of the work, the locals advised that the goat sacrifice was necessary before the big day. The sacrifice was made with uncanny results. The lake came to a dead calm that afternoon. Everything worked as it was supposed to, even the electronics.

  • Contact us







    Website © AVEC 2016