Fukushima Daiichi crisis: Simulation
Events at unit 126 August 2011
A PC-based simulator used to analyze the event evolution at Fukushima Daiichi unit 1 has produced results broadly consistent with reported data. In particular, it found that had TEPCO not acted to inject seawater when it did, further core meltdown and even vessel penetration would have been likely. By Li-Chi Cliff Po
Fukushima Daiichi unit 1 is a General Electric BWR 3 rated 460 MW electric (1380 MW thermal). Units 2 through 5 are BWR 4s rated at 784 MWe (2381 MWt). Outside the reactor pressure vessel (RPV), there are two recirculation loops with pumps. Jet pumps inside the RPV downcomer enhance the core flow for better efficiency. All units have Mark I containment (that is, steel liners plus concrete drywells and torus-shaped suppression pools). Combined, the drywell and wetwell make up the primary containment vessel (PCV). There are safety/relief valves (SRV) to discharge steam from the steam line to the suppression pool during an over-pressure or forced depressurization condition. The outside housing is called the reactor building or the secondary containment. During only extreme over-pressure conditions, the operator may decide to open the drywell and/or wetwell vents. The consequences of such an action will be severe, since the containment air is already highly contaminated.
The emergency core cooling system contains a steam-driven high-pressure coolant injection (HPCI) system. It is also named the reactor core isolation cooling (RCIC) system in GE’s later designs. This system is triggered by a low reactor water-level signal and acts initially to draw water from the condensate storage tank. When that tank’s content is exhausted, it begins to draw from the suppression pool. It provides for coolant makeup either until battery depletion or until the suppression pool heats up enough to lose its suction head.
There is another isolation condenser system. After an isolation event, steam from the reactor steam dome is led into a heat exchanger located high up in the reactor building. The steam in the tubes gets condensed and returns to the core, so the reactor never loses its inventory. Water in the heat exchanger tank’s secondary side is heated and boiled off in a few hours, and therefore requires a low-pressure source of makeup water.
On the active side, the diesel generator-powered residual heat removal (RHR) pumps are triggered by a low reactor water-level signal to inject water via the low-pressure coolant injection (LPCI) mode. In this mode, large-volume flow is used to re-flood the core after the reactor pressure is lowered by either a pipe break or forced depressurization. The same RHR pumps may be used to cool the suppression pool operating in the torus-cooling mode or to depressurize the containment in the containment spray mode. There is also an independent low-pressure core spray system also drawing water from the suppression pool. Heat exchangers are lined up to remove the heat. Service water is used to cool the secondary side, which also requires AC power.
Description of events
The earthquake, which struck on the afternoon of 11 March 2011, caused an instant loss of offsite power and a reactor shutdown at Fukushima unit 1. The Richter scale-9 shock exceeded the plant design limit of 8.2. Onsite emergency diesel generators began to provide AC power for residual heat removal, but the diesel tanks powering these generators were soon knocked out by the tsunami. From that point on, limited DC battery power provided for lighting and some control functions of the emergency cooling systems. A complete station blackout (SBO) then followed, leaving no means for coolant makeup and heat removal.
In the immediate aftermath of the Fukushima emergency, the author used a PC-based simulator (PC-TRAN) to analyze the event evolution. The simulation run assumed boundary conditions starting from a SBO induced by an earthquake and a tsunami, followed by the venting of the over-pressurized containment unit and the later injection of seawater. The key transient variables, such as pressure, temperature, water level, and (most importantly) extent of core-melt and radiological release, are projected.
This article’s focus is limited to unit 1. The outcomes for units 2 and 3 are similar in nature with slight variations in individual event timing and the extent of core damage . A simulation of the events at the Fukushima unit 4 spent fuel pool has also been carried out , and is discussed briefly.
The first four hours
Below are outlined the events that were conducted during the first four hours of the simulation. These events are based on the sequence of events reported by the Nuclear Industrial Safety Agency of Japan in a 21 March 2011 report .
Immediately after the earthquake hits, the reactor manually scrams and both recirculation pumps, all main feedwater pumps, and the reactor coolant clean-up system are shut off. The turbine bypass valve is isolated on loss of condenser vacuum and the main steam isolation valve closes. The RCIC and isolation condenser system start when the reactor water level falls to below a designated set point, L2 (see Figure 4).
At 3360 seconds all AC-operated systems stop working as the emergency diesel generators have been damaged by the tsunami. According to NISA’s later report , the DC battery for the RCIC was flooded by the tsunami. So the isolation condenser was the only system to remove the decay heat. But its secondary side water inventory boiled dry around the same time. (A TEPCO spokesperson told NEI that it connected a diesel-driven pump to the secondary side of the isolation condenser and started pumping in water. But the pump ran into trouble after several hours and stopped working. The spokesperson said that had the workers been able to continue to pump water in, the disaster might have been avoided, or at least would have provided more time to get the situation under control.)
Figure 1 shows PCTRAN Fukushima unit 1 model NSSS mimic in the first hour; the isolation condenser and RCIC are working to remove the decay heat.
Four hours from the initiating event (14,400 seconds) the first band of safety relief valves are cycling to maintain the RPV pressure at around 74 bar. The containment pressure increases slowly by the relief valves’ steam discharge into the suppression pool (Figure 2). The drywell air temperature increases slightly because of the flipping of the vacuum breaker between the drywell and wetwell. The vacuum breaker is a free-hanging hinge that balances the pressure difference. The temperature of the suppression pool water increased because of the reactor steam discharge from the SRV. In the first four hours the fuel remained under water so its temperature remained constant, as illustrated in Figure 3.
Venting and seawater injection
No plant data could be collected from the later part of day 1 (March 11) until early morning of the next day. At that point the operators noticed that PCV pressure was becoming high enough to threaten failure, so they decided to depressurize. Our assumptions in the simulation led us to depressurize the RPV by opening the first set of SRVs at 43,100 seconds and the containment vent valve at about 59,000 seconds (Figure 5). In reality, according to later reports, the Fukushima staff actually sent one volunteer to hand-crank the valve in a very high radiation area.
Regardless, the venting of the RPV caused the hydrogen concentration in the PCV to increase. The subsequent venting of the PCV will have caused the hydrogen concentration in the reactor building (secondary containment), to increase as well (Figure 6). Ideally when venting of the primary containment takes place, material should go directly through the stack to the atmosphere. However in the case of Fukushima unit 1, there was no power for the fans to push the mixture of hydrogen, nitrogen, steam and fission gas (noble gases, iodine-131 and caesium-137) mixture through the piping, so a leak could have occurred from the piping into the reactor building. At 89400 seconds the sound of an explosion was heard. At this time the simulation shows that the concentration of hydrogen in the reactor building had exceeded 10%, making an explosion chemically possible.
Figure 7 shows that by 60,000 seconds the average temperature of the fuel had reached 2500°C, the melting point of uranium. Fuel melting would have likely begun at the centre of the core. Please note the temperature is in logarithmic scale for easy comparison with the original containment temperature. The simulated temperature transient agrees with the TEPCO report dated 15 May . At 106,440 seconds seawater injection into the core began via the fire water line. The simulation assumed large quantity and no leakage at the vessel bottom. More recent reports have indicated lower water levels than expected, suggesting that injected water may have drained out. Because we have no way to determine the leakage size, we have terminated the simulation run without going any further.
Near the end of the simulation (56,000 seconds), the containment mimic (Figure 8) shows that there is near entire core-melt. This finding is consistent with the later disclosure (15 May) by the Fukushima authorities .
Moreover, our sensitivity study shows that delaying the seawater injection time would result in complete melt through the vessel bottom and even corium-concrete interaction (CCI) in the drywell floor. Until now, there has been no evidence that such an event has occurred. If the injected water has actually leaked out of the RPV and stayed in the drywell bottom pedestal, it would have quenched the debris and prevented CCI from occurring. At this moment without further evidence from the plant, all of the above scenarios are quite possible.
Unit 4 spent fuel pool
Another significant event, the loss of coolant at Fukushima unit 4’s spent fuel pool, caused clad oxidation and radiological release.
Micro-Simulation used its SFP mimic to analyze the events that happened at Fukushima unit 4 on 18 March. The scenario assumed that the pool was filled with fuel that was freshly unloaded as well as fuel from previous cycles. Their combined decay heat was removed by the cooling systems.
The simulation shows that upon a loss of cooling or coolant event, the pool will heat up to boiling. Continued boiling would expose the top of the fuel assemblies. Heating up of the exposed fuel could in turn lead to cladding oxidation and radiological gas release [see also p16].
The scale of radiological release from a spent fuel pool could be even more serious than from a reactor, since a pool contains many more assemblies than a core. In addition, cracks can develop at the bottom of a spent fuel pool—especially in the cases of Mark I and II containments where the pool is located high above ground.
Based on the simulation, the author feels that a supplemental system, with its own water storage and lines to outside make-up tanks, should be installed to spray water on fuel pools in case coolant is lost. Its piping and power supply should be independent and hardened to assure effectiveness in adversity. With this in place there would be no need for helicopters or fire engines, both of which were employed at Fukushima.
The simulations described above were carried out using PCTRAN, PC-based nuclear simulation software for nuclear power plant training and analysis. Since its introduction in 1985, PCTRAN has been constantly upgraded and expanded. The current scope of the software covers severe accidents and dose release in performance as well as numerous types of PWR and BWR plant designs, including Generation III plants.
PCTRAN models have been supplied for BWR 2 plants at Oyster Creek in the United States and Tsuruga 1 in Japan. Several BWR 4 and 6 models have also been supplied for plants in Taiwan (Chinshan and Kuosheng), Switzerland (Liebstadt and Muehleberg), and the United States (Hope Creek and River Bend). Each one went through detailed benchmarking and verification. A modest modification of the specifications for those models in size and power level allows us to accurately model the Daiichi unit 1 plant.
PCTRAN has been installed at over 100 utilities, research organizations, and government agencies. Since 1998, it has been used by the International Atomic Energy Agency’s (IAEA) annual Advanced NPP Simulation Workshop as one of the sample models in the curriculum. Recently, one model of VVER-1000 simulator was ordered by the IAEA and delivered to Vietnam for its first plant, to be built by Russia.
Li-Chi Cliff Po is the founder and president of Micro-Simulation Technology, a PC-based simulation software developer in Montville, New Jersey, USA. He worked for GPU—the owner of Three Mile Island—for 20 years in its post-accident modification.
The present analysis marks the first time that a simulation tool has been used to analyze what has happened during a nuclear power plant emergency according to a brief set of boundary conditions. The results appear to be reasonable and consistent with later disclosures. At this moment of a couple months after the event, new findings are still being released. Extensive analytical work over all four of the worst-affected units, and the use of other sophisticated tools, are necessary to reach complete understanding of the events at Fukushima.
 Micro-Simulation Technology, Fukushima Daiichi Unit 3 PCTRAN Analysis: http://www.microsimtech.com/downloads/Fuku3.htm, visited 16 May 2011