The IEEE website will place cookies on your device in order to provide you with the best user experience. By using our website, you agree to the placement of these cookies. To learn more, please read our privacy policy.

This is part of the IEEE Spectrum Special Report: Revisiting Nuclear Power

By next summer, Chicago’s Exelon, the largest operator of nuclear power plants in the United States, and three international partners will make a decision that may initiate a nuclear renaissance. Under the leadership of Eskom, the national utility company in Johannesburg, South Africa, the four will decide whether to continue investing in pebble bed modular reactor technology. This concept originated in West Germany in the 1960s and 1970s and is now experiencing an amazing revival.

Specifically, the four companies will decide whether to provide funding for the full-scale prototype pebble bed reactor at the site of the 17-year-old dual pressurized water reactor nuclear power plant in Koeberg and Eskom near Cape Town, South Africa. Another important milestone, assuming that these companies grant approvals, they will appear more or less at the same time. By mid-2002, the South African government is expected to allow the construction of a pebble bed demonstration plant based on the company’s feasibility and environmental impact studies. The technical works of the design are displayed and can be commercialized.

Companies and national laboratories continue to develop other evolutionary and exotic reactor concepts [see "Other advanced reactor concepts"]. But so far, Exelon is the only nuclear operating company that fully accepts a specific new technology and announces that it has committed to building multiple nuclear power plants if the technology is deemed feasible and economically competitive. If the pebble bed project in South Africa continues, Exelon will seek design certification from the U.S. Nuclear Regulatory Commission (NRC) in Rockville, Maryland, so that the technology can be sold as a standardized off-the-shelf design, such as the three types that have been pre-certified by the NRC. Advanced light water design [see "Advanced Reactor Development Rebound", IEEE Spectrum, November 1997, pages 41-48].

Exelon hopes to submit an application for a pebble bed modular reactor license to the committee at the end of 2002. Construction will begin around mid-2005, and the first unit will start operating in about three years.

The pebble bed concept is based on the high-temperature gas-cooled reactor technology developed in Europe and the United States in the 1960s. The typical representative of this technology is the use of helium coolant and graphite moderator. In the pebble bed version, the fuel-uranium dioxide enriched at 235U to 8-10%-consists of particles coated with two layers of carbon and a layer of silicon carbide, embedded in a carbon matrix, as the main barrier to prevent radioactive release [see picture].

Pioneers in cobblestone bed design produced some mixed results. At the German National Laboratory in Jülich near Aachen, a 15-megawatt experimental research reactor has been in operation for 21 years. Its German successor is a 300 MW demonstration reactor that has only been in operation for four years.

In the United States, General Atomics evaluated the high-temperature gas-cooled design in two reactors built in San Diego, California: the Taodi No. 1 nuclear power plant, a demonstration plant that operated in Pennsylvania from 1967 to 1974, and Fort St. Louis. Vrain operated in Colorado from 1974 to 1989. After some initial errors in the fuel design, the demonstration of air-cooling technology was generally successful in the Taodi No. 1 project. But this success did not transfer to the expanded Fort St. Vrain, the project was also at a time of poor management.

Exelon describes its pebble bed design as a hybrid that combines parts of past gas-cooled reactors with new engineering. The plant consists of a steel reactor pressure vessel with a diameter of 6 meters and a height of about 20 meters. It is located in a 48-meter-high reactor building, about half of which will be underground [see figure]. Helium extracts heat from the core and drives a turbine.

This fuel is based on work done in Germany and the United States and is made using tiny pieces of coated uranium dioxide. Approximately 15,000 of these fuel particles are encapsulated in a protective graphite layer to form a 60 mm "pebble", about the size of a billiard ball. There will be 330,000 fuel elements in the core, surrounded by approximately 110,000 solid graphite spheres of the same size, acting as a neutron moderator.

The concept is modular, and each reactor vessel and turbine system provides approximately 130 MW. A control room can serve up to five modules, and a site can have up to 10 modules. Such modular factories are considered by some to be the trend of the future because they can be constructed relatively cheaply, the components are manufactured in the factory and quickly welded together on the factory site, and they allow the addition of additional units as the power demand increases.

Exelon said that another cost advantage of this technology is its continuous online refueling system. The fuel pebbles circulate through an automatic inspection system every 87 days. The system will measure the reactivity of each pebbles to determine how much energy is consumed, transfer the fuel that still has enough energy to the fueling pipe, and kick the spent fuel into a separate The fuel tank, and then the damaged fuel in the funnel is put into another container.

Exelon expects each fuel ball to circulate 10 times before it runs out. According to the preliminary design plan, due to the continuous loading and unloading of fuel, the plant only needs to be closed for maintenance shutdown for 30-50 days every six years. (Light water reactors must be refueled every 18 to 24 months, and they usually shut down for about three weeks.)

Like most other reactors, the pebble bed relies on a control rod system to regulate reactivity. In addition, it will be equipped with about 17 neutron absorption spheres designed to be placed in the holes of the side reflector to cool the reactor for maintenance.

The proposed containment structure includes internal steel coverings on reactor pressure vessels, and perhaps reinforced concrete external structures, designed to withstand earthquakes or small aircraft crashes. In general, the design of the container is not as leak-proof as standard containers. Ward Sproat, vice president of international projects at Exelon Generation, said that this is because the pebble bed reactor uses helium, and there will never be the kind of steam pressure accident that traditional containment designs can withstand.

The fuel in the pebble bed requires a very high temperature and does not decompose

In contrast to light water reactors, in light water reactors, the loss of water coolant will cause a large amount of heat to accumulate and core melting, while the loss of helium coolant in a pebble bed reactor will not cause fuel to melt. Even if the control rod is not inserted, the maximum temperature of the core will be 1600°C, which is much lower than the fuel's melting point of about 2000°C.

The terrorist attacks of September 11 naturally heightened concerns about whether the proposed pebble bed containment — or indeed any containment — could withstand the impact of a fully-fueled commercial airliner. Exelon is considering placing more factories underground.

Among all the characteristics of pebble bed technology, the most attractive is its fuel, which can withstand very high temperatures without decomposing and releasing radionuclides. According to its engineers, coupled with the use of passive systems (non-machine activation) to eliminate the heat accumulated in the reactor core in the event of an accident, the fuel pebbles provide a guarantee against melting. They say that even in one of the worst cases—the complete loss of helium coolant—the waste heat will naturally flow out of the reactor vessel and containment building into the environment.

The gas-cooled reactor competing with the pebble bed plant is the Gas Turbine Modular Helium Reactor (GT-MHR), which is based on four years of development work by General Atomics.

The reactor is being developed under a project jointly funded by the U.S. Department of Energy and the Ministry of Atomic Energy of the Russian Federation (Minatom), which uses it to destroy excess Russian weapons plutonium. General Atomics is working to translate this concept into a commercially viable reactor that burns low-enriched uranium fuel instead of plutonium. The company hopes to bring its GT-MHR and pebble bed models to the market at the same time.

The GT-MHR has larger modules than the pebble bed design, and the reactor will be located completely underground in a ventilated concrete "silo", thus removing it from the willful aircraft as a target. A standard GT-MHR power plant is conceived as having four modules, each with an output power of 285 MW, and a common control room.

Like its pebble bed competitors, the GT-MHR is a helium-cooled reactor that uses ceramic-coated fuel, but is lumpy. Each fuel block is about 0.75 meters high, and 10 blocks high are stacked in hexagonal graphite elements to form a core. The 102 pillars are arranged in three concentric circles, like a honeycomb, consisting of 1020 blocks.

The fuel will be placed in the inner ring with inert graphite blocks in the center and outer ring. This arrangement provides passive cooling, eliminating the need for a mechanical system to remove waste heat in an accident. Lawrence Palme, the safety and permit manager of General Atomics’ power reactor division, said that refueling may take place every 18 months.

So far, the size of the factory is still being drafted. Nonetheless, designs in the 1990s showed that the reactor silo (which also included the power conversion system used to generate electricity) was more than 40 meters below the ground and 26 meters wide.

The turbine, generator and compressor will all be mounted on a shaft using magnetic bearings. Parme believes that this is an area that may require demonstration tests in the United States and may be proven in Russian projects. The latter is tentatively scheduled to start construction in 2006.

According to Parme, GT-MHR will be ready for the market within a few years, and may build a factory in the United States by the end of 2010 or earlier, "if the US market is suitable for more active deployment and NRC certification," he said .

The passive safety features of these high-temperature gas-cooled reactors—most importantly, natural air convection will cool their cores, without the real danger of core melting—may be their big selling point among the public. "It's like a cup of hot coffee. You turn it off and it cools down," said Corbin A. McNeill Jr., chairman and co-CEO of Exelon, when talking about the design of the cobblestone bed.

However, for utilities that may or may not order such reactors, economy is of the utmost importance. Here, as the leaders of Exelon and General Atomics have seen, the decisive factor is the higher thermal efficiency of the air-cooled reactor. The efficiency of the pebble bed is estimated to be 42%, which is about 50% higher than that of the light water reactor, while the HT-MHR of General Atomics may be as high as 48%.

If these efficiencies are achieved, the two companies believe they can build factories at a capital cost that is competitive with other forms of power generation used for large-scale power production. The target has been around US$1,000 per kilowatt, but some industry officials have stated that this number needs to be higher.

In any case, the cost of the first gas-cooled reactor may be at least $1,200 per kilowatt. This is not ideal, but it is still higher than the estimated cost of building a new light water reactor in the United States, which is thought to be as high as $1800 to $2000.

JENNY WEIL serves as the Executive Editor of Inside NRC, responsible for regulatory and commercial affairs of the nuclear energy industry. The newsletter is published in Washington, DC by Platts, a division of McGraw-Hill Corporation.

For more information on reactor technology, please visit the following websites: http://www.pbmr.com is the pebble bed reactor, http://www.gat.com/gtmhr.html is the GT-MHR of General Atomics.

For more information on pebble beds, see Dave Nicholls, "Pebble Bed Modular Reactor", Nuclear News, September 2001, pages 35-40.

Chinese tech giants are struggling to gain autonomy from U.S. chipsets

Craig S. Smith is a former New York Times reporter and host of the podcast Eye on AI.

View of Yitian 710, an ARM-based server processor developed by Alibaba.

With the announcement of Alibaba's design of a 5nm technology server chip based on Arm Ltd.'s latest instruction set architecture, China has taken another step towards semiconductor independence.

However, despite this impressive feat, a more important chip design development by the Chinese technology giant may be providing source code for the RISC-V CPU core designed by its own engineers. This means that other companies can use it in their own processor designs-and avoid architectural licensing fees. (The company made two announcements at its annual cloud conference in its hometown of Hangzhou last month.)

The Chinese government is funding many start-up companies that are designing various chips. In the first five months of 2021, the number of newly registered Chinese chip-related companies more than tripled over the same period last year. China's largest technology companies such as Alibaba, Baidu and Huawei are developing their own chips instead of relying on chips from Intel, Nvidia and other American companies.

"Flagship technology companies like Alibaba can help start the semiconductor industry by manufacturing very advanced chips," said semiconductor consultant Linley Gwennap.

China intends to develop semiconductor independence in the design and manufacture of the most advanced chips. The urgency of this is due to the US sanctions on Chinese telecom giant Huawei, which makes it impossible for the company to use foreign-made chips. The sanctions apply to any Huawei supplier that uses US parts or technology.

The United States was shocked by China's actions to place Taiwan under its control. After allowing most of the semiconductor manufacturing industry to migrate to Taiwan, it also embarked on an ambitious plan to "re-support" its semiconductor manufacturing industry. About 80% of the world's semiconductor production capacity is in Asia, and almost all the most advanced logic chip production is in Taiwan. Currently, no Chinese semiconductor foundry has reached the 5-nanometer process required to manufacture Alibaba's new ARM-based chips, so it still relies on Taiwanese manufacturing.

But in the long run, the influence of Alibaba on the general choice of Arm and RISC-V instruction set architecture may be more important. The instruction set architecture or ISA is the language of software and hardware dialogue, so it determines the type of software that can run on a particular chip. Most servers use CPUs based on Intel's x86 instruction set architecture. But the British-based Arm authorized its instruction set architecture to chip designers and has established a firm foothold in this market.

The RISC-V instruction set architecture has fewer additional strings. RISC-V refers to the fifth-generation open source simplified instruction set computer architecture created by American researchers. It is free, so it is not affected by geopolitical crosswinds.

China has two industrial groups that promote RISC-V: China Open Command Ecosystem Alliance and China RISC-V Industry Alliance. In June of this year, China hosted the fourth annual RISC-V summit, where industry, academia, and government gathered to discuss the future of architecture.

After the US sanctions, Huawei was also banned from using Google’s Android operating system. Huawei released its first RISC-V development platform to help engineers use its own Harmony operating system for smartphones, IoT gadgets, and other The so-called edge device. Due to sanctions unable to purchase Intel chips, Huawei recently sold its x86 server unit to a company in China's Henan Province.

Alibaba launched its first RISC-V processor in 2019 and was hailed as the most advanced RISC-V chip at the time. From the beginning, the company expressed its intention to open the source code of the CPU-a hardware description language that describes the structure and behavior of the electronic circuit of the CPU core. It has done so now...without fanfare.

"If Intel made the same statement on the design of the x86 instruction set microprocessor, it would be a big deal," said David Patterson, one of the creators of RISC-V.

As more and more chip and software vendors adopt this architecture, RISC-V is gradually surpassing Arm and Intel. Patterson pointed out that all NVIDIA GPUs use RISC-V, Samsung phones use RISC-V, and most open source tools are suitable for RISC-V. "RISC-V shipments have reached billions," he said, adding that Alibaba alone has shipped more than 1 billion cores using RISC-V. At the same time, several other open source RISC-V kernels are already available on the Internet.

With RISC-V processors for low-power tasks and custom Arm server CPUs for general-purpose computing, Alibaba now has a full range of computing infrastructure. Its Yitian 710 server system-on-chip (SoC) manufactured by Taiwan Semiconductor Manufacturing Co., Ltd. will have a total of 128 Arm-based cores, integrate 60 billion transistors, and have a maximum clock speed of 3.2GHz. Alibaba said that this is the first server processor compatible with the latest Armv9 architecture.

Alibaba said that the SoC scored 440 points in SPECint2017 (a standard benchmark for measuring CPU integer processing power), surpassing the current state-of-the-art Armv8-based Arm server processor by 20% and more than 50% in performance. . energy efficiency.

The company also announced the development of a proprietary server called Panjiu developed specifically for the next generation of cloud-native infrastructure. By separating computing from storage, the server is optimized for general and dedicated AI computing and high-performance storage.

At the same time, by opening the source code of its RISC-V Xuantie series of IP cores, developers will be able to build their own prototype chips to customize for different IoT applications. Alibaba has also opened up the software stack related to Xuan Tie, supporting multiple operating systems, including Linux, Android, RTOS and Alibaba's own AliOS. The company promises to provide more services and support for RISC-V development tools, software development kits and custom kernels in the future.

Consultant Gwennap said that Alibaba's Arm and RISC-V efforts are more experiments than commercial efforts, and pointed out that most of Alibaba's internals are still using x86 Intel chips. Gwennap said: "These companies talk a lot about Intel's alternatives." "But in the final analysis, they don't want to eat their own dog food."

Alibaba's newly launched Arm-based server chip will be used in Alibaba's data center to provide cloud services to customers. The company will continue to provide Intel-based services, so customers can choose Arm instead of x86-based chips. When Amazon did something similar a few years ago, Arm-based chips were hardly adopted.

But true semiconductor independence will require China to develop its own extreme ultraviolet lithography machine and to etch microcircuits on silicon. China's main chip foundry, SMIC, cannot provide any products smaller than 14 nanometers. SMIC claims to have mastered the 3nm chip process in the laboratory and is trying to purchase the EUV lithography machine needed for production from the Dutch company ASML, which currently monopolizes key equipment. But the United States intends to block this deal. (3 nm refers to the next reduction and tighter spacing of the smallest semiconductor feature size to allow increased transistor density, but does not refer to the actual size of the transistor gate or other features on the processor.)

The Chinese Academy of Sciences has an EUV lithography research team, and Tsinghua University has developed a new particle accelerator light source that can be used for EUV lithography. However, it will still take many years to get this technology out of the laboratory and into the machine.

Japanese startup is committed to developing autonomous robots that can perform useful tasks inside and outside the space station

At the end of last year, Japanese robotics startup GITAI sent their S1 robotic arm to the International Space Station as part of a commercial airlock expansion module to test some useful space autonomy. Everything on the International Space Station runs very slowly, so it wasn’t until NASA astronauts installed the S1 arm last month that GITAI was able to let the system complete its pace—or rather, sitting on a comfortable chair on Earth. , Watching the arm do the most thing its task itself, because that’s a dream, right?

The good news is that everything went well, and the arm did everything GITAI wanted it to do. So what is the next step for commercial autonomous robots in space? The CEO of GITAI tells us what they are doing.

One of the advantages of working in space is that it is a highly structured environment. Microgravity may be a little unpredictable, but you know the characteristics of objects (and even lighting) very well, because everything there is over-defined. Therefore, things like using a two-finger gripper to complete relatively high-precision tasks are entirely possible because the system has to deal with very little changes. Of course, things always go wrong, so GITAI also tested the Teleop program from Houston to ensure that participation is also an effective way to complete the task.

Since complete autonomy is much more difficult than almost complete autonomy, occasional telemetry may be crucial for various space robots. We spoke with Sho Nakanose, CEO of GITAI, to learn more about their approach.

IEEE Spectrum: How much autonomy do you think the robots working in the International Space Station should have?

Sho Nakanose: We believe that combining 95% of autonomous control and 5% of remote judgment and remote operation is the most efficient way of working. In this demonstration of the International Space Station, all work is done under 99% autonomous control and 1% remote decision-making. However, in the actual operation of the International Space Station, there will be irregular tasks that cannot be handled by autonomous control. We believe that such irregular tasks should be handled by ground remote control. Therefore, we believe that the final ratio is about 5%. Judgment and remote control will be the most effective of.

GITAI will apply the universal autonomous space robot technology, know-how and experience gained through this technology demonstration to develop an extravehicular robot (EVR) that can perform docking, repair and maintenance tasks for on-orbit services (OOS) or carry out lunar exploration and Various activities such as the construction of the moon base. -Sho Nakanose

I’m pretty sure that you tested the system many times on the ground before sending it to the International Space Station. How is operating a robot on the International Space Station different from your tests on Earth?

The biggest difference between ground experiments and ISS experiments is the microgravity environment, but it is not that difficult to deal with. However, experiments conducted on the International Space Station in an unknown environment that we have never been to will encounter various unexpected situations that are difficult to handle, such as communication interruption due to the failure of the thruster ignition experiment. On the Russian module. However, we were able to solve all problems because the development team made careful preparations for violations in advance.

It seems that robots are performing many tasks using equipment designed for humans. Do you think it would be better to design things like screws and control panels to make the robot easier to see and operate?

Yes, I think so. Unlike the International Space Station built in the past, it is expected that humans and robots will be built to work together in the lunar space station Gateway and the lunar base in the future. Therefore, it is necessary to design and implement an interface that is easy to use for both humans and robots. In 2019, GITAI received an order from JAXA to develop guidelines for easy-to-use interfaces for humans and robots on the International Space Station and gateways.

What are you going to do next?

We plan to conduct an out-of-orbit demonstration in 2023 and a moon demonstration in 2025. We are also working on space robot development projects for some of our customers who have already received orders.

Transform more than 55,000 substations into independent micro data centers to solve the most pressing power distribution, security, and cost challenges that plague the U.S. power grid. Learn how virtualization provides a total solution to modernize and expand the power grid, with the capacity, protection, and reliability needed to keep power flowing smoothly.