By Ma Shiqing, Edited by Ma Guangyu
Since September, the global energy storage industry has been hit by three consecutive fires.
The Moss Landing Energy Storage Facility in Monterey County, California, once the world's largest lithium battery energy storage power station, caught fire again, burning approximately 1,300 battery modules. This marks at least the fifth fire at the facility within five years since it was commissioned.
A few days later, an energy storage system at the Korea Electric Power Corporation (KEPCO) Power Research Institute campus caught fire, destroying 28 battery packs.
Shortly after, a thermal runaway and minor explosion occurred at an energy storage facility in Southern California, triggering large-scale local evacuation and shelter-in-place orders.
All three incidents point to the sensitive issue of energy storage system safety. To reduce safety risks, the key is to identify hidden dangers as much as possible before grid connection.
Currently, in the industry, energy storage safety simulation testing is a common method, but simulation models can also contain errors.
Chen Xiaobo, President of the Xiamen Empirical Energy Storage Technology Research Institute, revealed that during a CATL project in the UK, simulation model errors were discovered through third-party verification. Only by combining simulation with empirical testing can system hidden dangers be exposed before grid connection.
Beyond safety verification, energy storage projects also face long-term operation and profitability issues. "Battery lifespan can now reach 20 years, but it is still quite difficult for power electronic equipment to achieve the same lifespan," said Wu Dianfeng, Director of the Energy Storage Technology Center at CATL.
Batteries are responsible for storing energy, while power electronic equipment such as PCS is responsible for converting electrical energy and connecting to the grid. Premature aging of either party will increase maintenance and replacement costs, directly eroding system availability and project profitability.
In Wu Dianfeng's view, the industry must focus on extending the lifespan of power electronic equipment, aligning it as closely as possible with battery cells, to avoid customers having to pay out of pocket for equipment replacement in the 10th or 15th year of project operation.
The reason lifespan becomes an issue is that energy storage projects are a business of calculating accounts, generating revenue by participating in electricity market transactions. For investors, energy storage is both a set of equipment and an asset: it must operate long-term and continuously generate returns.
To increase profitability, in addition to extending the service life, costs can also be reduced. Currently, the consistent action in the industry is to increase battery cell capacity and expand system scale, trading scale for cost reduction.
However, Lin Jiubiao, Chief Technology Officer of CATL's Domestic Energy Storage Business Unit, pointed out that this increase in profitability is not without limits. The increase in battery cell capacity is accompanied by heat generation issues; and as the system scale expands, costs such as transportation and hoisting may also eat up a portion of the cost reduction benefits.
How to find a balance among safety, cost, and long-term profitability has become a question that system R&D must answer.
At the same time, new fields such as grid-forming technology and Artificial Intelligence Data Centers (AIDC) are raising the bar: energy storage must not only help maintain grid stability but also keep up with the rapid changes in power consumption at data centers, while holding the bottom line of power supply switching and fire safety.
From the lifespan matching of batteries and power electronic equipment, to the trade-offs between cost and safety in scale expansion, and then to empirical verification and new scenario applications, the energy storage industry faces the same question: how to make the system withstand long-term operation and continuously deliver project profitability?
Recently, CATL held a special Energy Storage session for its Global Open Day. During a small-scale communication meeting, Wu Dianfeng, Director of the Energy Storage Technology Center at CATL, Chen Xiaobo, President of the Xiamen Empirical Energy Storage Technology Research Institute, and Lin Jiubiao, Chief Technology Officer of CATL's Domestic Energy Storage Solutions, exchanged views with media including Leiphone. The following content has been edited without changing the original meaning:
01. Battery Cell Capacity and Power Station Scale Are Not the Bigger, the Better
Q: Regarding the 3S (EMS, BMS, PCS) system for energy storage, what work has CATL done? How do you view the future technical research direction of the 3S system?
Wu Dianfeng: CATL has also deployed its own 3S system. The original intention was to bring out the best performance of the battery cells, rather than managing them with unified algorithms using off-the-shelf solutions.
In the 3S system, the energy storage battery is responsible for energy storage, the Power Conversion System (PCS) is responsible for energy conversion, and the upper-level EMS is responsible for energy management and dispatch. Nowadays, overseas projects may also involve automated trading modules, basically these four major parts.
We have some basic logic in power electronics, mainly pursuing three dimensions.
The first is reliability. Power devices are relatively fragile, susceptible to electrical breakdown and thermal breakdown. Therefore, we do not blindly pursue cost reduction but focus more on how to improve their reliability.
The second is efficiency. The PCS converts and outputs the input power, and this process needs to improve efficiency. Improving efficiency not only significantly increases customer profitability but also ensures that all power losses are dissipated as heat in power electronic devices or consumed by EMI (Electromagnetic Interference), both of which affect equipment lifespan. When pursuing a higher RTE (Round Trip Efficiency), it can both reduce power loss and help extend equipment lifespan.
The third is the grid, because power electronic devices achieve grid connection on the low-voltage AC side through their own power outer loop, voltage outer loop, and current outer loop control. This process requires interaction with the grid, responding quickly and accurately to grid requirements, which mainly relies on software algorithms.
We focus on developing power electronic equipment from these three directions: reliability, efficiency, and grid interaction.
Based on energy dispatch, there are currently PMS and EMS. Our logic is to divide control functions into slow and fast. We implement the fast ones through PMS. We have adjusted the architecture of the PMS so that it can interact directly with the grid without going through the EMS as a relay, thereby reducing one link and improving reliability and response time. The slow ones can be placed in the EMS because the EMS has massive data storage capabilities and strategic calculations.
We do not just develop a software algorithm; we also integrate battery cell data, lifespan models, and SOH (State of Health) with the PMS and EMS.
Q: What are the main aspects of achieving high-quality R&D for energy storage systems? Please summarize from the perspective of product R&D.
Wu Dianfeng: The most important aspect of an energy storage system is still its reliability. From the perspective of R&D principles, the first is what is called "less is more" in English. Minimalist design brings greater reliability because the fewer components you have, the less there is to maintain.
The second is that we understand an energy storage system as a large system that can be divided into many subsystems. Each subsystem can be further broken down into more subsystems. Breaking it down level by level will reach the underlying elements such as materials, processes, and algorithms. Then, we look at what needs to be stacked up from the bottom, what is necessary, what is redundant, and whether some things can be designed more simply.
The third is high integration. The design of an integrated system is nothing more than electrical, thermal, and mechanical structures. Your electricity is charged into or drawn out, and during this process, the equipment generates heat. How do I take the heat away through liquid cooling plates to prevent battery lifespan degradation or excessive side reactions?
In this process, if there is a thermal runaway, where do I need to direct the heat? The mechanical structure is more about thermal expansion and contraction. The thermal pads between the battery cells and the cooling plates face high requirements for the viscosity of the thermally conductive pads because the battery cells will expand and increase in temperature when heated. There is a lot of system integration know-how involved here, and putting all this together constitutes the know-how of an energy storage system.
Ultimately, operating a system harmonically, knowing when each subsystem should fade, when it should retreat, and how much power it should output—this is what makes system integration a high-quality system.
Q: Currently, the single-unit scale and investment of energy storage power stations are growing larger, basically from the megawatt-hour to the gigawatt-hour scale; the capacity of single battery cells is also increasing. In the future, as the scale continues to grow, what new and greater challenges will be faced at the level of energy storage system integration?
Lin Jiubiao: 280Ah, 314Ah, 587Ah—these are the battery cells currently in mass production. Of course, we also see some even larger ones. Making battery cells larger has its benefits, but there are also issues that need to be weighed.
The benefit is that as the power station scale expands, using larger capacity battery cells can reduce the number of cells, bringing a certain cost reduction. This is a revenue item.
However, the larger the battery cell, the more safety issues need attention. Another point is that the marginal benefit of profitability decreases after making the cells larger. Increasing from 280Ah to 314Ah yields a relatively obvious cost reduction benefit; going further up, the benefits will decrease.
Heat generation issues must also be considered. We see that 2-hour battery cells have taken shape, basically staying at the 587Ah capacity size specification. For 4-hour systems, there are some iterative directions in the industry now, and we are also researching them. For 6-hour energy storage, battery cell capacity may continue to increase. Because the capacity of the battery cell divided by the number of hours equals the current, and the current magnitude also affects heat generation, this is from the perspective of heat production.
The core of making the system smaller or larger is still the cost issue. One is the cost of the system integration itself, and the other is the cost of the entire EPC footprint. Expanding the system from 3 MW to 5 MW or 6 MW expands the profit margin; continuing to expand to 7 MW or 10 MW, the marginal cost reduction benefit has been calculated to be basically around 1 point something, which is relatively small.
Making the system larger will also increase the difficulty of transportation and on-site hoisting. Currently, a 6.25 MW system weighs about 45 tons. In the future, going up to 6.9 MW will be about 50 tons, and reaching 7 to 8 MW will basically be over 60 tons. Transportation and hoisting costs may offset the earlier cost reduction benefits.
Therefore, I believe that the current system scale of 6 to 8 MW is relatively reasonable. Continuing to make it larger may not have technical obstacles, but the key is to solve the limitations of external conditions such as transportation.
Q: In the process of system R&D, what is the most important indicator?
Wu Dianfeng: In product R&D, it is actually a multi-objective optimization process. A single indicator cannot produce the best product. We have a model; for example, cost is certainly one dimension, but a product cannot be designed solely based on the cost dimension.
We incorporate indicators such as cost, SOH (State of Health), and RTE into the product design model, seeking the optimal balance point among multiple objectives, which is what we call the "sweet spot," and determine the product solution accordingly.
Personally, I think it is irrational to blindly enlarge the mechanical structure. Making battery cells and cabinets larger does not mean the number of components will decrease; it just combines more components into a larger structure. Capacity increases may reduce some shared components, but besides the mechanical structure, there are many other dimensions worth our consideration. Therefore, I advise the industry not to blindly pursue larger sizes.
02. Passing Simulation Does Not Necessarily Mean the Energy Storage System Is Safe
Q: How do laboratory empirical testing and on-site field testing complement each other? Why can the laboratory verify some operating conditions that are difficult to cover on-site?
Chen Xiaobo: It is hard to predict when on-site failures will occur, and grid-connected equipment dare not be operated to the limit of its capability. Energy storage charge and discharge cycles are relatively slow, and extreme conditions such as sandstorms and aging often require a very long time.
The laboratory can reproduce the most severe historical operating conditions, and testing allows for equipment failure or even combustion; it can also evaluate the long-term operational reliability of equipment in a relatively short time through environmental simulation and accelerated aging.
Risks at the system integration level mostly come from matching issues between components. For example, the conductivity of the coolant changes during operation, leading to sparking in the coolant and fire protection pipelines; water flow elbows may also spark with the enclosure; incorrect switching sequences during commissioning may also cause fuse blowouts.
A large number of hidden dangers in detailed design need system testing to be exposed. Accumulating extensive empirical and failure experience can feed back into design specifications, achieving risk front-loading. The laboratory mainly verifies extreme boundary conditions, while on-site field testing verifies long-term operation in real environments; the two complement each other.
Q: Many companies conduct energy storage testing through simulation and laboratory simulation. How can the collected data be made more comprehensive and targeted?
Chen Xiaobo: Some integrators' proprietary simulation platforms can build models to run on boards, but there is a lack of horizontal comparison among similar models, and a lack of third-party verification.
In a CATL project in the UK, we discovered simulation model errors through third-party verification. If these errors were left until the grid connection stage, it would prolong the delivery cycle and even cause grid oscillation.
Simulation alone cannot reproduce all real operating conditions. Combining simulation with empirical data can trace back on-site failures, reproduce the causes of accidents, and carry out targeted improvements.
Q: Currently, battery cells, PACKs, and PCS all have corresponding certifications. Compared to system-level empirical laboratories, what shortcomings still exist in component certification?
Chen Xiaobo: FMEA (Failure Mode and Effects Analysis) is an effective engineering design method, but it heavily relies on experience accumulation. At this stage, the industry still has many cognitive blind spots regarding component matching and failure mechanisms within the system. Many risks can only be exposed through system-level actual testing.
Q: How does the laboratory constrain the related risks of financial institutions, and what plans are there to carry out docking cooperation? In the context of the construction of a new power system, does the laboratory dock with relevant national platforms to carry out grid connection testing?
Chen Xiaobo: The laboratory is open to the entire industry. Currently, the bidding prices in the energy storage industry are fiercely competitive, and equipment manufacturers are highly sensitive to testing fees. Several integrated equipment enterprises have visited, and subsequent plans include sending samples for testing.
Currently, our cooperation with BOC Insurance and large EPC operators is still in its early stages. Subsequently, we will focus on high-altitude energy storage projects, relying on empirical devices to study risk premium balancing solutions.
During the construction of the laboratory, we received support from experts at the System Institute, New Energy Institute, and Energy Storage Institute of the China Electric Power Research Institute (CEPRI). We have also joined the National New Energy Storage Innovation Platform Alliance and participated in the Manufacturing New Energy Storage Innovation Center. Subsequently, we will promote the grid connection commissioning of SST and the implementation of joint debugging empirical testing.
Q: The implementation of grid-forming technology needs to go through stages such as standards, simulation, and third-party field testing. What verifications has CATL carried out? What is the differentiation compared to companies with a power electronics background?
Chen Xiaobo: The energy storage chain is divided into three links: grid-side grid connection control, power electronic regulation, and batteries. CATL's core advantage lies in the battery side. Our simulation and empirical system can penetrate deep into the DC side and the battery side. The energy storage DC belongs to an active system, and failure risks will propagate to the DC side. This penetrating capability is a key point in driving industry progress.
The Xiamen Empirical Laboratory obtained CMA qualification in September and is expected to complete CNAS certification by the end of the year, gradually establishing the credibility of test data. We have also obtained permission to carry out certification cooperation with TÜV Rheinland. The international certification system cannot be bypassed at this stage.
03. Grid-Forming Meets AIDC: How to Withstand the New Test?
Q: How does CATL view the development of grid-forming energy storage?
Wu Dianfeng: Personally, I am firmly optimistic about grid-forming technology. It can not only be used for energy storage but also for various power electronic equipment.
The grid is moving towards power electronification, originating from two ends: one is the photovoltaic and wind power on the power generation side; the other is the power consumption side, because the vast majority of electricity in our lives is converted into DC power for application, coupled with the current electrification processes of AIDC, electric vehicles, etc.
In this power electronification process, traditional power generation sources mainly rely on synchronous generators to achieve coordinated control through electromagnetic induction. The grid-forming technology of power electronic equipment can simulate synchronous generators and better adapt to the existing power system.
Therefore, my judgment is that grid-forming technology will become a mandatory item in the grid connection standards of the vast majority of national grids in the future. This is a trend.
In addition, grid-forming technology is not just about the power electronic equipment itself; it should be viewed from the perspective of the entire station. Power electronics are only a part of the control. The control of the PCS itself mainly regulates the frequency and voltage on the AC side of the grid connection point, but this is only local control. There is a larger control loop outside the power station, which also needs to coordinate with the grid.
I think grid-forming will gradually develop from a single power electronic component to the coordinated control of the entire power station, including the coordinated operation of PV-storage power stations and wind-storage power stations.
Q: Under the development trend of grid-forming technology and industry price pressure, what is the biggest challenge in the R&D of energy storage AC-side products?
Wu Dianfeng: Grid-forming technology does not bring significant changes to hardware design, unless the "3 times for 10 seconds" requirement brings about oversizing issues. But in our understanding, this requirement needs to be reasonably selected according to the grid application scenario and installation location.
If you look at many international grid-forming standards or the Grid Codes recommended by countries, there is no such requirement. Everyone is still able to provide support to the grid on the existing hardware basis by improving control algorithms.
For control algorithms, we will definitely make grid-forming an important direction for CATL's product R&D. In the future, you will see that our grid-forming control algorithms are somewhat different from the industry.
Another challenge is the lifespan mismatch between power electronic equipment and batteries. Battery lifespan can already reach 20 years, but power electronic equipment contains a large number of power device chips, and it is currently difficult to achieve the same lifespan. In the later stage of equipment operation, the failure rate may become higher and higher, and O&M costs will also increase; the decline in equipment availability will also bring losses to customers.
Therefore, I think the most important thing for the power electronics industry to improve is how to make the service life consistent with the battery cells, so that customers do not have to replace the power electronic equipment in the 10th or 15th year.
Editor's Note: "3 times for 10 seconds" usually refers to the short-term overload capability of grid-forming energy storage equipment. When the grid needs support, the equipment can output about 3 times the rated current or power for a short time and last for 10 seconds.
Q: Regarding the application of energy storage in Artificial Intelligence Data Centers (AIDC), what solutions does CATL have?
Wu Dianfeng: The AIDC customers we have contacted include both domestic and foreign ones, and their load characteristics in training and inference are different. But the first principle is that the GPU itself is a capacitive load. Simply put, when GPUs are training, they often perform a large amount of computation simultaneously, and the power demand will also rise or fall synchronously.
The load changes brought by this synchronous operation were relatively rare in the past. The original grid load could be very large, but it was a randomly statistical load and would not fluctuate like this. AIDC, especially training loads, may reach the gigawatt level and rise and fall together in a short time. The grid design is more adapted to slow load changes; it is not about how many GW, but how fast it changes and how large the amplitude is.
We will use grid-forming technology and energy storage to adapt to the fluctuations in frequency and voltage at the grid connection point. On the back side, we have power-type energy storage to solve the power fluctuations of AIDC. Through battery charging and discharging, energy storage can buffer the changes between the grid and AIDC loads.
Therefore, we define energy storage as the "decoupler" and "energy cabin" between the grid and AIDC, and design according to this idea.
Q: In CATL's view, what is the future picture of a zero-carbon society?
Wu Dianfeng: "Zero carbon" is the direction CATL has always adhered to, and it is also our vision.
According to our deduction, if carbon neutrality is achieved around 2050, global electricity consumption may grow from the current approximately 30,000 TWh to about 58,000 TWh, or even more. In this case, the entire electricity consumption will double.
But humanity has already spent more than 100 years building the existing power grid. It is not realistic to build another set of power grids to meet the new electricity demand. Land acquisition is very difficult, considering the investments of so many years.
Therefore, when there is no second power grid to support so much incremental electricity consumption, we believe that "zero carbon" will definitely move towards being centered on terminal loads, fully utilizing the wind and solar resources of terminal loads to convert them into electricity to solve the terminal electricity demand.
If the future energy system is dominated by renewable energy such as wind and solar, the demand for energy storage will be very large. Specific data is inconvenient to disclose externally. The energy storage industry has only just started in the short few years. This is a big mountain, and we are still at the foot of the mountain. The amount needed by the whole society in this area will be huge.
Why do we adhere to "zero carbon"? Because current fossil fuels are energy stored by solar energy on Earth many years ago. We extract it, transport it to power plants through ships and ports, burn it to generate heat, drive steam turbines to generate electricity, and finally send it to the power consumption end through the power grid. The entire closed loop is equivalent to wastefully burning a large amount of solar energy stored over a million or several million years ago, rather than applying it efficiently.
From the perspective of treating energy kindly, we feel that CATL has the obligation and responsibility to create a more zero-carbon society through technological innovation.