The speech content is as follows:
Jiang Jiuchun:
I'm talking about battery energy storage. Our university, Jiaotong University, has been working on energy storage, from power systems and electric vehicles to rail transit. Today, we'll be discussing some of the things we're doing in power system applications.
Our main research areas are microgrids and battery applications. In battery applications, the earliest electric vehicles we rode in used power system energy storage.
The most important issues regarding battery energy storage are, firstly, safety; secondly, lifespan; and then, efficiency.
For energy storage systems, safety is the primary concern, followed by efficiency. This includes maintaining high efficiency, transformer performance, lifespan, and energy utilization even after battery degradation. While there may not be a quantifiable metric to describe this, it is crucial for energy storage. We aim to address safety, lifespan, and efficiency through several key initiatives: a standardized energy storage system, and a battery status analysis system. Energy storage systems are widely used in electric vehicles and public transportation.
The use of energy storage systems, node controllers, and smart distribution boxes, which are currently in widespread use, improves the overall economy and stability of the system, enhances the core value of system integrators, and allows for seamless integration with backend cloud platforms.
This is a centralized energy dispatch system. The hierarchical structure was explained clearly this morning. We can achieve optimized dispatching of multi-energy storage power stations and microgrids over extended periods through multi-node controllers.
Now it's made into a standard intelligent power distribution cabinet. This is a basic feature of power distribution cabinets, which includes a variety of functions, such as charging and discharging functions, automatic protection, and interface functions. This is the standard configuration.
The node controller is the core device for local energy management, responsible for important data acquisition, monitoring, storage, execution of management strategies, and uploading. There's an issue that requires in-depth research: the data sampling rate and the timing of data sampling during upload. This allows for the analysis of battery data in the battery's backend, transforming battery maintenance into intelligent maintenance. We are also working on determining the optimal sampling rate and storage speed to fully describe the battery's current state.
If I drive an electric car, you'll notice that many electric car statuses frequently change and fluctuate. Energy storage in power systems faces the same problem, and we hope to solve it through data. Here, we need to determine the appropriate number of samples for a BMS (Battery Management System).
Next, I'll discuss flexible energy storage. Everyone claims a single battery cell can last 6,000 cycles, or even 1,000 cycles in a car, but that's hard to say. Now, you're building an energy storage system that claims 5,000 cycles, but what is the actual utilization rate? Because batteries have a major problem: during degradation, the degradation process is random. Each battery degrades differently, leading to increasingly large differences between individual cells. The degradation patterns of batteries from different manufacturers are also inconsistent. How much of the battery pack can actually be used, and what the usable energy is, is a question that requires careful analysis. Currently, when electric vehicles are first introduced, they are used from 10% to 90% capacity, but after degradation to a certain point, only 60% or 70% can be used, posing a significant challenge to energy storage.
Could we group batteries according to their degradation patterns, making a compromise to determine the optimal grouping size for better performance and efficiency? We hope to group them according to battery degradation patterns—is 20 batteries a suitable node, or 40? This involves balancing and optimizing efficiency and power electronics. Therefore, we are working on flexible energy storage, which is one of our projects. Another advantage is the possibility of tiered utilization. I think tiered utilization has had some value in the last two years, but whether it's worthwhile in the future remains to be seen. Once charge/discharge efficiency and battery prices decrease, tiered utilization faces some challenges. Flexible grouping can solve many of these problems. Another advantage is high modularity, which reduces the overall system cost. The biggest benefit is improved utilization.
For example, car batteries degrade by less than 8% after three years, with a utilization rate of only 60%. This difference is due to the varying performance of the batteries. By combining five battery modules, the utilization rate can reach 70%, thus improving efficiency. Connecting battery modules in series can also increase battery utilization. Energy storage can increase by 33% after maintenance.
Look at this example: after equalization, it can improve by 7%, and after flexible assembly, it can improve by 3.5%. Equalization can improve by 7%. Flexible assembly can bring one benefit: in fact, the battery degradation trajectories of different manufacturers are different. You need to understand in advance what the battery pack will become or what the parameter distribution is, and then you can make a targeted optimization.
This is a method that uses independent current control for the entire module's power, but it is not suitable for high-power applications.
The module features independent current control for some of its power outputs, making this circuit suitable for medium to high voltage and reusability. This is a method for MMC battery energy storage to be suitable for high voltage and high power applications.
Regarding battery state analysis, I've always emphasized the inconsistency in battery capacity and the randomness of degradation. Inconsistent battery aging leads to significant reductions in capacity and internal resistance. These parameters are commonly used to characterize this, with capacity and internal resistance being two frequently used metrics. To maintain consistency, you need to evaluate the SOC differences of each individual battery. How do you evaluate the SOC of a single cell to understand the inconsistencies and the differences in maximum power? We use SOC to maintain battery performance, but how is a single SOC determined? Current practices place the BMS on the battery system to estimate the SOC online in real-time. We want to describe it using a different method. We hope to use the sampled data from the system to analyze the battery's SOC and SOH, and then optimize the battery based on this analysis. Therefore, we hope to use automotive battery data—not quite big data, but a data platform—to expand the SOH estimation model through machine learning and data mining, and provide full charge/discharge management strategies for the battery system based on the estimation results.
With the data available, another benefit is that I can provide early warnings about battery health. Battery fires still occur frequently, so energy storage systems must prioritize safety. We hope to use backend data analysis to establish real-time information and medium- to long-term early warning systems, identify online warning methods for safety hazards on both short and long time scales, and ultimately improve the safety and reliability of the entire system.
In this way, I can significantly improve several aspects: firstly, increase the system's energy utilization rate; secondly, extend battery life; and thirdly, ensure safety, so that the energy storage system can work reliably.
How much data needs to be transmitted to meet my requirements? I need to find the minimum acceptable battery operating state. This data is necessary to support subsequent analysis, but it can't be too large, as sending a large amount of data would put a significant load on the entire network. Sampling the voltage and current of each battery in tens of milliseconds and transmitting it to the backend is impractical. We've found a solution: we can tell you the appropriate sampling frequency and the specific data characteristics you need. We then perform simple compression on this data before transmitting it to the network. Battery curve parameters in one millisecond are sufficient for battery evaluation, requiring very, very little data recording.
Finally, let's talk about the BMS (Battery Management System). The cost of energy storage is becoming more important than the cost of batteries. If you add all the functions to the BMS, you won't be able to reduce its cost. Since we can send the data up, and there's a powerful analysis platform in the backend, we can simplify the frontend. The frontend only needs data sampling or simple protection, performing a very simple SOC (State of Charge) calculation. All other data is sent up from the backend. This is what we're doing now: the entire state estimation and the sampling by the BMS go through the energy storage node controller and are finally transmitted to the network. The energy storage node controller has certain algorithms; the following is basically detection and balancing. The final calculation is performed on the backend network. This is the entire system architecture.
Let's look at the most basic, effective, and simple process: balancing, from low-voltage data acquisition and equalization data acquisition to current acquisition. The energy storage node controller tells the downstream components how to process the data, including performing one step of the SOC (State of Charge) analysis here, and then another step in the background. This is the smart sensor, battery management unit, and smart node controller we're already developing, which significantly reduces the cost of energy storage.
