We all know that Taiwan faces power shortages, but I wouldn’t say that our company directly “assists” the government. Our original goal was quite simple—to sustain the company, make a profit, and secure projects.
Fortunately, Taipower introduced a PPP (Public-Private Partnership) policy, which allows private companies to receive reasonable investment returns. Our team worked extremely hard, overcoming civil and electromechanical constraints, installing battery systems, and participating in Taipower’s power dispatch operations. As a result, the company became profitable.
Since planning and design are our core strengths, we retain control over all key systems. In February and May of last year, we ranked first on Taipower’s trading platform. This achievement would not have been possible without the government’s private trading platform and our team’s persistence in solving problems and overcoming challenges. 

From my perspective, a personnel management system is essentially the simplest form of automation. It's like going to an ATM. When you withdraw money, do you need 0.1 seconds to get the cash? Even so, some procedures are required; it might take 10 or 15 seconds. So, the first thing we need is stability. Secondly, we need the equipment to withdraw money from an ATM, but we don't need it. If it malfunctions or breaks down, then we need to predict what will happen. Of course, with automation and AI integration, we can make predictions. For instance, if we know tomorrow will be sunny, we can charge our batteries more today and more tomorrow. But if we predict tomorrow will be rainy, we know we won't need to charge as much, so we'll charge less. Therefore, because the entire environment operates very quickly, we need more information about the surroundings and future needs, which allows us to make such predictions and plans. Therefore, for me, this kind of thing might be what Schneider Electric called PLC automation 20 years ago, or what PLC automation was 10 years ago, but now it's still called PLC, and it still controls a few Ols, that is, a few inputs and outputs. The control is still fixed. But how does it improve? It's because it can predict the next potential loss or the next expected situation, so it can avoid it early. For instance, when we're using tires, we know the pressure, and when the pressure reaches a certain level, we know what the car should do first. The same goes for batteries. Batteries have full charge and discharge cycles, and they have DODs. To maintain a certain signal, it can't accept an immediate power outage; I want to maintain the system's operation. Thus, I might reduce the load and then use it. Therefore, for me, the first point is a bit niche; it's an automation, a loop-avoidance system. We just need to add many sensors to control and prevent the situations we might encounter. 

The first point is having your own BMS. If we have bought someone else's system, it will be very difficult to control their battery management system. For example, if I want it to discharge, I won't be able to control it. If the system could only control one area, you wouldn't be able to achieve your ultimate goal. Yet, if you could control every part of the system, it would be much more complete. As for AI, to me, it is a more complex, more sophisticated, and smarter system. You can do more than just internal self-protection and external protection. You can also incorporate air quality, weather conditions, and various judgments. Therefore, to me, it's like a fusion, only the software couldn't be this smart before. Nowadays, software and hardware development are intertwined, operating together. For instance, hardware development must complement software, unlike traditional systems, where hardware only does hardware. Right now, these two are just like twins. The collaboration between software and hardware is somewhat like a duo, just like when we dance. The trace and dance moves are actually the same. When I move forward, and when I spin, that's another part, a module. I'm not good at describing it, but for me, it's like a duo. We must own our own BMS. Nevertheless, because we can do both, they can be AI-enabled. If one side stops moving, only the other side is controllable, but it's somewhat flexible. No matter what you say or do, it just won't move. Therefore, this kind of multidimensional, current generation requires every part, every finger, and every joint to be able to engage. The more you control, the more functional your system and your AI become. Same idea, the fewer items you control, the less functional you are.   

Firstly, due to the fact that we are a private business, we will need to conduct soil testing and an architect's assessment before installing our equipment. After the equipment is installed, all the hanging decorations must first pass Taipower's communication and capacity tests. The tests can only be put online after the meeting. Before going online, there are many different regulations. For instance, the inner ring of the electrical system must be less than ten times the outer ring's resistance. Our standard is less than four times the resistance, because we believe that the lower the grounding impedance of our electrical equipment can solve problems. Therefore, regardless of whether it's a conventional or AC vehicle, we want the grounding impedance to be less than four ohms. Yes, but four ohms isn't the end. Afterwards, once the equipment is up and running, you need to test its HY, also known as the hybrid impedance. Currently, in our factory, the factory-installed equipment has a tolerance of around 1500 to 3000 ohms. Thus, after the equipment is tested and the entire circuit is completed, we will need to ask Teclast to perform communication capability testing. Only after the test is completed will it issue an "OK" signal, and only then will Teclast power on the equipment. In conclusion, there is a specific logic involved.  

Currently, in the overall green energy ecosystem, some solar panels have been blown away by typhoons, resulting in longer power generation times. Yet, we are still discussing power generation products with GE, a gas-fired power plant company. Thermal power plants take at least 30 minutes to start up, due to combustion and other factors. Therefore, I firmly agree that energy storage is currently the fastest energy source, as its response time from 0 to 1 is only 0.1 seconds. We will be able to react within 100ms. Furthermore, energy storage offers better regulation and utilization. For example, we currently have hydropower and thermal power generation. Taiwan requires that mechanical power generation should be 15%, but due to current power shortages in Taiwan, you often see mechanical power generation dropping to 6% or 5%. However, at this point, our energy storage capacity, in Taiwan, is 1: 2.5's 1: 2 change to 1:6, meaning that for every hour of power generation we use, we have a 6 hour energy storage system as a backup, the effect might be that, for example, if there is a power outage in the northern or central regions, because we have a 1:6 backup, we can meet the current power shortage for most equipment during immediate power generation. Therefore, the power supply to equipment will not drop below 15%, but may increase from 5% to 8% to 10%. Thus, I currently believe that energy storage is still one of the most important components in the system. 

Beyond these issues, and in all our planning and design, going back to what I've just mentioned, the more sensors we have, the more things we will be able to monitor, which will allow us to better we can predict the stability of our environment. Therefore, we invest heavily in sensors, such as fire detection, temperature sensing, voltage sensing, and current sensing, including mutual inductance. In many places, we might not just use one connection for sensing, but two connections. There are also various temperature sensors. For instance, the voltage range we usually see looks like this. If one day a wire suddenly drops or rises, you can predict whether there's a problem and pinpoint the stable point. Therefore, in fire detection and fire protection models, our strategy is to use sensors as much as possible, increasing the number of sensors to avoid human error. Due to that, sometimes, seeing a dense array of data, like the voltage of 6000 batteries, is less intuitive than seeing a single, unexpected failure. For instance, 1500 volts requires 1000 batteries. Let's imagine we need 1000 batteries. In a three-legged race, everyone moves forward in a 1-2-1-2 pattern. But with 1000 people, it's 1000 people on one leg, so everyone's steps need to be more monitored. If I only observe everyone taking the correct steps, it's not easy to find a stable point. Yet, if I can detect anomalies using sensors, like someone having a stomachache or twisting their ankle, my device can predict and immediately correct the steps, preventing mishaps.  

Regarding the company's development, given the current situation in Taiwan due to the issues with TSMC and NVIDIA, our GDP suppressed South Korea's last week — a feat that took 22 years to achieve. For me, this development is thanks to TSMC's manufacturing processes and NVIDIA's technology. Therefore, my future goal is to work closer to these companies. The company is indeed focusing on development in ITC redundancy, BPU, and the utilization of TSMC's facilities. The way that we build a redundancy system for TSMC is a key direction for the entire company. Regarding talent within the company, we are collaborating with the National Taiwan University (NTU) Cultural and Creative Industries Area and with National Central University (NCNU) on a personnel management system. We hope to work with university faculty and students, as the market is constantly changing, to collaborate on courses and implement new technologies, electronics, and ideas. For instance, in my collaboration with the NTU racing team, I noticed they no longer have mass-produced batteries, which saddened me. My goal is to help students find a competitive battery and apply it to their vehicles. So, my first point is that we need to be proactive; everyone needs to be involved. Having ideas alone is useless. We need to develop practical skills — physical involvement and hands-on-experience. Therefore, I hope to use my experience as a model to encourage more people to participate in our endeavors. Thank you.  

I've been working with the National Taiwan University (NTU) racing team for four years. The first year, a group of students came to me about their batteries and what had happened. Because I didn't have enough resources at the time, I asked them to contact manufacturer A and manufacturer B for specific tasks. Since we had established connections, I was able to resolve the students' problem. During the second year, the students came to me again and requested their progress for the following year. Since my company had more resources that year, I said I would help them to get some batteries from manufacturer X, and I would help them plan specific steps.

In the third year, I told the teachers that the interaction between students and us wasn't enough. I thought we should find a group of passionate students to do something together. Professor Chang from National Taiwan University had this idea, so I decided to start a course using unmanned vehicle concepts to attract students. In less than twenty days, we found about eighteen students. Now it's the second year, and we've developed an unmanned vehicle center. We're currently collaborating with the government. For example, Chenxi invests 10 million NT dollars annually, and the government also invests 10 million NT dollars annually. Therefore, over three years, the university can receive about 60 million NT dollars in collaborative funding. I hope that through this project and production, everyone will be willing to get involved and learn more.
  
We just saw these two topics. One is our cross-domain testing. In cross-domain testing, we disassemble the entire cross-domain testing process and then reassemble it. However, since we don't use anything made in mainland China, we will reverse engineer everything and manufacture these parts locally in Taiwan. The second is the larger-scale HV (High-Vehicle Vehicle). The two major issues are that Taiwan hasn't conducted research on these large vehicles, and the second difficulty is the durability of 50 to 60 tons. To ensure that it can withstand 50 and 60 tons, this requires design, planning, and algorithms.

We will not reuse the disassembled batteries. However, for frequency-modulated energy storage, if we plan to use them in remote areas such as mountains or schools, solar cells are very suitable. Our batteries may have a lifespan of 15 years. Even if some devices have only been used for about 5 years, they still have some residual capacity. We can use these batteries for a second phase of application. Therefore, remote areas, or relatively remote Wi-Fi 5G stations, are very suitable for this kind of utilization.