Micro -network -based artificial intelligence approach for power outages?

Micro -network -based artificial intelligence approach for power outages
Micro networks are a system model in which electrical energy in the distributed energy resources group is produced. This network can be connected to the main power network, and when the connection is cut, it can work again with different modes. It is not affected by problems that will affect the main network while being supported by alternative energy sources. Uc Santa Cruz Electrical and Computer Engineering Department. Yu Zhang used the phrase for micro networks, micro Micro networks, both in industry and academy, focus on future energy distribution systems ”. Dr. Yu Zhang and his team developed a new approach for power for power by taking advantage of an artificial intelligence (AI) -based approach to increase the efficiency, reliability and durability of power systems. In the magazine IEEE Transactions on Control of Network Systems Zhang's student, Shourya Bose, made new definitions of artificial intelligence in his article and wanted to show that there was a better power restoration than old techniques in artificial intelligence developed.


Today, the electricity used by mass settlements is still dependent on public service companies that produce local electricity in infrastructure. In the use of local electricity, any natural disaster or weather events cause power outages. During the repair process, there is a disruption in daily life and people are victimized. Some households try to overcome these outages with minimal damage with generator or energy batteries. It acts as a distributor to re -provide electricity as a mixture of power supplies such as micro networks, generators or energy batteries. The spreading area of micro networks may vary according to the size. Zhang said, “Essentially, we would like to bring electricity production closer to the demand side to get rid of long transmission lines. This can improve power quality and reduce power losses on lines. In this way, we will make the network smaller but stronger and more durable. ” it aims to increase the distribution power with smaller size micro networks. In order to increase the function of micro networks, Zhang's laboratory has developed an artificial intelligence -based technique called “deep reinforced learning olan, which forms the basis of large language models. Furthermore, it has been seen that the reinforced learning controller responded much faster than traditional optimization methods at the time of power outage.

Reinforced learning, a system that perceives the environment in it and can solve the self, is a method that can learn to make the right decisions in solving the current problem. Likewise, it depends on the rewarding of the algorithm that sends successful response when the program is run in the programming language. Reinforced learning is an important factor in the power exchange of the network. Deep learning process, which can be adapted in accordance with the network problem, should be identified and the balance of power must be made in accordance with it. Bose said, “We model many things such as solar energy, wind, small generators and batteries, and at the same time we model when people's demand for electricity changes. Innovation is that this special variety of reinforced learning, which we call limited policy optimization (CPO), is used for the first time. ” 2 CPO approaches say that the use of machinery to find long -term patterns that affect the changing demand and renewable energy production in the networks over time. This is often different from the traditional systems that use a technique called model predicted control (MPC), which is generally based on existing conditions during optimization. The CPO method predicts that the brightness and temperature of the sun will increase at a certain period of time and increase the use of solar energy. In a day when the air is cloudy, long -term energy saving is being tried to ensure a different strategy. This system, which also provides information about different systems, found that CPO techniques perform significantly better than traditional MPC methods. It is aimed to use the artificial atomic algorithm, which is a metacular algorithm with a reinforced learning algorithm, and to see if this algorithm will be appropriate for such problems by applying it on hunter problem. In this study, the solution of the hunting problem was obtained by using the Q Learning Algorithm which is one of the reinforced learning algorithms. As a result of the study, the algorithm was successful in solving the problem and can be easily applied to such optimization problems.

Zhang and the research team proved that electricity transmission in power networks with reinforced learning and CPO techniques was made more efficiently in the competition. UC Santa Cruz researchers now see this as an indication that large -scale network operators can start to move towards artificial intelligence and artificial intelligence.

The research team, which now develops a successful algorithm in simulations, is trying to test its models on micro networks in their laboratories. In the long run, researchers hope to implement their solutions to the energy system of the UCAN Santa Cruz campus to resolve the interruption problems faced by the campus community.