Improving Nuclear Power Plant Outage Scheduling with Artificial Intelligence

2019-10-30

The maturity of artificial intelligence technologies such as machine learning and natural language processing has enabled them to automate more complex and impactful tasks. Artificial intelligence has tremendous potential in reducing repetitive tasks, allowing engineers to focus more on the core engineering work rather than being consumed by tedious mechanical tasks. OPG, a power generation company headquartered in Ontario, has a total power generation capacity of over 16GW, with 5.7GW coming from nuclear power installations. Recently, the potential of artificial intelligence was fully demonstrated in a specific step of OPG's nuclear unit shutdown planning process. Within OPG's nuclear facilities, there are planned maintenance outages lasting two to three months each year. Each outage requires the scheduling of approximately 20,000 to 25,000 individual tasks, with most of them being similar to previous planned tasks. This means that a professional team consisting of highly skilled personnel must work tirelessly to complete these repetitive tasks. In the past, this kind of human resource waste was necessary. However, with the emergence of artificial intelligence, this inefficient scheduling pattern has been transformed. OPG has deployed an artificial intelligence solution for outage planning, focusing on predicting the logical relationships between tasks within the outage window to create a scheduling plan that includes all tasks. This approach not only reduces manual work but also maintains sufficient supervision and emergency functionality during subsequent revisions of the outage planning process, thereby reducing risk indicators. This solution is a customized, cloud-hosted application that seamlessly integrates with OPG's existing IT infrastructure. It leverages elements of AI, machine learning, NLP, and intelligent automation to address repetitive work throughout the maintenance process. It consists of three fundamental pillars: 1. Creating virtual tasks to identify missing tasks from the existing historical outage schedule. 2. Automatically scheduling tasks that have not yet been planned. 3. Eliminating logical dependencies to avoid meaningless loops in the schedule. These three pillars are at the core of the entire solution. They work through a data transmission mechanism that allows candidate nuclear unit outage scheduling instructions to be sent to the cloud-hosted environment. The key feature of this mechanism is its seamless integration with the existing technology stack and efficient prioritization of computer-assisted scheduling tasks for outage work. Finally, during the development of this intelligent maintenance solution, the collective team considered three key factors to ensure its longevity: 1. Improving labor efficiency to ensure the value of the solution. 2. Building solution scalability to plan maintenance for other power generation sites. 3. Developing a framework that considers scope changes to further enhance the capabilities of the current solution. Currently, the greatest advantage of artificial intelligence lies in its powerful computing capabilities, which are precisely the "nemesis" of repetitive work. Although today's artificial intelligence may not be "intelligent" enough, its powerful computing power has already made contributions in multiple fields.

RECENT AI NEWS

RECENT AI TOOLS