Presentations

Abstract

Inefficiencies in data-driven decision-making reduce profitability across the wind energy industry. While wind farms generate extensive operational data, much of it remains inaccessible for systematic learning, trapped in databases, scattered Excel files, or maintenance reports written in free text. As a result, valuable knowledge from past events often stays with individuals rather than being institutionalized, leading to repeated failures and missed opportunities for economies of scale across portfolios. This study introduces an artificial intelligence (AI) agent, KEN, designed to establish a central knowledge base of operational insights for wind farms. KEN focuses on maintenance records, event comments, and work orders written by operators and technicians. By processing unstructured text, normalizing terminology, and connecting records across sites and years, KEN explains in plain language the dynamics of failures, identifies recurring root causes, and highlights corrective actions that proved effective in comparable cases. For example, recurring main shaft bearing issues described differently across farms—using shorthand, typos, or local jargon—were unified under a single failure mode category. This enabled portfolio-wide statistics on how often such failures occur, which root causes are identified by maintenance crews during inspections, how long repairs typically take, and which interventions are most frequently applied. Through this process, KEN reconstructs event histories and transforms fragmented records into structured, reusable knowledge. The approach addresses a persistent industry gap: the absence of mechanisms to connect and learn from human-authored operational records across sites and teams. By converting underutilized textual information into institutional memory, the system enables asset managers and operators to enhance decision-making, prevent repeated failures, and realize the benefits of scale in renewable energy portfolios.