Project Title: Krypton Shine
Funding Opportunity: Technology to Market 3
Solar Subprogram: Technology to Market
Location: San Francisco, CA
Amount Awarded: $855,711
Awardee Cost Share: $914,193

This project is focused on improving the efficiency and efficacy of solar operations and maintenance (O&M) through Krypton Shine, an advanced O&M decision engine to reduce costs. The most practical way to achieve cost reduction in this area while simultaneously improving reliability is to assist solar O&M teams to automatically learn from ongoing activities such that every O&M data point helps increase asset performance and reliability and supports reduction of costs.

With native support for ingestion and processing of solar monitoring data through a high-scale stream processing engine, Krypton Shine translates operations anomalies into an intermediate, machine-learning enabled signature pattern that allows owners, operators, and maintenance professionals to better detect and diagnose potential reliability, performance, quality or safety issues. Krypton Shine is designed to improve mean-time-to-detection of reliability issues and performance degradation, improve accuracy of diagnosis, and reduce mean-time-to-repair by increasing both the application and improving the efficacy of remote issue remediation.

This project will introduce several key innovations:

• A highly scalable, fault-tolerant, distributed stream processing engine with native support for common solar monitoring data formats, protocols, and integration options;
• A robust expression language for defining complex evaluation of solar monitoring data;
• A machine-learning engine with native capability to “baseline normal” for every time-series event using a dynamically-selected best-fit algorithm;
• A purpose-built data pipeline to deconstruct solar monitoring data into a logical substructure that defines reliability, performance, quality, and safety patterns of interest to solar O&M teams; and
• A distributed, high-speed decision engine designed to constantly and iteratively derive, identify, clean, and link information culled from historical O&M interactions including work history, root cause failure, operator logs, and weather.