What is Kavaken?

Renewable energy asset owners face decreasing revenues, aging assets, and increased downtime. Their assets generate significant operational data that could help address these challenges, yet most lack the capabilities to exploit this data efficiently.

Kavaken is "an Artificial Intelligence-powered solution that increases revenue by optimizing asset management with a data-centric approach." The platform integrates three core functions: production forecasting, asset management, and predictive maintenance, working together to drive operational decisions focused on revenue optimization.

Module 1: Revenue Cockpit

This dashboard provides live monitoring of wind power plants with features including:

• General overview of all wind power plants on a live map
• Real-time data: wind speed, production, temperature, production summaries, and lost revenue tracking
• Summary table reflecting KPIs across all sites focused on revenue metrics (Production, Capacity Factor, Revenue-Based Availability, Energy-Based Availability, Time-Based Availability, Power Efficiency)
• Site pages with live maps showing individual turbine information
• Live turbine metrics: wind speed, production, temperature

Module 2: Forecast+

This AI-based module helps wind energy clients minimize forecast uncertainty. The module combines machine learning models with commercially available forecast models, recognizing that each has distinct strengths for specific scenarios.

Kavaken's proprietary three-step approach includes:

• Layer one: inputs from Plant, SCADA, and Weather data plus physical weather models
• Layer two: optimization of weather and physical models using machine learning
• Layer three: combining ML-based meteorological results with turbine-specific information

The system provides Day Ahead (GÖP) and Intraday (GIP) production forecasts. Users access a dedicated online portal with Kavaken Forecast tools and historical model accuracy comparisons.

Module 3: Power Booster

Using AI and machine learning, this module analyzes power curves and identifies unoptimized areas. The approach detects causes of underperformance and provides recommendations to improve power curves and increase turbine outputs.

Key capabilities include:

• Baseline power curve benchmarking for turbine performance
• Detection of power curve shifts
• Non-performance analysis (icing, curtailment, hysteresis)
• Yaw performance analysis

Module 4: Maintenance Planner

A maintenance management tool for entering, visualizing, and tracking maintenance activities. Features include:

• Status tracking (on target, delayed, planned)
• Commenting and documentation attachment capabilities
• No training data required; ready for use after integration
• One-stop portal for accessing maintenance-related comments and documents

Module 5: Predictive Maintenance

This module generates early warnings when major components (generator, gearbox) may experience breakdown. Using ML-based techniques, primarily anomaly detection, it analyzes both high-frequency vibration and SCADA data.

Methods employed include:

• Normal behavior model development
• Clustering techniques
• Anomaly detection methods

The goal is maximizing uptime by early detection of potential failures, reducing unplanned maintenance and downtime, and preventing losses from costly component failures and revenue interruption.

While initially focused on wind energy, "the capabilities are easily transferable to other renewable energy sources, making it a solution available to an $80B global market."