Overview
Literal Labs LBN: Wind Turbine Power Prediction
Short Summary
An AI model for predicting wind turbine generated power. Developed for Advantech's WEDA containers using Literal Labs' Logic-based Network (LBN) algorithm and ModelMill training platform.
Overview
Modern wind turbine systems are equipped with Supervisory Control and Data Acquisition (SCADA) systems that continuously monitor operational parameters such as wind speed, rotor speed, temperature, and power output. These systems generate rich time-series data that can be leveraged to build accurate predictive models.
Predicting wind power generation is critical for grid stability, energy trading, and efficient asset management. Variability in wind conditions makes generation inherently uncertain, creating challenges for operators who must balance supply and demand in real time. AI models can enable wind turbine farm operators to predict power generation in advance, and in doing so reduce curtailment, and optimize maintenance scheduling.
This Logic-based Network (LBN) AI model uses proprietary algorithms from Literal Labs. The model was trained using public data from Kaggle Wind Turbine SCADA dataset, which is a collection of real-time sensor inputs at a wind turbine in Turkey. The model predicts short-term wind power generation, and is designed to run efficiently within a small MCU, enabling scalable deployment across distributed wind assets.
Key Customer Benefit
Power generation prediction enables grid operators to balance supply and demand more accurately, reduce reliance on backup generation, and optimize energy trading and storage decisions under fluctuating wind conditions.
By downloading and deploying this model within the Advantech container, users can evaluate the capability and efficiency of Literal Labs's Logic-based Network (LBN) technology. This model is provided for evaluation purposes only and is subject to the licence agreement in the licence file within the SDK.
Key Features of the LBN AI model
- CPU-based: The model is able to execute directly on a CPU target and delivers AI inference execution on a single Cortex-M33 core in the Advantech ROM-2620 platform.
- Deterministic and Explainable: The logic-based model is deterministic in its operation, providing the same output for the same data input.
- Low energy: Our logic-based algorithm delivers significant energy efficiency (over 50x higher efficiency than neural networks), making it suitable for battery-operated applications.
ModelMill
ModelMill is a unified and highly automated AI platform that trains, validates and deploys Logic-based Networks based on Literal Labs' proprietary algorithms.
Environment Requirements
Hardware
This model is compatible with Advantech ROM-2620 and will run on the Cortex-M33 core.
Host Environment
- OS: Yocto Linux
Container Quick Start
For more details regarding Literal Labs LBN, please refer to : Advantech Container GitHub Repository