Smart Behavior Analytics Platform with Machine Learning for Utilities Applications (ART/242CP)

Smart Behavior Analytics Platform with Machine Learning for Utilities Applications (ART/242CP)

Smart Behavior Analytics Platform with Machine Learning for Utilities Applications (ART/242CP)
11 / 10 / 2017 - 31 / 03 / 2019

Dr Kangheng WU

1. Platform requirement and system architecture documents 2. Smart behaviour analytics key technical document 3. Smart behaviour analytics platform system a. Cloud-based system platform (Stream processing capacity >= 1M messages / second) b. Online machine learning (stream learning algorithms and supervised learning based on utility application data) c. Anomaly detection engine (trimming of error-prone data with pattern analysis algorithms (learning module on top of SBA)) d. Real-time Monitoring and reporting (node number >= 10K) e. High-frequent data analytics devices (sample rate >= 50/second) f. Extensible API for supporting outside applications 4. Tools and modules for smart utilities applications 5. CS from Nebula Technology Ltd a. Behavior Analytics System for IoT (SBA cloud platform for outdoor sensor/devices health status monitoring and data analysis) b. Extensible API and document to support IoT devices (software module to support data analysis of different formats from multiple devices and sensors) 6. Technical report a. Data analytics and modeling report for smart water applications (e.g. WSD) b. Behavior analytics algorithm research for utilities applications (e.g. HKU) c. The target production_loss_rate will be below 20% after the new system is in place in the deployment area (e.g., the metallurgical plant area, Hengyang City, Hunan Province. Its production_loss_rate is 40% currently. The SBA platform contribution should be at least 5% in the reducing overall-target production_loss_rate (20% = 40% - 20%)).

Zhuzhou Southern Valve Co. Ltd

Due to the improving performance of data storage, computation hardware and network communication, lots of advanced, data-intensive applications are available in utilities systems nowadays. These applications aggregate a wide spectrum of time-serial data that can be used to online alert, real-time monitoring and rapid fault location. This project aims to establish a smart behavior analytics platform with machine learning for real-time processing, discovering, understanding and analyzing the continuous time-serial data from utilities applications. The real-time intelligence is added to data analytics, giving user access to the cutting-edge technologies in machine learning. It provides online machine learning, anomaly detection engine, deep learning, real-time monitoring and high-frequent data analytics devices. The platform enables partners to build, develop, and commercialize new utilities products and applications on a common cost-effective platform.