Leveraging NVIDIA’s Software Platforms For Veritone CDI Exemplifies Our Commitment To Developing Cutting-Edge Technologies For Grid Optimization
Veritone, Inc, the creator of the world’s first operating system for artificial intelligence (AI), aiWARE™, today announced the migration of Veritone’s aiWARE and patented Cooperative Distributed Inferencing (CDI), a predictive smart grid optimization and control technology, to NVIDIA robust software stack and parallel-processing GPUs in order to meet the challenges of complex real-time AI model and device control optimization in energy delivery and generation. Addressing the dynamic environments of today’s complex electrical grids, which incorporate distributed energy resources such as solar, batteries, wind and hydropower, requires predictive, responsive AI to ensure that all devices under control across the grid are operating optimally and safely. Based on initial testing, Veritone projects that the CDI running on the NVIDIA EGX edge AI platform, which consists of NVIDIA GPUs optimized with NVIDIA software platforms and tools, will increase processing speeds for updating device models by up to 100x compared with multi-threaded CPUs. This accelerated processing will significantly improve battery control, forecasting, autonomous dispatch, and overall grid performance.
“Leveraging NVIDIA’s software platforms for Veritone CDI exemplifies our commitment to developing cutting-edge technologies for grid optimization and accelerated clean energy adoption,” said CDI inventor and Veritone Chief Data Scientist Dr. Wolf Kohn. “From autonomous, resilient grids to local battery optimization, Veritone sees a significant opportunity to effectively manage our nation’s fragile network of energy systems that grows more complex by the minute. We have more than a decade of engineering invested in this technology, and we are excited to see our responsive, real-time device modeling made possible through strong collaboration with NVIDIA engineers and their technology stack.”
The two companies are currently working to enable Veritone’s accelerated CDI rule-based model optimizer, which uses reinforcement learning for real-time modeling of solar batteries and other distributed energy resources, to run on NVIDIA EGX. Veritone’s solutions model a wide range of battery parameters, including charging/discharging states, operating conditions such as temperature and humidity, and warranty constraints. With the accelerated processing capabilities made possible by NVIDIA, Veritone’s CDI models update continuously in real-time based on constantly changing operating conditions, and the most optimal model is used at any given time to ensure the best operational state. This model optimization is done at the device level, and also synchronizes across multiple devices in a complex energy network so that the entire grid’s resources are operating at optimal performance levels.
“We are very excited about how Veritone is leveraging the power of the NVIDIA accelerated computing platforms to streamline the management and optimization of clean energy. We believe that this will help provide more predictable and reliable energy,” said Keith Cockerham, Utilities Industry Lead at NVIDIA. “The combination of Veritone CDI and NVIDIA EGX provides the millisecond-level autonomous decision-making capabilities required for smart grid management and resiliency.”
Veritone’s energy solutions use patented CDI technology to deliver real-time dynamic modeling and control of energy devices for predictable, cost-effective and resilient energy dispatch. Some clean energy sources, such as solar and electric car batteries, require extremely fast model refresh times to ensure proper energy charging, discharging and health monitoring. Energy spikes, anomalies and extreme weather events require a real-time approach to device model updating and device control. Applications for AI model updating using GPUs are abundant. Veritone is focusing their initial NVIDIA migration efforts on energy optimization of complex smart grids, and plan to develop optimized solutions for grid and electric car batteries in the future. For grid batteries, fast and accurate modeling of batteries on NVIDIA EGX are planned to increase battery utilization, reduce the risk of thermal events and extend battery life. For electric car batteries, the two companies plan to integrate Veritone’s dynamic modeling with NVIDIA EGX, which is expected to improve range, battery longevity and reduce the risk of thermal events.
NVIDIA and Veritone are co-hosting a webinar on March 4, 2021, “Optimizing Real-Time Models of Dynamic Energy Systems” Click here to register for the webinar, which will cover the following topics:
- How high-performance computing (HPC) on NVIDIA GPUs can solve large optimization problems for dynamic grid systems in real-time;
- Why the battery Data Tomograph modeling system is well suited for NVIDIA GPUs because of its high computational density and high data flow requirements; and
- How dynamic modeling can dramatically improve the economics of operating smart grids.
The combination of Veritone AI technology and NVIDIA EGX will open doors to real-time AI applications running on-premise, hybrid or in the public cloud, for use in markets such as energy, security, smart cities, media and entertainment, contact centers, and industrial and manufacturing.
This news was originally published at Business Wire