Listen
❯
Sensor Network - IOT
❯AI-dispatcher (operator) of large-scale distributed energy system infrastructure
Product Design Using AI
Large Companies
Knowledge representation - Bayesian Network
AI-dispatcher (operator) of large-scale distributed energy system infrastructure
Large Companies
Operations - Intelligent Monitoring of Infrastructures
Reducing Food Recalls With AI-Dri
Predictive analytics for the behaviour and psycho-emotional conditions of eSports players using heterogeneous data and artificial intelligence
Robot consciousness
Ontologies for smart buildings
Unmanned protective vehicle for road works on motorways
AI components for vehicle platooning on public roads
Use of robotic solution for traffic policing and control
Behavioural and sentiment analytics
From Weeks Down to Hours: How insurers innovate with AI to shorten claims processing time and improve customer experience
AI (swarm intelligence) solution for attack detection in IoT environment
Robotic solution for replacing human labour in hazardous conditions
Animal Disease Detection
Ecosystems management from causal relation inference from observational data
Weather Forecasting in Agriculture
Robotic vision scene awareness
AI to understand adulteration in commonly used food items
AI based text to speech services with personal voices for people with speech impairments
Large Companies
Manufacturing and Factories – Predictive Maintenance
Self-driving aircraft towing vehicle
Improving productivity for warehouse operation
AI-dispatcher (operator) of large-scale distributed energy system infrastructure
For:
Energy companies focused on AI solutions to drive energy production, transition
and distribution in large territoriesGoal:
Improve Operation Efficiency, Increase Revenues, Reduce costsProblem addressed
To develop an effective industrial AI solution that is able to recommend the
optimal control of energy infrastructure systems in real-time in order to:
satisfy the energy demands of consumers;
minimize possible negative impacts on the environment;
reduce operational costs through systems real-time continuous.
Scope of use case
Monitoring, optimization and control of large-scale distributed energy systems
using deep reinforcement learning (gas, oil, power, heat, and water transmission
and distribution infrastructure systems).
Description
Motivation
The existing technologies do not provide an effective
solution to the problem of optimization of distributed energy
systems in real time. At the same time, the effects of
optimization in the energy sector are substantial.
Objects (systems) under consideration
Real large-scale distributed energy systems (gas, oil, power,
heat, water transmission and distribution infrastructure
systems). The main features of systems under consideration:
Territorial distribution and a large number of
interconnected units of equipment with individual
characteristics
The complex physics of technological processes
Huge amounts of real-time information from various
sensors
Problem statement
The central goal of the AI solution that is being developed is
formulated as follows: to ensure the supply of energy of a
certain quality at the right time to all consumers of a
distributed energy system, taking into account all
technological limitations and minimizing the operational
costs through systems real-time continuous optimization.
Solving this problem requires solving a number of subtasks.
Solution approach
The AI solution uses an approach of industrial system
modelling based on hybrid models that combine the benefits
of traditional physics-based modelling and machine learning
capabilities. We use the reliable digital twins of energy
systems and virtual simulators to simulate the systems
physics (dynamics) and we train deep reinforcement
learning models of these systems.
Current results
PoC of the AI system has been developed, which consists of:
digital twins of real gas infrastructure systems;
reliable physics-based models and virtual simulators of
these systems, actively used in industry;
model-free deep reinforcement learning algorithms,
connected with the above-mentioned virtual simulators;
services for training models, visualizing and analysing
the results.
Computational experiments proved that the initial objective
can be achieved with the help of modern AI technologies. The
results show the effectiveness using AI based technologies to
optimize and control of distributed energy systems and that
these solutions can outperform both human capabilities and
traditional optimization algorithms that were proposed
earlier.
Technologies
Physics-based modelling, deep reinforcement learning
technologies, deep learning frameworks, big data
technologies, streaming platforms, cloud-native architecture
of AI-system.
Sensor Network - IOT
Raw Data
AI: Perceive
Times Series
AI: Understand