Dr. Razzaqul Ahshan

Associate Professor, College of Engineering

In Oman, as elsewhere in the world, the electric power grid is no longer a one-way street where power plants generate and deliver electricity to homes and businesses. Instead, with solar panels, energy storage systems, electric vehicles and flexible demand on the horizon, our distribution networks are becoming active, dynamic systems that require much smarter management.
Recent research at SQU presents an advanced AI-driven framework designed specifically for the type of distribution network we see here in Oman (and more broadly in the Gulf region) – where renewable energy and changing loads are increasingly important.

What’s the Challenge?

Let us step back: imagine a neighbourhood in Oman with renewable generation, some homes with batteries (or EVs), a few flexible loads (for example, air conditioners that can shift usage), and devices that support voltage and reactive power control. The network must coordinate all of this in real time, keeping voltage and current safe, losses low, costs low and emissions down.
Traditional scheduling methods in power grids are often designed for predictable generation and loads. But with high penetration of renewables (especially solar), increased variability, and distributed devices, the scheduling task becomes complex. The research tackles this complexity by combining the following powerful AI tools:

Multi-modal input: Instead of relying only on electrical measurements (voltages, currents, etc.), the system also uses visual data – images of solar panels – to detect “soiling loss” (dust, shading, dirt reducing solar panel output) via a CNN–Vision Transformer hybrid.
Multi-agent reinforcement learning (MARL): Multiple “agents”, each controlling a resource (for example, energy storage, flexible load, PV unit, static var compensator), learn to act cooperatively to optimise multiple objectives (operational cost, losses, carbon emissions) rather than a single agent doing everything.

Why the “Multi-Modal – Multi-Task” Approach Matters

The term “multi-modal” refers to using different types of data (electrical measurements and images), whereas “multi-task” refers to solving several goals simultaneously (cost, losses, emissions).
Images of solar panels help the AI to recognise when solar output might drop due to soiling or shading – realistic in Oman’s dusty climate. Electrical grid data offers the state of the network (voltages, currents, power flows). By combining both, the model builds a richer “state representation” of the grid environment. Each agent then decides – to charge or discharge a battery, to adjust a flexible load, to switch a PV unit or adjust reactive compensation – in coordination with the others. The result is a smarter, more flexible, more resilient active distribution network (ADN).

Implementation in an Omani Network

The authors validated their approach on a large-scale distribution network in Oman – so the research is not purely theoretical but tested in an approximate real-world scenario in the country.
The results showed that the developed AI framework:
• Achieved superior cooperative scheduling performance – meaning the multiple agents worked well together.
• Balanced multiple operational objectives (not just one).
• Enhanced grid flexibility – meaning the system could adapt to changing conditions.
• Outperformed traditional scheduling strategies (which treat resources independently or rely on simpler optimisation).

What Does This Mean for Oman and the Region?

For Oman’s electricity sector, and for readers interested in energy, innovation or smart grids, this research signals several things:
• Renewables + storage + flexible demand = complexity. As solar integration grows, and as energy storage becomes more common, distribution networks need more advanced management.
• Visual data adds value. In dusty, sun-intensive climates like Oman’s, panel soiling is real. Being able to detect and respond to that via AI is valuable.
• Coordination of resources is key. It is not enough to optimise each device alone; the system-wide view is critical.
• Local relevance. Since the testbed is in Oman, the research has direct regional applicability.
• Towards smarter grids. This contributes to the vision of an “active distribution network” – where the grid is adaptive, resilient, efficient and low-carbon.

Looking Ahead: From Research to Practice

Here are some of the pathways and considerations for turning such research into everyday practice in Oman:
• Data infrastructure: Collecting reliable images of solar panels and electrical measurements in real time requires sensors, cameras or drones, communications infrastructure, and data processing capabilities.
• Agent deployment: Each controllable device (storage, flexible load, PV unit) needs to be “agent-capable” (i.e. controllable, monitored).
• Training and adaptation: The multi-agent system needs training (possibly simulation + live data) and must adapt as devices, loads and renewable resources change.
• Regulatory and operational integration: Utilities and grid operators will need to integrate such AI systems into operations, ensuring safety, reliability and regulatory compliance.
• Local climate and context: Dust accumulation, shading from local geography, and load patterns in Oman must all be factored in.
• Stakeholder engagement: Utility engineers, maintenance staff, renewable project developers, regulators and consumers must all be involved.

Why Readers Should Care

Even if you are not an electrical-engineering specialist, this research is relevant because:
• It shows how advanced AI is being applied to real problems (not only laboratories) in Oman.
• For businesses and industries: it indicates that energy management will become smarter – potentially reducing costs and emissions and increasing reliability.
• For consumers: smarter grids mean fewer outages, better integration of renewables, and possibly lower bills in the long run.
• For policymakers and regulators: it signifies the importance of preparing the grid, infrastructure, regulation and workforce for a more dynamic future.
• For students and innovators: this is a frontier where AI meets power systems – offering opportunities in research, careers and start-ups.

In Summary

The article “Multi-modal multi-task AI model for active distribution network scheduling with multi-agent reinforcement learning” presents an exciting step forward in how we can manage modern, complex electricity networks – especially in regions like Oman which are embracing renewables, storage and smart grid technologies. By combining image-based solar panel data with electrical measurements, and using a cooperative AI-agent framework, the research demonstrates better performance, flexibility and cost-effectiveness in a large distribution network case study.
As Oman continues its energy transition – moving towards more renewable energy, lower emissions and smarter grids – such AI-based frameworks may play a major role in ensuring that our power systems keep up with the changing landscape.