The global energy transition is fundamentally a power systems engineering challenge. Replacing dispatchable fossil fuel generation with variable renewable sources — solar and wind — requires transforming grid architecture, control systems, and storage infrastructure in ways that are technically demanding and economically complex.
The Variability Problem
Solar and wind are abundant and clean, but they share a fundamental characteristic that challenges grid operators: output depends on weather, not demand. This variability problem becomes qualitatively more difficult as renewable penetration increases. At 10% penetration, variability can be managed with existing dispatchable generation. At 60–80% penetration — the level required for deep decarbonisation — the grid needs fundamentally different architecture. Three technical responses are emerging: long-duration energy storage, continental-scale transmission interconnection, and demand-side flexibility managed through digital control systems.
Energy Storage: Beyond Lithium-Ion
Lithium-ion batteries have dominated short-duration grid storage and their cost has fallen by 97% over the past decade. But the variability problem of a high-renewable grid is not primarily a daily challenge; it is a seasonal one. Long-duration energy storage (LDES) technologies are at various stages of maturity. Iron-air batteries use the reversible rusting reaction to store energy at estimated capital costs below $20/kWh — approximately one-tenth of lithium-ion. Flow batteries offer almost unlimited duration scalability. Green hydrogen can be stored for months in underground caverns and reconverted to electricity when needed.
Smart Grid Architecture and AI Control
A smart grid is a fundamentally different system architecture characterised by bidirectional power and information flows, distributed decision-making, and real-time optimisation across millions of nodes. Our research group at Tsinghua developed a hierarchical reinforcement learning framework for real-time grid optimisation that coordinates distributed energy resources across three timescales simultaneously: second-by-second frequency regulation, minute-by-minute voltage management, and hour-by-hour economic dispatch. In simulation studies, the RL controller reduced total operating costs by 8.3% compared to conventional model predictive control while maintaining all voltage and frequency constraints within regulatory bounds.
Power Electronics: The Invisible Infrastructure
Power electronics — the technology for converting electrical power between different voltages, frequencies, and forms — is the enabling infrastructure of the energy transition that receives far less public attention than batteries and turbines. Wide-bandgap semiconductors, particularly silicon carbide (SiC) and gallium nitride (GaN), are enabling step-change improvements in power electronics performance. SiC-based inverters are already deployed in electric vehicles by Tesla and BYD; their application to grid-scale power conversion is an active research frontier with significant implications for HVDC transmission efficiency.
Conclusion
The energy transition is achievable with technologies that are either commercially mature or approaching maturity. Building the smart grid infrastructure required for a high-renewable energy system requires policy frameworks that price the value of flexibility, long-duration storage, and transmission capacity. Power systems researchers have both the expertise and the responsibility to engage with these policy dimensions, not merely the technical ones.
