Spike Talk: A Sustainable "One-Solution-Fits-All" Technology for Electronic Grids

Power electronics and digitalization are crucial for the green transition by optimizing energy efficiency and integrating renewable energy sources into the grid. Nevertheless, power electronic grids need enormous technology scaling to reduce energy and carbon footprint without preventing democratized use of artificial intelligence (AI), which has in fact led to a large-scale technology-induced bias and business-propelling optimism that deviates from sustainability. We hereby discuss a novel paradigm inspired by computational neuroscience, Spike Talk, and overview its applications in energy management, where spiking neuron models are leveraged and converters are enabled to “comprehend” the dynamics of grid, as a “one-solution-fits-all” technology making sustainable use of AI and the grid infrastructure. We model the power electronic grid into a single spiking neural network, with the line parameters mapped to the synaptic weights. The operation states of remote converters are essentially interpreted from the networked power flows, as the “language” for agents “talking” with each other. Apart from remarkable decrease in energy consumption, Spike Talk also shows advantages in the capability of Hebbian learning as well as enhanced system resilience. Furthermore, its sparse computational abilities can be further extended to other converter-level applications in power electronics, showing promising prospects in promoting the energy-efficiency of the systems.