{"record_key":"inductor-free-self-powered-modular-multilevel-converter-with-ai","identifier":"07046","external_key":"07046","title":"Inductor-Free, Self-Powered Modular Multilevel Converter with AI-Optimized Control for Compact, Low-Cost High-Voltage Power Conversion","summary":null,"url":"https://inventions.pitt.edu/technologies/inductor-free-self-powered--07046","json_url":"https://inventions.pitt.edu/technologies/inductor-free-self-powered--07046.json","object_key":"technologies","description":"This invention is a modular multilevel power converter that removes inductors from both its DC side and its submodules, paired with self-powering submodules and a reinforcement learning-trained controller. It aims to make high-voltage and medium-voltage power conversion smaller and simpler by cutting the passive components that dominate converter size and cost.\nDescription\nUnmet Need - Conventional modular multilevel converters depend on bulky inductors and large submodule capacitors that drive up footprint, cost and failure risk. This technology removes the inductors and uses learned control to minimize submodule capacitance.\nThe converter uses a modular multilevel architecture with arms of series-connected submodules, each built from capacitors and power semiconductor devices. It has no inductors on the DC side or within the submodules. Removing inductors normally causes sharp current spikes and larger voltage swings on the submodule capacitors. The technology addresses this in two ways.\nFirst, each submodule can include a spike absorber and energy harvesting circuit. Magnetic cores with coils absorb current spikes, and an electric circuit converts the harvested energy to constant DC. This captures the spike energy and uses it for local power, so the submodule can power itself.\nSecond, a reinforcement learning-based control policy is trained offline and loaded into the controller. The converter has many redundant switching states, and the policy selects among them to minimize submodule capacitance and suppress voltage variation. A reward function based on submodule capacitance and/or voltage ripple guides the training. In a two-layer scheme, a virtual impedance-source control layer sets the overall arm voltage profile and the learned policy chooses the specific switching states that achieve it. Each submodule can behave as a voltage source at low frequency, a current source at mid frequency to limit fault current, and a resistive source at high frequency.\nApplications\n- High-voltage DC (HVDC) transmission converter stations\n- Medium-voltage DC (MVDC) distribution systems\n- Flexible AC transmission systems (FACTS)\n- Medium- and high-voltage motor drives\n- Renewable energy grid interconnection and power conversion\nAdvantages\n- Inductor-free converter architecture on both the DC side and in the submodules\n- Self-powering submodules that harvest spike energy for local power\n- Reinforcement learning-optimized switching that minimizes submodule capacitance and suppresses voltage variation\n- Layered control that combines virtual impedance-source control with learned submodule-level selection\n- Fault-transient mitigation and fault-current limiting through frequency-dependent submodule control\nInvention Readiness\nThe technology is at the simulation and concept-validation stage. Simulation results show the inductor-free converter with spike absorption and energy harvesting suppressing arm current ripple. Further work is needed to build and test a hardware prototype of the submodule and converter, train and validate the reinforcement learning policy against the full converter model, and demonstrate performance at scale in a relevant operating environment.\nIP Status\nPatent Pending","attributes":[{"key":"file_number","label":"Reference Number","values":["07046"]},{"key":"technology_type","label":"Technology Type","values":["Engineering Technology"]},{"key":"technology_subtype","label":"Technology Subtype","values":["Energy & Power"]},{"key":"tags","label":"Tags","values":["Artificial intelligence (AI)","Machine learning","Sustainability"]},{"key":"lead_inventor_name","label":"Lead Inventor","values":["Fang Peng"]},{"key":"lead_inventor_dept","label":"Department","values":["Electrical and Computer Engineering"]},{"key":"all_tech_innovators","label":"All Innovators","values":["Brandon M. Grainger","Yuan Li","Fang Zheng Peng"]},{"key":"technology_readiness_level","label":"Technology Readiness Level","values":["3. Prototype development"]},{"key":"date_submitted","label":"Date Submitted","values":["Jan 23, 2025"]}],"related":[]}