Reinforcement learning and heuristic based real time power grid management

A system comprising:

• at least one memory storing programmed instructions;
• one or more Input/Output (I/O) interfaces;
• and one or more hardware processors operatively coupled to the at least one memory, wherein the one or more hardware processors are configured by the programmed instructions to:
• receive a real time data pertaining to a power grid network using a real time data acquisition system;
• simulate a maximum line current overflow ratio of the power grid network for a future timestep based on the real time data using a simulation technique;
• compute a critical value of the power grid network for the future timestep based on a comparison between the simulated maximum line current overflow ratio and a predefined safety threshold value, wherein the critical value is set to one if the simulated maximum line current overflow ratio is greater than the predefined safety threshold value and zero otherwise;
• generate a normalized data based on the received real time data using a normalization technique if the critical value is set to one, wherein one of a) line reconnection action and b) grid recovery action is performed if the critical value is set to zero;
• predict a remedial power grid network topology based on the normalized data using a pretrained Reinforcement Learning (RL) technique, wherein the RL technique selects a remedial power grid network topology with a maximum reward;
• recompute the critical value of the power grid network by simulating the remedial power grid network topology using the simulation technique;
• apply a remedial power grid topological reconfiguration action associated with the simulated remedial power grid network in the power grid network in conjunction with a line action if any, only if, a) a recomputed critical value is less than the predefined safety threshold value and b) the predicted remedial power grid network topology is different from a current topology;
• select an optimal heuristic based topology from a plurality of topology heuristics based on the real time data if the critical value is greater than the predefined safety threshold obtained by the pretrained RL technique;
• and
• apply a topological reconfiguration action associated with the optimal heuristic based topology in conjunction with the line action if any only if the corresponding critical value is less than the critical value associated with the remedial power grid network topology predicted by the pretrained RL technique, wherein the remedial power grid network topology predicted by the pretrained RL technique is selected otherwise.

One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors causes:

• receiving a real time data pertaining to a power grid network using a real time data acquisition system;
• simulating a maximum line current overflow ratio of the power grid network for a future timestep based on the real time data using a simulation technique;
• computing a critical value of the power grid network for the future timestep based on a comparison between the simulated maximum line current overflow ratio and a predefined safety threshold value, wherein the critical value is set to one if the simulated maximum line current overflow ratio is greater than the predefined safety threshold value and zero otherwise;
• generating a normalized data based on the received real time data using a normalization technique if the critical value is set to one, wherein one of a) line reconnection action and b) grid recovery action is performed if the critical value is set to zero;
• predicting a remedial power grid network topology based on the normalized data using a pretrained Reinforcement Learning (RL) technique, wherein the RL technique selects a remedial power grid network topology with a maximum reward;
• recomputing the critical value of the power grid network by simulating the remedial power grid network topology using the simulation technique;
• applying a remedial power grid topological reconfiguration action associated with the simulated remedial power grid network in the power grid network in conjunction with a line action if any, only if, a) a recomputed critical value is less than the predefined safety threshold value and b) the predicted remedial power grid network topology is different from a current topology;
• selecting an optimal heuristic based topology from a plurality of topology heuristics based on the real time data if the critical value is greater than the predefined safety threshold obtained by the pretrained RL technique;
• and
• applying a topological reconfiguration action associated with the optimal heuristic based topology in conjunction with the line action if any only if the corresponding critical value is less than the critical value associated with the remedial power grid network topology predicted by the pretrained RL technique, wherein the remedial power grid network topology predicted by the pretrained RL technique is selected otherwise.