Research Projects
Scenario Forecasting of Residential Load Profiles
We proposed a generative learning approach to model residential customers' energy consumption behavior based on historic load profiles from ERCOT using the flow-based conditional Generative Neural Network (GNN). This approach not only offers informed future forecasts but also serves as a valuable supplement for inadequate measured meter data.
​
A Convex Neural Network Solver for DCOPF with Generalization Guarantees
We proposed a neural network-based algorithm for solving DCOPF that guarantees the generalization performance. First, by utilizing the convexity of the DCOPF problem, we train an input convex neural network. Second, we construct the training loss based on Karush–Kuhn–Tucker optimality conditions. By combining these two techniques, the trained model has provable generalization properties.
​
An Efficient Learning-Based Solver for Two-Stage DC Optimal Power Flow with Feasibility
We proposed a learning method to solve the two-stage problem in a more efficient and optimal way. A technique called the gauge map is incorporated into the learning architecture design to guarantee the learned solutions’ feasibility to the network constraints.
​
An Iterative Approach to Improving Solution Quality for AC Optimal Power Flow Problems
A widely used strategy to overcome nonconvexity of ACOPF is finding a good initial point for iterative algorithms or nonlinear programming local solvers. We developed both iterative and NN-based algorithms to generate warm starts that geometrically guarantee ACOPF solvers escape from local optima. The key observation is the partial Lagrangian function, which dualizes nonlinear equality constraints, has a flatter landscape. Using its optimal solution as an initial point, the solver is more likely to escape local minima and converge toward a global optimum or at least improve the solution’s optimality.
​
Read more on the iterative algorithm
​
Read more on the learning algorithm
​
Convex Restriction of Feasible Sets for AC Radial Networks
We proposed an analytical method to construct the convex restriction of the feasible set for AC power flows in radial networks. The construction relies on simple geometrical ideas and is explicit, in the sense that it does not involve solving other complicated optimization problems. We also show that the construct restrictions are in some sense maximal, that is, the best possible ones.
​