Optimized LinDistFlow for High-Fidelity Power Flow Modeling of Distribution Networks

The DistFlow model accurately represents power flows in distribution systems, but the model's nonlinearities result in computational challenges for many optimization applications. Accordingly, a linear approximation known as LinDistFlow is commonly employed. This paper introduces an algorithm for enhancing the accuracy of the LinDistFlow approximation, with the goal of aligning the outputs more closely with those from the nonlinear DistFlow model. Using sensitivity information, our algorithm optimizes the LinDistFlow approximation's coefficient and bias parameters to minimize discrepancies in predictions of voltage magnitudes relative to the nonlinear DistFlow model. The algorithm employs the Truncated Newton Conjugate-Gradient (TNC) optimization method to fine-tune coefficients and bias parameters during an offline training phase in order to improve the LinDistFlow approximation's accuracy in optimization applications. This improves the model's effectiveness across various system scenarios, leading to a marked improvement in predictive accuracy. Numerical results underscore the algorithm's efficacy, showcasing accuracy improvements in L₁-norm and L∞-norm losses of up to $92%$ and $88%$, respectively, relative to the traditional LinDistFlow model. We assess how the optimized parameters perform under changes in the network topology and also validate the optimized LinDistFlow approximation's efficacy in a hosting capacity optimization problem.