Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Black Box to Clarity: AI-Powered Smart Grid Optimization with Kolmogorov-Arnold Networks

Xiaoting Wang, Yuzhuo Li, Yunwei Li, Gregory Kish

TL;DR

This work tackles the challenge of applying AI to smart grid optimization under uncertainty while preserving interpretability. It introduces a Kolmogorov-Arnold Network (KAN) based framework that centralizes interpretability and learns distributions of grid responses with physically meaningful activations. Through a stochastic OPF case study in a hybrid AC/DC grid, the approach yields accurate PDFs/CDFs for key variables, with means, variances, and confidence intervals that capture rare events. The work is the first energy-system application of KAN, offering a transparent digital replica for planning, forecasting, and monitoring with potential benefits for reliable grid operation.

Abstract

This work is the first to adopt Kolmogorov-Arnold Networks (KAN), a recent breakthrough in artificial intelligence, for smart grid optimizations. To fully leverage KAN's interpretability, a general framework is proposed considering complex uncertainties. The stochastic optimal power flow problem in hybrid AC/DC systems is chosen as a particularly tough case study for demonstrating the effectiveness of this framework.

From Black Box to Clarity: AI-Powered Smart Grid Optimization with Kolmogorov-Arnold Networks

TL;DR

This work tackles the challenge of applying AI to smart grid optimization under uncertainty while preserving interpretability. It introduces a Kolmogorov-Arnold Network (KAN) based framework that centralizes interpretability and learns distributions of grid responses with physically meaningful activations. Through a stochastic OPF case study in a hybrid AC/DC grid, the approach yields accurate PDFs/CDFs for key variables, with means, variances, and confidence intervals that capture rare events. The work is the first energy-system application of KAN, offering a transparent digital replica for planning, forecasting, and monitoring with potential benefits for reliable grid operation.

Abstract

This work is the first to adopt Kolmogorov-Arnold Networks (KAN), a recent breakthrough in artificial intelligence, for smart grid optimizations. To fully leverage KAN's interpretability, a general framework is proposed considering complex uncertainties. The stochastic optimal power flow problem in hybrid AC/DC systems is chosen as a particularly tough case study for demonstrating the effectiveness of this framework.
Paper Structure (6 sections, 1 equation, 5 figures)

This paper contains 6 sections, 1 equation, 5 figures.

Table of Contents

  1. Introduction
  2. Interpretable AI-Powered Smart Grid Optimization
  3. Mathematical Formulation of Smart Grids Optimization
  4. Proposed KAN-based Framework
  5. Case Studies -- Stochastic OPF Application
  6. Conclusions

Figures (5)

  • Figure 1: The KAN-based framework for smart grid optimization tasks li2022smartliu2024kan
  • Figure 2: Hybrid AC/DC system case study. single line diagram of the modified 5-bus test system. The blue lines indicate DC lines Ergun2019.
  • Figure 3: Training and test loss for different sample sizes and KAN configurations. For different KAN configurations, the training sample size is set as 4000.
  • Figure 4: Examples of activation function in for layer $l=0$ before and after training. Blue: before. Red: after.
  • Figure 5: The probability distribution functions (PDFs) and cumulative distribution functions (CDFs) of estimated SOPF solutions: generator output $P_{g_1}$. CI: confidence interval.