Abstract:This paper presents a novel physics-informed diffusion model for generating synthetic net load data, addressing the challenges of data scarcity and privacy concerns. The proposed framework embeds physical models within denoising networks, offering a versatile approach that can be readily generalized to unforeseen scenarios. A conditional denoising neural network is designed to jointly train the parameters of the transition kernel of the diffusion model and the parameters of the physics-informed function. Utilizing the real-world smart meter data from Pecan Street, we validate the proposed method and conduct a thorough numerical study comparing its performance with state-of-the-art generative models, including generative adversarial networks, variational autoencoders, normalizing flows, and a well calibrated baseline diffusion model. A comprehensive set of evaluation metrics is used to assess the accuracy and diversity of the generated synthetic net load data. The numerical study results demonstrate that the proposed physics-informed diffusion model outperforms state-of-the-art models across all quantitative metrics, yielding at least 20% improvement.

What problem does this paper attempt to address?

### Main Purpose and Problem Addressed by the Paper This paper proposes a new Physics-informed Diffusion Model (PDM) for generating synthetic net load data, aiming to address data scarcity and privacy protection issues. ### Key Features of the Solution - **Embedding Physical Models**: PDM directly embeds physical models into the denoising network, utilizing the Photovoltaic System Performance Model (PVSPM) to enhance the model's interpretability and generalization ability. - **Conditional Denoising Network Design**: The study designs a conditional denoising neural network that jointly trains the transformation kernel parameters of the diffusion model and the physical information function parameters to better leverage the physical model. - **Comprehensive Evaluation Metrics**: The proposed method's effectiveness is validated through a series of comprehensive numerical studies and compared with existing advanced generative models, including Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Normalizing Flows, and a well-calibrated baseline diffusion model. ### Technical Background and Methods - **Advantages of Diffusion Models**: Compared to other generative models such as VAEs, GANs, and Normalizing Flows, diffusion models overcome many key challenges, such as the posterior distribution alignment issue in VAEs and the unstable adversarial objectives in GANs. - **Physics-informed Diffusion Framework**: The paper proposes a diffusion framework that integrates physical models by embedding the basic solar power generation profiles into the denoising network, enabling the model to learn both the transformation kernel parameters and the physical information function parameters simultaneously. - **Design of the Denoising Network**: To capture the complex dynamic characteristics in time series data, a denoising network is designed that includes Long Short-Term Memory (LSTM) units, multi-head self-attention mechanisms, Multi-Layer Perceptrons (MLP), and physical model components. ### Overview of Numerical Studies - **Dataset**: The study uses a real-world smart meter dataset provided by Pecan Street, which includes daily energy consumption, solar photovoltaic generation, and net load time series data for 25 residential users in Austin, Texas, USA. - **Preprocessing**: The data recording interval is 15 minutes, and missing data points are filled using the average net load of the adjacent hour. ### Main Contributions - Proposes a physics-informed diffusion model framework that effectively utilizes physical models to improve the model's interpretability and generalization ability. - Designs two denoising network architectures—baseline denoising network and physics-informed denoising network—that can effectively capture the temporal correlation of net load curves. - Extensive numerical experiments validate that the proposed PDM model outperforms state-of-the-art generative models on all evaluation metrics, with performance improvements of at least 20%. ### Conclusion The method proposed in this paper not only addresses data scarcity and privacy protection issues but also performs excellently in generating synthetic net load data, providing valuable reference data for operational planning tools in power systems.

Generating Synthetic Net Load Data with Physics-informed Diffusion Model

Denoising Diffusion Straightforward Models for Energy Conversion Monitoring Data Imputation

Physics-Informed Diffusion Models

Synthetic Active Distribution System Generation Via Unbalanced Graph Generative Adversarial Network.

Creating synthetic energy meter data using conditional diffusion and building metadata

Denoising diffusion probabilistic models for probabilistic energy forecasting

Customized Load Profiles Synthesis for Electricity Customers Based on Conditional Diffusion Models

Synthetic Data Generation for Residential Load Patterns via Recurrent GAN and Ensemble Method

DiffNILM: A Novel Framework for Non-Intrusive Load Monitoring Based on the Conditional Diffusion Model

DiffCharge: Generating EV Charging Scenarios via a Denoising Diffusion Model

NetDiffus: Network traffic generation by diffusion models through time-series imaging

Text2PDE: Latent Diffusion Models for Accessible Physics Simulation

Towards Generative Modeling of Urban Flow through Knowledge-enhanced Denoising Diffusion

Denoising Diffusion Step-aware Models

Natural generative noise diffusion model imputation

AirPhyNet: Harnessing Physics-Guided Neural Networks for Air Quality Prediction

DiffPLF: A Conditional Diffusion Model for Probabilistic Forecasting of EV Charging Load

Electric load forecasting under false data injection attacks via denoising deep learning and generative adversarial networks

NetDiffusion: Network Data Augmentation Through Protocol-Constrained Traffic Generation

STDEN: Towards Physics-Guided Neural Networks for Traffic Flow Prediction