Large Foundation Models for Power Systems

Chenghao Huang,Siyang Li,Ruohong Liu,Hao Wang,Yize Chen

2023-12-12

Abstract:Foundation models, such as Large Language Models (LLMs), can respond to a wide range of format-free queries without any task-specific data collection or model training, creating various research and application opportunities for the modeling and operation of large-scale power systems. In this paper, we outline how such large foundation model such as GPT-4 are developed, and discuss how they can be leveraged in challenging power and energy system tasks. We first investigate the potential of existing foundation models by validating their performance on four representative tasks across power system domains, including the optimal power flow (OPF), electric vehicle (EV) scheduling, knowledge retrieval for power engineering technical reports, and situation awareness. Our results indicate strong capabilities of such foundation models on boosting the efficiency and reliability of power system operational pipelines. We also provide suggestions and projections on future deployment of foundation models in power system applications.

Systems and Control,Machine Learning

What problem does this paper attempt to address?

The paper primarily explores how to leverage large foundational models (such as GPT-4 and other large language models) to address various challenging problems in the power system. Specifically, the paper attempts to solve the following key issues:

1. **Optimizing the operation of power systems**: By developing application pipelines, the study investigates how these foundational models can play a role in simplifying Optimal Power Flow (OPF), Electric Vehicle (EV) scheduling, technical report knowledge retrieval, and situational awareness tasks in power systems. For example, it demonstrates how to use large language models as optimizers to solve OPF problems and how to set optimal EV charging plans guided by user preferences.

2. **Enhancing the capabilities of power system operators**: By utilizing the powerful natural language processing capabilities of these models, the paper aims to help operators and decision-makers better understand and manage complex power networks, thereby improving efficiency and reliability. This includes but is not limited to quickly adapting to new optimization problems, understanding technical documents, and handling multimodal data (such as images and text) to achieve situational awareness.

3. **Lowering the barrier to use**: The paper also explores how these models can reduce the need for data collection and model training in power system applications, making it easier for non-experts to interact with power systems and receive instant feedback.

In summary, the goal of the paper is to showcase the potential of these large foundational models in the field of power systems and to present a series of specific case studies to demonstrate their effectiveness and applicability. Additionally, it discusses some limitations of current methods and possible future development directions.