Interpretable deep learning models for hourly solar radiation prediction based on graph neural network and attention

Gao, Yuan; Miyata, Shohei; Akashi, Yasunori

doi:10.1016/j.apenergy.2022.119288

Cited by 35 publications

(5 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DNNPerf 22 uses the Direct Acyclic Graph (DAG) representation of the neural network and execution platform specifications to predict the execution time and GPU memory consumption of DL models. The DAG is automatically created by a parser, and features describing the model and the execution platform are encoded into its nodes and edges by an attention‐based node‐edge encoder.…”

Section: Related Workmentioning

confidence: 99%

A lightweight performance proxy for deep‐learning model training on Amazon SageMaker

Keller Tesser,

Marques,

Borin

2024

Concurrency and Computation

View full text Add to dashboard Cite

SummaryCloud computing has become popular for training deep‐learning (DL) models, avoiding the costs of acquiring and maintaining on‐premise systems. SageMaker is a cloud service that automates the execution of DL workloads. Its features include automatic hyperparameter optimization and use of spot instances. Nonetheless, it does not assist in selecting the right instance type for a workload. In public clouds, rent price depends on the configuration of the chosen instance type. Advanced and faster instances are typically more expensive, but not always the best choice. To select the optimal instance type, users must compare the workload's relative performance (and hence cost) on several candidates. Building on the execution profiles of multiple DL applications, we model the performance and cost of training DL applications on SageMaker and propose a lightweight technique to estimate these at low temporal and monetary cost. This method is a performance proxy that can be used to replace more expensive performance measurement procedures. So, it could speed up any technique that relies on such measurements. We show how it can help cloud customers seeking suitable instance types to train DL models, and that it can accurately predict the performance of different instance types when training these models on SageMaker.

show abstract

Section: Related Workmentioning

confidence: 99%

A lightweight performance proxy for deep‐learning model training on Amazon SageMaker

Keller Tesser,

Marques,

Borin

2024

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…Solar energy is a crucial and well-established form of renewable energy. As a result, its integration into building energy systems has become an increasingly vital strategy for reducing building energy consumption (Gao et al, 2022b). However, the integration of renewable energy sources into building energy systems has presented significant challenges in terms of control and optimization.…”

Section: Introductionmentioning

confidence: 99%

Multi-step solar radiation prediction using transformer: A case study from solar radiation data in Tokyo

Dong,

Tang,

et al. 2024

Journal of Building Physics

View full text Add to dashboard Cite

The widespread advancement of computer technology resulted in the increasing usage of deep learning models for predicting solar radiation. Numerous studies have been conducted to explore their research potential. Nevertheless, the application of deep learning models in optimizing building energy systems, particularly in a multi-step solar radiation prediction model for model predictive control (MPC), remains a challenging task. This is mainly due to the intricacy of the time series and the possibility of accumulating errors in multistep forecasts. In this study, we propose the development of a transformer-based attention model for predicting multi-step solar irradiation at least 24 h in advance. The model is trained and tested using measured solar irradiation data and temperature forecast data obtained from the Tokyo Meteorological Agency. The findings indicate that the transformer model has the capability to effectively mitigate the issue of error accumulation. Additionally, the generative model exhibits a significant improvement in accuracy, with a 62.35% increase when compared to the conventional regression LSTM model. Additionally, the transformer model has been shown to attain superior prediction stability, mitigate the effects of error accumulation in multi-step forecasting, and circumvent training challenges stemming from gradient propagation issues that can occur with recurrent neural networks.

show abstract

“…There are various studies on solar and wind energy sources, which are the two leading elements of renewable energy sources. When the studies on solar power plants are examined, it is seen that early studies of prediction associated with solar power are made with artificial neural networks (ANN) before the current methods of deep learning [4]. While one of these studies takes meteorological data as input [5], another study has established prediction structures that support the ANN structure with optimization techniques [6].…”

Section: Introductionmentioning

confidence: 99%

Comprehensive Assessment and Comparative Analysis of Deep Learning Models for Large-Scale Renewable Energy Power Generation Prediction: A National Perspective

Aksoy,

Genc

2024

Preprint

View full text Add to dashboard Cite

In forecasting the future energy consumption and generation at the national level, strategic planning for both the medium and long term becomes imperative. The trajectory of renewable energy contribution to the smart grid, whether in the short or long term, significantly influences the grid's operational dynamics. This study is dedicated to the development of deep learning-based power prediction models tailored for a nation characterized by extensive reliance on renewable energy sources. Specifically, four distinct deep learning methodologies—namely,Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), and Gated Recurrent Unit (GRU)—have been formulated and rigorously examined for their efficacy within this domain. These models have been individually tailored for the prediction of power generation from solar photovoltaic plants and wind turbines, leveraging the inherent advantages of architectures featuring memory cells. The outcomes of these predictive models, which encompass the entire spectrum of renewable energy sources, exhibit remarkable precision. Furthermore, an exhaustive analysis of the performance metrics derived from these models has been conducted, affording a comprehensive and nuanced comparison. The findings contribute valuable insights into the suitability and effectiveness of the aforementioned deep learning methodologies in forecasting power generation from renewable sources at a national scale.

show abstract

Interpretable deep learning models for hourly solar radiation prediction based on graph neural network and attention

Cited by 35 publications

References 32 publications

A lightweight performance proxy for deep‐learning model training on Amazon SageMaker

A lightweight performance proxy for deep‐learning model training on Amazon SageMaker

Multi-step solar radiation prediction using transformer: A case study from solar radiation data in Tokyo

Comprehensive Assessment and Comparative Analysis of Deep Learning Models for Large-Scale Renewable Energy Power Generation Prediction: A National Perspective

Contact Info

Product

Resources

About