Probabilistic load forecasting with a non-crossing sparse-group Lasso-quantile regression deep neural network

Lü, Shixiang; Xu, Qifa; Jiang, Cuixia; Liu, Yezheng; Kusiak, Andrew

doi:10.1016/j.energy.2021.122955

Cited by 27 publications

(5 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unlike the approach of incorporating soft constraints on quantile crossing, as demonstrated by Lu et al (2022), we enforce strict monotonicity by placing hard constraints on quantiles. This is achieved by defining higher quantiles as the sum of lower quantiles and non-negative increments.…”

Section: Discussion On Non-crossing Constraintsmentioning

confidence: 99%

“…However, this approach necessitates estimating a separate model for each quantile level, often resulting in quantile-crossing phenomena where higher quantiles are smaller than lower quantiles. For that, Lu et al (2022) have proposed a non-crossing sparse-group Lasso-quantile regressive deep neural network, incorporating constraints on the monotonicity of quantiles to mitigate quantile crossing. Wen et al (2022b) have proposed a continuous and distribution-free probabilistic forecasting approach, capable of predicting the entire distribution at once by transforming the base distribution to the desired one, thereby naturally avoiding quantilecrossing phenomena.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Probabilistic wind power forecasting resilient to missing values: an adaptive quantile regression approach

Wen

2023

Preprint

View full text Add to dashboard Cite

Probabilistic wind power forecasting approaches have significantly advanced in recent decades. However, forecasters often assume data completeness and overlook the challenge of missing values resulting from sensor failures, network congestion, etc. Traditionally, this issue is addressed during the data preprocessing procedure using methods such as deletion and imputation. Nevertheless, these ad-hoc methods pose challenges to probabilistic wind power forecasting at both parameter estimation and operational forecasting stages. In this paper, we propose a resilient probabilistic forecasting approach that smoothly adapts to missingness patterns without requiring preprocessing or retraining. Specifically, we design an adaptive quantile regression model with parameters capable of adapting to missing patterns, comprising two modules. The first is a feature extraction module where weights are kept static and biases are designed as a function of missingness patterns. The second is a non-crossing quantile neural network module, ensuring monotonicity of quantiles, with higher quantiles derived by adding non-negative amounts to lower quantiles. The proposed approach is applicable to both missing-at-random and missing-not-at-random cases. Case studies demonstrate that our proposed approach achieves state-of-the-art results in terms of the continuous ranked probability score, with acceptable computational cost.

show abstract

Section: Discussion On Non-crossing Constraintsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Probabilistic wind power forecasting resilient to missing values: an adaptive quantile regression approach

Wen

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…In recent years, the quantile regression neural networks [22,23] have gained popularity among researchers. Researchers have used them along with modern deep learning architectures for obtaining efficient probabilistic load forecasting [24][25][26][27][28], wind forecasting [29][30][31], and photovoltaic power forecasting [32,33]. + in the feature space for the estimation of the τth quantile, where I :  n o  is a mapping from the input space to a higher dimensional feature space  .…”

Section: Related Workmentioning

confidence: 99%

A ν-Support Vector Quantile Regression Model with Automatic Accuracy Control

Anand

Rastogi

Chandra

2022

Research Reports on Computer Science

View full text Add to dashboard Cite

Quantile regression models have become popular among researchers these days. These models are being used frequently for obtaining the probabilistic forecast in different real-world applications. The Support Vector Quantile Regression (SVQR) model can obtain the conditional quantile estimate using kernel function in a non-parametric framework. The ϵ-SVQR model successfully incorporates the concept of the asymmetric ϵ-insensitive tube in the SVQR model and enables it to obtain a sparse and accurate solution. But, it requires a good choice of the user-defined parameter. A bad choice of ϵ value may result in poor predictions in the ϵ-SVQR model. In this paper, we propose a novel ‘ν-Support Vector Quantile Regression’ (ν-SVQR) model for quantile estimation. It can efficiently obtain a suitable asymmetric ϵ-insensitive zone according to the variance present in the data. The proposed ν-SVQR model uses the ν fraction of training data points for the estimation of the quantiles. In the ν-SVQR model, training points asymptotically appear above and below the asymmetric ϵ-insensitive tube in the ratio of 1 − τ and τ. Apart from these, there are other interesting properties of the proposed ν-SVQR model, which we have briefly described in this paper. These properties have been empirically verified using simulated and real-world data sets also.

show abstract

“…Other studies report that popular technologies in deep learning (such as discard layer, batch training, etc.) can be used to improve quantile regression neural network (QRNN) methods to avoid overfitting Although there has been much research on probabilistic load forecasting [2,3], , no research on the main factors affecting probabilistic load forecasting and future directions has been done. The main contributions of our work are summarized as follows:…”

Section: Introductionmentioning

confidence: 99%

Study on Probabilistic Load Forecasting Model and Its Improvements

Chen

Zhu

Duan

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

The current research on probabilistic load forecasting models mostly combines machine learning algorithms and quantile regression methods to construct quantile models. This paper first summarizes and combs the commonly used quantile regression forecast models. Combined with the actual data, we analyzed the main factors affecting the performance of probabilistic load forecasting and attempted to elaborate on the influencing mechanism. Then, a quantile regression probabilistic load forecasting strategy based on sample weight is designed. According to the analysis and experimental verification, we propose the future research direction for probabilistic load forecasting.

show abstract

Probabilistic load forecasting with a non-crossing sparse-group Lasso-quantile regression deep neural network

Cited by 27 publications

References 53 publications

Probabilistic wind power forecasting resilient to missing values: an adaptive quantile regression approach

Probabilistic wind power forecasting resilient to missing values: an adaptive quantile regression approach

A ν-Support Vector Quantile Regression Model with Automatic Accuracy Control

Study on Probabilistic Load Forecasting Model and Its Improvements

Contact Info

Product

Resources

About