Temporal hierarchies with autocorrelation for load forecasting

Nystrup, Peter; Lindström, Erik; Pinson, Pierre; Madsen, Henrik

doi:10.1016/j.ejor.2019.07.061

Cited by 60 publications

(55 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Temporal reconciliation is used for accurate and robust load forecasting through auto and cross-correlation functions. 32 An ordinal regression method with a multi-criteria disaggregation method can be used for long-term demand forecasting, which outperforms the least square regression models and provides high accuracy. 33 ANN can be used to forecast the load demand at multiple time horizons, and ANN performs well in medium-and short-term forecasting with higher accuracy through which end users and system operators get benefited.…”

Section: Load and Demand Forecastingmentioning

confidence: 99%

A holistic review on energy forecasting using big data and deep learning models

Devaraj¹,

Elavarasan

Shafiullah

et al. 2021

Int J Energy Res

View full text Add to dashboard Cite

With the growth of forecasting models, energy forecasting is used for better planning, operation, and management in the electric grid. It is important to improve the accuracy of forecasting for a faster decision-making process. Big data can handle large scale of datasets and extract the patterns fed to the deep learning models that improve the accuracy than the traditional models and hence, recently started its application in energy forecasting. In this study, an in-depth insight is initially derived by investigating artificial intelligence (AI) and machine learning (ML) techniques with their strengths and weaknesses, enhancing the consistency of renewable energy integration and modernizing the overall grid. However, Deep learning (DL) algorithms have the capability to handle big data by capturing the inherent non-linear features through automatic feature extraction methods. Hence, an extensive and exhaustive review of generative, hybrid, and discriminative DL models is being examined for short-term, medium-term, and long-term forecasting of renewable energy, energy consumption, demand, and supply etc. This study also explores the different data decomposition strategies used to build forecasting models. The recent success of DL is being investigated, and the insights of paradoxes in parameter optimization during the training of the model are identified. The impact of weather prediction in the wind and solar energy forecasting is examined in detail. From the existing literatures, it has seen that the average mean absolute percentage error (MAPE) value of solar and wind energy forecasting is 10.29% and 6.7% respectively. Current technology barriers involved in implementing these models for energy forecasting and the recommendations to overcome the existing system barriers are identified. An indepth analysis, discussions of the results, and the scope for improvement are provided in this study including the potential directions for future research in the energy forecasting.

show abstract

Section: Load and Demand Forecastingmentioning

confidence: 99%

A holistic review on energy forecasting using big data and deep learning models

Devaraj¹,

Elavarasan

Shafiullah

et al. 2021

Int J Energy Res

View full text Add to dashboard Cite

show abstract

“…"As the purpose of temporal aggregation is to exploit important information about time series at different frequencies", Nystrup et al (2020) propose other formulations in order to include potential information in the autocorrelation structure. The matrices considered in this paper 7 are:…”

Section: Alternative Approximations Of the Covariance Matrix For Poin...mentioning

confidence: 99%

“…We refer this as optimal combination cross-temporal forecast reconciliation. In addition, grounding on the existing literature on this topic (Wickramasuriya et al, 2019, Athanasopoulos et al, 2017, and Nystrup et al, 2020, we discuss some simple approximations of the covariance matrix to be used in the statistical point forecast reconciliation, with focus on those making use of the in-sample residuals (when available) of the models used to get the base forecasts. The strictly, and very important related issue of probabilistic forecast reconciliation (Gamakumara et al, 2018, Hong et al, 2019, Jeon et al, 2019, Roach, 2019, Ben Taieb et al, 2020 is not considered in this paper, and will be dealt with in the next future.…”

Section: Introductionmentioning

confidence: 99%

Cross-temporal forecast reconciliation: Optimal combination method and heuristic alternatives

Fonzo¹,

Girolimetto²

2020

Preprint

View full text Add to dashboard Cite

Forecast reconciliation is a post-forecasting process aimed to improve the quality of the base forecasts for a system of hierarchical/grouped time series (Hyndman et al., 2011). Contemporaneous (cross-sectional) and temporal hierarchies have been considered in the literature, but -except for Kourentzes and Athanasopoulos (2019) -generally these two features have not fully considered together. Adopting a notation able to simultaneously deal with both forecast reconciliation dimensions, the paper shows two new results: (i) an iterative cross-temporal forecast reconciliation procedure which extends, and overcomes some weaknesses of, the two-step procedure by Kourentzes and Athanasopoulos (2019), and (ii) the closed-form expression of the optimal (in least squares sense) point forecasts which fulfill both contemporaneous and temporal constraints. The feasibility of the proposed procedures, along with first evaluations of their performance as compared to the most performing 'single dimension' (either cross-sectional or temporal) forecast reconciliation procedures, is studied through a forecasting experiment on the 95 quarterly time series of the Australian GDP from Income and Expenditure sides considered by Athanasopoulos et al. (2019).

show abstract

“…A better solution would be to produce forecasts at all hierarchical levels and suitably combine them so that the forecasts across the various aggregation levels are coherent . Combining point forecasts across different cross-sectional [1, 2, 4, 11, 12], temporal [3, 13–19] as well as cross-temporal [20–22] aggregation levels has been extensively studied in the literature. More recently, studies have examined the case of probabilistic and density hierarchical forecasts as opposed to point forecasts [5, 23].…”

Section: Background Researchmentioning

confidence: 99%

Improving the forecasting performance of temporal hierarchies

2019

View full text Add to dashboard Cite

Temporal hierarchies have been widely used during the past few years as they are capable to provide more accurate coherent forecasts at different planning horizons. However, they still display some limitations, being mainly subject to the forecasting methods used for generating the base forecasts and the particularities of the examined series. This paper deals with such limitations by considering three different strategies: (i) combining forecasts of multiple methods, (ii) applying bias adjustments and (iii) selectively implementing temporal hierarchies to avoid seasonal shrinkage. The proposed strategies can be applied either separately or simultaneously, being complements to the method considered for reconciling the base forecasts and completely independent from each other. Their effect is evaluated using the monthly series of the M and M3 competitions. The results are very promising, displaying lots of potential for improving the performance of temporal hierarchies, both in terms of accuracy and bias.

show abstract

Temporal hierarchies with autocorrelation for load forecasting

Cited by 60 publications

References 46 publications

A holistic review on energy forecasting using big data and deep learning models

A holistic review on energy forecasting using big data and deep learning models

Cross-temporal forecast reconciliation: Optimal combination method and heuristic alternatives

Improving the forecasting performance of temporal hierarchies

Contact Info

Product

Resources

About