Forecasting hourly global solar radiation using hybrid k-means and nonlinear autoregressive neural network models

Benmouiza, Khalil; Cheknane, Ali

doi:10.1016/j.enconman.2013.07.003

Cited by 226 publications

(78 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…[18], [19], [20]), it is possible to model more complex time series. There are practical cases where the single linear modeling approach is not enough for systems that present different behaviors along time [10], [11], [12], [13]. These behavior changes might be produced by changes on dynamical regimes.…”

Section: A Backgroundmentioning

confidence: 99%

“…In previous works related to regime recognition, the identification of regimes behaviour has used qualitative information. Benmouiza and Cheknane [10] proposed the implementation of a global NAR (NonLinear Autoregressive) neural network predictor to estimate the regimes associated to another local NAR neural network predictor for the hourly global solar radiation. Kumar and Patel [11] propose a predictive algorithm using data clustering and local training models that combined produce the forecast.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Flow meter data validation and reconstruction using neural networks: Application to the Barcelona water network

Rodríguez

Puig

Flores

et al. 2016

2016 European Control Conference (ECC)

View full text Add to dashboard Cite

Abstract-The use of false or erroneous data can lead to wrong decisions when operating a system. In case of a water distribution network, the use of incorrect data could lead to errors in the billing system, waste of energy, incorrect management of control elements, etc. This paper is focused on detecting flow meters reading abnormalities by exploiting the temporal redundancy of the demand time series by means of artificial neural networks (ANN). Communication problems with the sensor generate missing data and bad maintenance service in the flow meters produce false data. In this work, a methodology to detect the false data (validate) and replace the missing or false data (reconstruct) is proposed. As a core methodology, ANNs are used to model the time series generated from the water demand flow meters, and use the confidence intervals to validate the information. To illustrate the proposed methodology, the application to flow meters in the water distribution network of Barcelona is used.

show abstract

Section: A Backgroundmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Flow meter data validation and reconstruction using neural networks: Application to the Barcelona water network

Rodríguez

Puig

Flores

et al. 2016

2016 European Control Conference (ECC)

View full text Add to dashboard Cite

show abstract

“…With respect to the limitations of the sole prediction models, many scholars have proposed some combination models to fill the gap of each model. Khalil [18] applied a method of the combination of unsupervised k-means clustering algorithm and ANN to forecast hourly global horizontal solar radiation. Xiang et al [19] developed a vector-angle-cosine hybrid model for thermal error prediction.…”

Section: Mathematical Problems In Engineeringmentioning

confidence: 99%

A New Error Prediction Method for Machining Process Based on a Combined Model

Zhou

Zhu

Wang

et al. 2018

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Machining process is characterized by randomness, nonlinearity, and uncertainty, leading to the dynamic changes of machine tool machining errors. In this paper, a novel model combining the data processing merits of metabolic grey model (MGM) with that of nonlinear autoregressive (NAR) neural network is proposed for machining error prediction. The advantages and disadvantages of MGM and NAR neural network are introduced in detail, respectively. The combined model first utilizes MGM to predict the original error data and then uses NAR neural network to forecast the residual series of MGM. An experiment on the spindle machining is carried out, and a series of experimental data is used to validate the prediction performance of the combined model. The comparison of the experiment results indicates that combined model performs better than the individual model. The two-stage prediction of the combined model is characterized by high accuracy, fast speed, and robustness and can be applied to other complex machining error predictions.

show abstract

“…Others investigated implementing forecasting at a real photovoltaic power plant [4]. Recently, hybrid models based on some algorithms have been proposed [5,6]. A comparison of different methods [7], very short-term prediction [8,9], and online prediction [10,11] have also been researched.…”

Section: Introductionmentioning

confidence: 99%

SVM-based Multi-point Prediction System of Solar Irradiance using DTW-based Cluster Analysis

Tanaka

Takahashi

2017

Transactions of ISCIE

View full text Add to dashboard Cite

Unexpected short-term fluctuation of solar irradiance may negatively affect an electric power system; hence, predicting the solar irradiance is necessary. Considering that large quantities of solar power will be introduced across wide areas in the future, this paper addresses multi-point predictions of solar irradiance. Support vector machines (SVMs) perform well for predicting solar irradiance, but involve relatively high computational complexity. In addition, SVMs have to be repeatedly computed at multiple points because prediction models need to be updated regularly to adapt to the season. Here, we introduce a multi-point prediction system that considerably reduces the amount of calculation. Instead of constructing prediction models at all grid points, this system establishes clusters based on similarities of time series data by executing the dynamic time warping algorithm, and then constructs a small number of models representing each cluster. Further, it automatically tunes the number of clusters by inspecting prediction accuracy. Simulation results reveal that the system saves a considerable amount of calculation while maintaining high prediction accuracy.

show abstract

Forecasting hourly global solar radiation using hybrid k-means and nonlinear autoregressive neural network models

Cited by 226 publications

References 30 publications

Flow meter data validation and reconstruction using neural networks: Application to the Barcelona water network

Flow meter data validation and reconstruction using neural networks: Application to the Barcelona water network

A New Error Prediction Method for Machining Process Based on a Combined Model

SVM-based Multi-point Prediction System of Solar Irradiance using DTW-based Cluster Analysis

Contact Info

Product

Resources

About