Structure optimisation of input layer for feed-forward NARX neural network

Li, Zongyan; Best, Matt C.

doi:10.1504/ijmic.2016.075814

Cited by 4 publications

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A thorough survey of literature suggests that for the prediction of time series, dynamical neural networks are most efficient as they can be trained and tuned to predict time-dependent data. Amongst the various developments of dynamical neural networks, the Non-linear AutoRegressive model with exogenous inputs (NARX) neural network has gained great popularity in the research community [154][155][156][157][158][159].…”

Section: Artificial Neural Networkmentioning

confidence: 99%

“…Then the number of neurons in the hidden layer of the neural network were optimized until no further improvement was achieved. In order to set the input and feedback delays, a correlation analysis was performed on the data, and then through a trial and error procedure, the best performing delays were selected for each model [154,157]. It should be noted that artificial neural networks often suffer the two problems of overfitting and premature convergence to local solutions.…”

Section: Artificial Neural Networkmentioning

confidence: 99%

See 1 more Smart Citation

Machine-learning methods for integrated renewable power generation: A comparative study of artificial neural networks, support vector regression, and Gaussian Process Regression

Sharifzadeh

Sikinioti-Lock

Shah

2019

Renewable and Sustainable Energy Reviews

262

View full text Add to dashboard Cite

Renewable energy from wind and solar resources can contribute significantly to the decarbonisation of the conventionally fossil-driven electricity grid. However, their seamless integration with the grid poses significant challenges due to their intermittent generation patterns, which is intensified by the existing uncertainties and fluctuations from the demand side. A resolution is increasing energy storage and standby power generation which results in economic losses. Alternatively, enhancing the predictability of wind and solar energy as well as demand enables replacing such expensive hardware with advanced control and optimization systems. The present research contribution establishes consistent sets of data and develops data-driven models through machine-learning techniques. The aim is to quantify the uncertainties in the electricity grid and examine the predictability of their behaviour. The predictive methods that were selected included conventional artificial neural networks (ANN), support vector regression (SVR) and Gaussian process regression (GPR). For each method, a sensitivity analysis was conducted with the aim of tuning its parameters as optimally as possible. The next step was to train and validate each method with various datasets (wind, solar, demand). Finally, a predictability analysis was performed in order to ascertain how the models would respond when the prediction time horizon increases. All models were found capable of predicting wind and solar power, but only the neural networks were successful for the electricity demand. Considering the dynamics of the electricity grid, it was observed that the prediction process for renewable power and wind was fast and accurate enough to effectively replace the alternative electricity storage and standby capacity.

show abstract

Section: Artificial Neural Networkmentioning

confidence: 99%

Section: Artificial Neural Networkmentioning

confidence: 99%

Machine-learning methods for integrated renewable power generation: A comparative study of artificial neural networks, support vector regression, and Gaussian Process Regression

Sharifzadeh

Sikinioti-Lock

Shah

2019

Renewable and Sustainable Energy Reviews

262

View full text Add to dashboard Cite

show abstract

“…Lagged variables are generated by a number of linear time-delayed input terms or past values, normally in ascending order, such as P(t−n)… P(t−2), P(t−1), to estimate the output value P(t) [34].…”

Section: Lagged Variable Sizementioning

confidence: 99%

Optimizing Predictor Variables in Artificial Neural Networks When Forecasting Raw Material Prices for Energy Production

et al. 2020

View full text Add to dashboard Cite

This paper applies a heuristic approach to optimize the predictor variables in artificial neural networks when forecasting raw material prices for energy production (coking coal, natural gas, crude oil and coal) to achieve a better forecast. Two goals are (1) to determine the optimum number of time-delayed terms or past values forming the lagged variables and (2) to improve the forecast accuracy by adding intrinsic signals to the lagged variables. The conclusions clearly are in opposition to the actual scientific literature: when addressing the lagged variable size, the results do not confirm relationships among their size, representativeness and estimation accuracy. It is also possible to verify an important effect of the results on the lagged variable size. Finally, adding the order in the time series of the lagged variables to form the predictor variables improves the forecast accuracy in most cases.

show abstract

“…ANN can achieve superb performance, eg in [9], but as with most black-box methods, do not give any insight into the virtually unknown model that has been identified. A comparison between black-box and grey-box identification in the automotive field can be found in [10], where Savaresi et al have successfully identified magnetorheological damper models through both a nonlinear semiphysical model and a Nonlinear AutoRegressive eXogenous (NARX) structure.…”

Section: *Corresponding Author: Department Of Aeronautical and Automomentioning

confidence: 99%

“…In recent times, the approach of trying to reproduce the mechanisms of human learning through artificial neural network methods has become increasingly popular. Artificial neural networks can achieve a superb performance (see, for example, the paper by Li and Best 9 ) but, as with most black-box methods, they do not give any insight into the virtually unknown model that has been identified. A comparison between black-box identification and grey-box identification in the automotive field can be found in the work by Savaresi et al, 10 who successfully identified magnetorheological damper models, using both a non-linear semiphysical model and a non-linear autoregressive exogenous (NARX) structure.…”

Section: Introductionmentioning

confidence: 99%

A new structure for non-linear black-box system identification using the extended Kalman filter

Bogdanski

Best

2017

Proceedings of the Institution of Mechanical Engineers, Part D:

View full text Add to dashboard Cite

The method works iteratively in the time domain using an Extended Kalman Filter. The model retains a state space structure in modal canonical form, which ensures that a minimal number of parameters need to be identified and also produces additional information in terms of system eigenvalues and dominant modes. This structure is completely black-box, requiring no physical understanding of the process for successful identification, and it is possible to easily expand the order and complexity of nonlinearities, whilst ensuring good parameter conditioning. A simple nonlinear example illustrates the method, and identification of a highly nonlinear brake model is also presented. These examples show the method can be applied as a mechanism for model order reduction; it is equally well suited as a tool for nonlinear plant system identification. In both capacities this new method is valuable, particularly as the generation of simplified models for the whole vehicle and its subsystems is an increasingly important aspect of modern vehicle design.

show abstract

Structure optimisation of input layer for feed-forward NARX neural network

Cited by 4 publications

References 16 publications

Machine-learning methods for integrated renewable power generation: A comparative study of artificial neural networks, support vector regression, and Gaussian Process Regression

Machine-learning methods for integrated renewable power generation: A comparative study of artificial neural networks, support vector regression, and Gaussian Process Regression

Optimizing Predictor Variables in Artificial Neural Networks When Forecasting Raw Material Prices for Energy Production

A new structure for non-linear black-box system identification using the extended Kalman filter

Contact Info

Product

Resources

About