A novel genetic-algorithm-based neural network for short-term load forecasting

Ling, Sai Ho; Leung, F.H.F.; Lam, Hak‐Keung; Lee, Yim-Shu; Tam, P.K.S.

doi:10.1109/tie.2003.814869

Cited by 108 publications

(37 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However, skeptics argue that NNs present practical inefficiencies related to their 'parameter' tuning process and the generalization of their performance. For that reason, researchers apply either novel NN algorithms that try to overcome some of these limitations (Ling et al(2003) or forecast combination techniques that seem able to combine the virtues of different networks for superior forecasts (see amongst others Harrald and Kamstra (1997),…”

Section: Literature Reviewmentioning

confidence: 99%

Stochastic and genetic neural network combinations in trading and hybrid time-varying leverage effects

Sermpinis

Dunis²

2014

Journal of International Financial Markets, Institutions and Mo

View full text Add to dashboard Cite

The motivation of this paper is threefold. Firstly, we apply a Multi-Layer Perceptron (MLP), a Recurrent Neural Network (RNN) and a Psi-Sigma Network (PSN) architecture in a forecasting and trading exercise on the EUR/USD, EUR/GBP and EUR/CHF exchange rates and explore the utility of Kalman Filter, Genetic Programming (GP) and Support Vector Regression (SVR) algorithms as forecasting combination techniques. Secondly, we introduce a hybrid leverage factor based on volatility forecasts and market shocks and study if its application improves the trading performance of our models. Thirdly we introduce a specialized loss function for Neural Networks (NNs) in financial applications. In terms of our results, the PSN from the individual forecasts and the SVR from our forecast combination techniques outperform their benchmarks in statistical accuracy and trading efficiency. We also note that our trading strategy is successful, as it increased the trading performance of most of our models, while our NNs loss function seems promising.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Stochastic and genetic neural network combinations in trading and hybrid time-varying leverage effects

Sermpinis

Dunis²

2014

Journal of International Financial Markets, Institutions and Mo

View full text Add to dashboard Cite

show abstract

“…Besides, some researchers claimed that a uniform auction worsens spot price volatility as compared to a discriminatory auction [7]. So, accuracy of MCP prediction cannot be compared with, for example, load forecast [9,13] or prediction of adequacy indices [14]. Although our obtained error values might seem relatively high, but considering complexities of the MCP signal and forecast errors obtained for it in the other research works (such as those mentioned in this paper), it can be seen that the MCP prediction accuracy of the proposed method is reasonable and promising.…”

Section: Numerical Resultsmentioning

confidence: 99%

“…Although some previous researchers presented hybrid forecast techniques relatively similar to our method (such as combination of GA and NN presented in Reference [9] for load forecast), however combination of the proposed techniques in the framework of the hybrid forecast method and especially its application for MCP prediction can be considered as the contribution of this paper.…”

Section: Introductionmentioning

confidence: 99%

Day‐ahead price forecasting of electricity markets by a hybrid intelligent system

Amjady

Hemmati

2008

Int Trans Elec Energy Syst

View full text Add to dashboard Cite

SUMMARYPrice forecasting is so valuable for both producers and consumers in the new competitive electric power markets. In this paper, a new hybrid intelligent system (HIS) is proposed for day-ahead prediction of market clearing price (MCP) in the pool-based markets. MCP has a volatile and time-dependent behavior owning many outliers. Prediction of such a complex signal is a challenging task requiring a qualified forecast tool, which not only fits well to the training data, but also can predict the stochastic behavior of the unseen part of the signal. The proposed hybrid system is composed of a real-coded genetic algorithm (RCGA), cross-validation, repetitive training, and archiving techniques combined in the framework of a neural network (NN). The proposed RCGA with enhanced stochastic search capability is used for training of the NN while the cross-validation, repetitive training and archiving techniques enhances generalization capability of it. The proposed forecast strategy is examined on the different periods of the Spanish electricity market. It is shown that the method can provide more accurate results than the other price forecasting techniques such as ARIMA time series, wavelet-ARIMA, fuzzy neural network (FNN), and generalized auto-regressive conditional heteroskedastic (GARCH) methods.

show abstract

“…ANN modelling is well established as a tool for electric load forecasting, both as a stand-alone tool (da Silva and Moulin, 2000;Hippert et al, 2001;Mori and Yuihara, 2001;Marin et al, 2002;Taylor and Buizza, 2002;Beccali et al, 2004;Hippert and Pedreira, 2004;Musilek et al, 2006;Hinojosa and Hoese, 2010) and as a hybrid (Khotanzad et al, 1998;Srinivasan et al, 1999;Huang and Yang, 2001;Ling et al, 2003;Chen et al, 2004;Fan and Chen, 2006;Liao and Tsao, 2006;Amjady, 2007;Yun et al, 2008;Bashir and El-Hawary, 2009). While the ANN node-connection diagram is relatively simple (Hsieh, 2009), users might forget that a very large number of free parameters must be determined for ANNs.…”

Section: Introductionmentioning

confidence: 99%

Electric Load Forecasting for Western Canada: A Comparison of Two Non-Linear Methods

Bakhshaii

Stull

2012

Atmosphere-Ocean

View full text Add to dashboard Cite

Seven years of hourly temperature and electric load data for British Columbia in western Canada were used to compare two statistical methods, artificial neural networks (ANN) and gene expression programming (GEP), to produce hour-ahead load forecasts. Two linear control methods (persistence and multiple linear regression) were used for verification purposes. A two-stage (predictor-corrector) approach was used. The first stage used a single regression model that applied weather and calendar data for the previous hour to predict load for any hour of the day, and the second stage applied different corrections every hour, based on a high correlation of today's load with yesterday's load for any given hour. By excluding the day-before variables, the two-stage method reduced the total number of variables in the first-stage regression and gave better results. The first five years of data were used for training (regression finding) and validation (comparative testing of candidate functions to reduce overfitting) and the last two for verification (scoring with independent data).It was found that both non-linear methods worked better than the linear methods for the first stage. All methods worked well for the second stage; hence, persistence is recommended for the second stage because it is easiest to do. After both stages, the load error was less than 0.6% of the load magnitude for hour-ahead forecasts. When used iteratively to create forecasts up to 24 hours ahead, errors grew to about 3.5% of the load magnitude for both GEP and ANNs. We also experimented by training the statistical methods with a shorter period (one year) of past data to examine the over-fitting problem. Overall, ANNs were more useful in fitting the curves to robust data, and GEP was superior for short datasets and was less sensitive to the length of the dataset. We also found that a time-nested electric load forecasting model with different lead times kept maximum load errors to 2.1% of the load magnitude out to a horizon of 24 forecasts.RÉSUMÉ [Traduit par la rédaction] Nous avons utilisé sept années de données de température et de charge électrique pour la Colombie-Britannique dans l'ouest du Canada pour comparer deux méthodes statistiques, les réseaux neuronaux artificiels (ANN) et la programmation génétique (GEP), pour produire des prévisions de charge une heure à l'avance. Nous nous sommes servis de deux méthodes de vérification linéaires (persistance et régression linéaire multiple) à des fins de vérification. Nous avons utilisé une approche en deux étapes (prédicteur-correcteur). Dans la première étape, un modèle de régression simple appliquait des données météorologiques et de calendrier pour l'heure précédente pour prévoir la charge à n'importe quelle heure de la journée et la deuxième étape appliquait à chaque heure différentes corrections basées sur une forte corrélation de la charge d'aujourd'hui avec la charge d'hier pour une heure quelconque. En excluant les variables du jour précédent, la méthode à deux étapes a réduit le nombr...

show abstract

A novel genetic-algorithm-based neural network for short-term load forecasting

Cited by 108 publications

References 18 publications

Stochastic and genetic neural network combinations in trading and hybrid time-varying leverage effects

Stochastic and genetic neural network combinations in trading and hybrid time-varying leverage effects

Day‐ahead price forecasting of electricity markets by a hybrid intelligent system

Electric Load Forecasting for Western Canada: A Comparison of Two Non-Linear Methods

Contact Info

Product

Resources

About