Search

As global warming intensifies, the reduction of carbon emissions is imminent. Carbon price is directly related to whether carbon can be effectively reduced. Therefore, accurately forecasting carbon price has important practical significance. Aiming at the nonstationary and nonlinear characteristics of carbon price, this paper proposes a novel hybrid model for forecasting carbon price, which is based on improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), multiscale fuzzy entropy (MFE), complete ensemble empirical mode decomposition (CEEMD), improved random forest by salp swarm algorithm (SSARF), improved back propagation by cuckoo search (CSBP), improved extreme learning machine by whale optimization algorithm (WOAELM) and error correction (EC), named ICEEMDAN-MFE-CEEMD-SSARF-CSBP-WOAELM-EC. Firstly, carbon price is decomposed by ICEEMDAN, divided into high-, medium-, and low-complexity components by MFE. Secondly, high-complexity components are merged and secondarily decomposed by CEEMD, which are still recorded as high-complexity components. Then, SSARF, CSBP and WOAELM are used to forecast high-, medium-, and low-complexity components, respectively, and forecasting results are reconstructed. Finally, EC is carried out using an extreme learning machine to obtain the final forecasting results, and the Diebold-Mariano test is introduced for a comprehensive evaluation of the model. Taking carbon price in the pilot cities of Shenzhen and Hubei as examples, after 6 aspects and 20 comparative experiments, the results show that the proposed model has higher forecast accuracy, with MAPE, MAE and RMSE up to 0.03131, 0.00089 and 4.02e-06 in Hubei, and its forecasting ability is better than other commonly used international carbon financial price forecasting models, providing a theoretical and data basis for carbon pricing and formulating carbon reduction policies in China. The main contributions of this paper are the improved primary decomposition, the use of secondary decomposition, and the innovative combination of three optimal models to forecast carbon price, but it still needs to be optimized for practice. [Display omitted] • Propose improve CEEMDAN by using Hermite interpolation and calculating residuals first. • High complexity components are secondarily decomposed to reduce its complexity. • Improved RF by SSA, named SSARF, is proposed to solve the dilemmas of RF. • Propose a high-accuracy hybrid model for carbon price forecasting. • Compare multi-faceted experiments to prove the validity of the proposed model. [ABSTRACT FROM AUTHOR]

A major difficulty in integrating large scale wind power generation in an electrical power system is that wind generated power appears to be erratic, intermittent, and volatile. In this paper, we demonstrate the efficacies of a novel ultra short term 1-step ahead wind generated power prediction model, by combining two best of breed machine learning models in their respective areas of applications: a deep convolutional neural network (CNN) model, known to be effective in classification problems, and a bi-directional long short term memory (Bi-LSTM) model, known to be effective in 1-step ahead time series prediction problems, using a cross attention (XA) mechanism on three challenging practical datasets: the East-China dataset, the Yalova (Turkey) dataset, and the 16 MW dataset. There are two alternative cross attention models: (1) using the CNN features as the key–value pair, and the Bi-LSTM features as the query, this we called a XA model, and (2) using the Bi-LSTM features as the key–value pair, and the CNN features as the query, this we called a XA_I model; we found empirically for a properly chosen sliding window length, whichever CNN model or Bi-LSTM model performs well in the 1-step ahead prediction, then its features should be used as the query in the cross attention mechanism, thus, the XA achieves better results than XA_I on the East-China dataset, while XA_I achieves better results than XA on the Yalova dataset, and on the 16 MW dataset. Where comparable results using other state-of-the-art models are available on the Yalova dataset and the 16 MW dataset, we found that our XA_I results are superior. Moreover, it is found empirically, that the cross attention mechanism model produces stable k -step ahead predictions, for k ≤ W 2 , where W is the sliding window length. • A novel prediction model for wind generated power based on cross attention mechanism has been proposed. • The proposed method demonstrated on three practical datasets to have superior performance. • The cross attention mechanism based models give a stable multi-step ahead prediction. [ABSTRACT FROM AUTHOR]

This paper presents a hybrid machine learning model for predicting the pullback force in horizontal directional drilling (HDD) construction. The model combines the nondominated sorting genetic algorithm II (NSGA-II) and support vector machine (SVM). NSGA-II is used to optimize two hyperparameters in SVM. Different from other optimization algorithms, NSGA-II is a multi-objective optimizer, which considers prediction accuracy and stability. The proposed model is applied to two practical HDD projects in China. The prediction result is compared with the actual monitoring data, which shows that the mean absolute percentage errors (MAPE) are less than 7%. The primary conclusions are as follows: (1) The proposed model's accuracy and stability are better than those of the two benchmark models; (2) Machine learning models can predict the pullback force more accurately than the empirical model in the construction phase, and the maximum MAPE does not exceed 17%; (3) The running time of the proposed model is short, and it is feasible in practical application. Moreover, this paper discusses the practical application of machine learning models in HDD construction and the future development direction. • A hybrid machine learning model is proposed to predict pullback force during HDD construction. • Prediction accuracy and stability are considered. • Proposed model is applied to two HDD projects in China. [ABSTRACT FROM AUTHOR]

Integrating return prediction of traditional time series models in portfolio formation can improve the performance of original portfolio optimization model. Since machine learning and deep learning models have shown overwhelming superiority than time series models, this paper combines return prediction in portfolio formation with two machine learning models, i.e., random forest (RF) and support vector regression (SVR), and three deep learning models, i.e., LSTM neural network, deep multilayer perceptron (DMLP) and convolutional neural network. To be specific, this paper first applies these prediction models for stock preselection before portfolio formation. Then, this paper incorporates their predictive results in advancing mean–variance (MV) and omega portfolio optimization models. In order to present the superiority of these models, portfolio models with autoregressive integrated moving average's return prediction are used as benchmarks. Evaluation is based on historical data of 9 years from 2007 to 2015 of component stocks of China securities 100 index. Experimental results show that MV and omega models with RF return prediction, i.e., RF+MVF and RF+OF, outperform the other models. Further, RF+MVF is superior to RF+OF. Due to the high turnover of these two models, this paper discusses their performance after deducting the transaction fee cased by turnover. Experiments present that RF+MVF still performs the best among MVF models and omega model with SVR prediction (SVR+OF) performs the best among OF models. Moreover, RF+MVF performs better than SVR+OF and high turnover erodes nearly half of their total returns especially for RF+OF and RF+MVF. Therefore, this paper recommends investors to build MVF with RF return prediction for daily trading investment. • Compares the performance of machine learning and deep learning in stock preselection. • Combining return prediction of machine learning and deep learning in portfolio formation. • Emphasis on advancing portfolio optimization with return prediction. • Advanced mean–variance model with random forest forecasts performs the best. [ABSTRACT FROM AUTHOR]

Motivated by reducing carbon emissions, carbon trading market have been opened to promote environmental protection. Accurate carbon trading volume and price forecasts have far-reaching implications for environmental and energy policy formulation. As the country with the most massive carbon emissions in the world, China's carbon price trends, carbon trading volume trends, and policy orientation are of great significance. As matters stand, there are few studies on carbon price and trading volume forecasts, and no research paper has covered carbon price and trading volume forecasts for all carbon trading markets in China. In order to provide a relatively complete reference for the carbon price and trading volume in China, this paper uses six machine learning models to predict the daily carbon price and trading volume of eight carbon markets in China, including Beijing, Shenzhen, Guangdong, Hubei, Shanghai, Fujian, Tianjin, Chongqing. The prediction models include extreme gradient boosting, random forest, kernel-based nonlinear extension of the Arps decline model optimized by grey wolf optimizer (GWO-KNEA), support vector machine optimized by particle swarm optimizer, support vector machine optimized by fruit fly optimizer and simulated annealing algorithm, and radial basis function neural network (RBFNN). Moreover, an advanced data denoising method, i.e., complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), is used in the models to smooth the raw data. Because the eight carbon trading markets have different carbon price and trading volume distribution characteristics in time series, by comparing the prediction results, more suitable models for the specific carbon market can be obtained. The prediction results show that CEEMDAN-GWO-KNEA and CEEMDAN-RBFNN are more competitive in predicting carbon prices and trading volume in China because of their best performance in many datasets. The average accuracy of CEEMDAN-RBFNN and CEMDAN-GWO-KNEA in carbon price prediction is 98.40% and 97.89%, respectively. The predictive stability of each model in different application scenarios is also discussed. The results show that high prediction accuracy does not mean better prediction stability, but CEEMDAN-RBFNN and CEMDAN-GWO-KNEA can still guarantee high stability in most datasets. [ABSTRACT FROM AUTHOR]