Xueheng Qiu

Ensemble deep learning for regression and time series forecasting

In this paper, for the first time, an ensemble of deep learning belief networks (DBN) is proposed for regression and time series forecasting. Another novel contribution is to aggregate the outputs from various DBNs by a support vector regression (SVR) model. We show the advantage of the proposed method on three electricity load demand datasets, one artificial time series dataset and three regression datasets over other benchmark methods.

Empirical Mode Decomposition based ensemble deep learning for load demand time series forecasting

Graphical abstractDisplay Omitted HighlightsAn ensemble deep learning method has been proposed for load demand forecasting.The hybrid method composes of Empirical Mode Decomposition and Deep Belief Network.Empirical Mode Decomposition based methods outperform the single structure models.Deep learning shows more advantages when the forecasting horizon increases. Load demand forecasting is a critical process in the planning of electric utilities. An ensemble method composed of Empirical Mode Decomposition (EMD) algorithm and deep learning approach is presented in this work. For this purpose, the load demand series were first decomposed into several intrinsic mode functions (IMFs). Then a Deep Belief Network (DBN) including two restricted Boltzmann machines (RBMs) was used to model each of the extracted IMFs, so that the tendencies of these IMFs can be accurately predicted. Finally, the prediction results of all IMFs can be combined by either unbiased or weighted summation to obtain an aggregated output for load demand. The electricity load demand data sets from Australian Energy Market Operator (AEMO) are used to test the effectiveness of the proposed EMD-based DBN approach. Simulation results demonstrated attractiveness of the proposed method compared with nine forecasting methods.

Empirical mode decomposition based adaboost-backpropagation neural network method for wind speed forecasting

Wind speed forecasting is a popular research direction in renewable energy and computational intelligence. Ensemble forecasting and hybrid forecasting models are widely used in wind speed forecasting. This paper proposes a novel ensemble forecasting model by combining Empirical mode decomposition (EMD), Adaptive boosting (AdaBoost) and Backpropagation Neural Network (BPNN) together. The proposed model is compared with six benchmark models: persistent, AdaBoost with regression tree, BPNN, AdaBoost-BPNN, EMD-BPNN and EMD-AdaBoost with regression tree. The comparisons undergoes several statistical tests and the tests show that the proposed EMD-AdaBoost- BPNN model outperformed the other models significantly. The forecasting error of the proposed model also shows significant randomness.

Oblique random forest ensemble via Least Square Estimation for time series forecasting

Abstract Recent studies in Machine Learning indicates that the classifiers most likely to be the bests are the random forests. As an ensemble classifier, random forest combines multiple decision trees to significant decrease the overall variances. Conventional random forest employs orthogonal decision tree which selects one “optimal” feature to split the data instances within a non-leaf node according to impurity criteria such as Gini impurity, information gain and so on. However, orthogonal decision tree may fail to capture the geometrical structure of the data samples. Motivated by this, we make the first attempt to study the oblique random forest in the context of time series forecasting. In each node of the decision tree, instead of the single “optimal” feature based orthogonal classification algorithms used by standard random forest, a least square classifier is employed to perform partition. The proposed method is advantageous with respect to both efficiency and accuracy. We empirically evaluate the proposed method on eight generic time series datasets and five electricity load demand time series datasets from the Australian Energy Market Operator and compare with several other benchmark methods.

Electricity load demand time series forecasting with Empirical Mode Decomposition based Random Vector Functional Link network

Short-term electricity load demand forecasting is a critical process in the management of modern power system. An ensemble method composed of Empirical Mode Decomposition (EMD) and Random Vector Functional Link network (RVFL) is presented in this paper. Due to the randomly generated weights between input and hidden layers and the close form solution for parameter tuning, RVFL network is a universal approximator with the advantages of fast training. By introducing ensemble approach via EMD into RVFL network, the performance can be significantly improved. Five electricity load demand datasets from Australian Energy Market Operator (AEMO) were used to evaluate the performance of the proposed method. The attractiveness of the proposed EMD based RVFL network can be demonstrated by the comparison with six benchmark methods.

Short-term Electricity Price Forecasting with Empirical Mode Decomposition based Ensemble Kernel Machines

Short-term electricity price forecasting is a critical issue for the operation of both electricity markets and power systems. An ensemble method composed of Empirical Mode Decomposition (EMD), Kernel Ridge Regression (KRR) and Support Vector Regression (SVR) is presented in this paper. For this purpose, the electricity price signal was first decomposed into several intrinsic mode functions (IMFs) by EMD, followed by a KRR which was used to model each extracted IMF and predict the tendencies. Finally, the prediction results of all IMFs were combined by an SVR to obtain an aggregated output for electricity price. The electricity price datasets from Australian Energy Market Operator (AEMO) are used to test the effectiveness of the proposed EMD-KRR-SVR approach. Simulation results demonstrated attractiveness of the proposed method based on both accuracy and efficiency.

Detecting Wind Power Ramp with Random Vector Functional Link (RVFL) Network

Due to the intermittent nature of the wind, the wind speed is fluctuating. Fluctuating wind speed cause even more fluctuation in wind power generation. The sudden changes of the wind power injected into the power grid within a short time frame is known as power ramp, which can be harmful to the grid. This paper presents algorithms to detect the wind power ramps in a certain forecasting horizon. The importance and challenges of wind power ramp detection are addressed. Several different Wind power ramps are defined in this paper. A random vector functional link (RVFL) network is employed to predict the future occurrence of wind power ramp. The forecasting methods are evaluated with a real world wind power data set. The RVFL network has comparable performance as the benchmark methods: random forests (RF) and support vector machine (SVM) but it has better performance than the artificial neural network (ANN). The computation time of training and testing is also in favor of the RVFL network.

Ensemble incremental learning Random Vector Functional Link network for short-term electric load forecasting

Abstract Short-term electric load forecasting plays an important role in the management of modern power systems. Improving the accuracy and efficiency of electric load forecasting can help power utilities design reasonable operational planning which will lead to the improvement of economic and social benefits of the systems. A hybrid incremental learning approach composed of Discrete Wavelet Transform (DWT), Empirical Mode Decomposition (EMD) and Random Vector Functional Link network (RVFL) is presented in this work. RVFL network is a universal approximator with good efficiency because of the randomly generated weights between input and hidden layers and the close form solution for parameter computation. By introducing incremental learning, along with ensemble approach via DWT and EMD into RVFL network, the forecasting performance can be significantly improved with respect to both efficiency and accuracy. The electric load datasets from Australian Energy Market Operator (AEMO) were used to evaluate the effectiveness of the proposed incremental DWT-EMD based RVFL network. Moreover, the attractiveness of the proposed method can be demonstrated by the comparison with eight benchmark forecasting methods.

Heterogeneous ensemble for power load demand forecasting

Electricity load demand is the fundamental building block for all utilities planning. The load demand data has nonlinear and non-stationary characteristics, which make it difficult to be predicted accurately by just computational intelligence or ensemble methods. Ensemble methods like Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is a powerful tool to forecast power load demand time series. Heterogeneous ensemble, a combination of two base models, will be distinct or more powerful in forecasting power load demand. In this paper, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) method is hybridized with three computational intelligence-based predictors: support vector regression (SVR), artificial neural network (ANN) and random forest (RF). The basis of this paper was to conduct a comparative study on the accuracy of the forecasting result from using heterogeneous ensemble method to individual computational intelligence or ensemble method for four different horizons. The performances of the heterogeneous method are compared and discussed. It shows that heterogeneous method has outperformed the individual computational intelligence and ensemble methods. Possible future works are also recommended for power load demand forecasting.

Stock Price Forecasting with Empirical Mode Decomposition Based Ensemble \(\nu \)-Support Vector Regression Model

Stock price forecasting is one of the most challenging tasks of time series forecasting due to the inherent non-linearity and non-stationary characteristics of the stock market and financial time series. In this paper, an ensemble method composed of Empirical Mode Decomposition (EMD) algorithm and \(\nu \)-Support Vector Regression (\(\nu \)-SVR) is presented for short-term stock price forecasting. First of all, the historical stock price time series were decomposed into several intrinsic mode functions (IMFs). Then each IMF was modeled by a \(\nu \)-SVR model to generate the corresponding forecasting IMF value. Finally, the prediction results of all IMFs were combined to formulate an aggregated output for stock price. The stock market price datasets of three power related companies are used to test the effectiveness of the proposed EMD-\(\nu \)-SVR method. Simulation results demonstrated attractiveness of the proposed method compared with six forecasting methods.