Uncertainty prediction of wind speed based on improved multi-strategy hybrid models

Karijadi I., Chou S., Dewabharata A.

A precise wind power forecast is required for the renewable energy platform to function effectively. By having a precise wind power forecast, the power system can better manage its supply and ensure grid reliability. However, the nature of wind power generation is intermittent and exhibits high randomness, which poses a challenge to obtaining accurate forecasting results. In this study, a hybrid method is proposed based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Empirical Wavelet Transform (EWT), and deep learning-based Long Short-Term Memory (LSTM) for ultra-short-term wind power forecasting. A combination of CEEMDAN and EWT is used as the preprocessing technique, where CEEMDAN is first employed to decompose the original wind power data into several subseries, and the EWT denoising technique is used to denoise the highest frequency series generated from CEEMDAN. Then, LSTM is utilized to forecast all the subseries from the CEEMDAN-EWT process, and the forecasting results of each subseries are aggregated to achieve the final forecasting results. The proposed method is validated on real-world wind power data in France and Turkey. Our experimental results demonstrate that the proposed method can forecast more accurately than the benchmarking methods.

Abdel-Basset M., Mohamed R., Jameel M., Abouhawwash M.

This work presents a novel nature-inspired metaheuristic called Nutcracker Optimization Algorithm (NOA) inspired by Clark’s nutcrackers. The nutcrackers exhibit two distinct behaviors that occur at separate periods. The first behavior, which occurs during the summer and fall seasons, represents the nutcracker’s search for seeds and subsequent storage in an appropriate cache. During the winter and spring seasons, another behavior based on the spatial memory strategy is regarded to search for the hidden caches marked at different angles using various objects or markers as reference points. If the nutcrackers cannot find the stored seeds, they will randomly explore the search space to find their food. NOA is herein proposed to mimic these various behaviors to present a new, robust metaheuristic algorithm with different local and global search operators, allowing it to solve various optimization problems with better outcomes. NOA is evaluated on twenty-three standard test functions, test suites of CEC-2014, CEC-2017, and CEC-2020 and five real-world engineering design problems. NOA is compared with three classes of existing optimization algorithms: (1) SMA, GBO, EO, RUN, AVOA, RFO, and GTO as recently-published algorithms, (2) SSA, WOA, and GWO as highly-cited algorithms, and (3) AL-SHADE, L-SHADE, LSHADE-cnEpSin, and LSHADE-SPACMA as highly-performing optimizers and winners of CEC competition. NOA was ranked first among all methods and demonstrated superior results when compared to LSHADE-cnEpSin and LSHADE-SPACMA as the best-performing optimizers and the winners of CEC-2017, and AL-SHADE and L-SHADE as the winners of CEC-2014.

Hu C., Xing F., Pan S., Yuan R., Lv Y.

Fault diagnosis of rolling bearings can be a serious challenge, as rolling bearings often work under complex conditions and their vibration signals are typically nonlinear and nonstationary. This paper proposes a novel approach to diagnosing faults of rolling bearings based on variational mode decomposition (VMD) and genetic algorithm-optimized wavelet threshold denoising. First, VMD was used to decompose the vibration signals of faulty rolling bearings into a series of band-limited intrinsic mode functions (BLIMFs). During the decomposition, the parameters of VMD were selected by Kullback–Leibler (K–L) divergence. Then, the effective BLIMFs were determined by the analysis of their correlation coefficients and variance contributions. Finally, genetic algorithm-optimized wavelet threshold denoising was proposed to optimize the selection of important parameters, and the optimized threshold function used not only ensures the continuity of the threshold function but also avoids the fixed deviation of the soft threshold. The validity and superiority of the proposed approach were verified by theoretical calculations, numerical simulations and application studies. The results indicate that the proposed approach is promising in fault diagnosis of rotary machinery.

Zhang K., Yu X., Liu S., Dong X., Li D., Zang H., Xu R.

In today’s increasingly serious world energy crisis, Renewable energy such as wind energy has gradually penetrated into life. Aiming at the uncertainty of wind power and the need of a mass of sample data in nonparametric kernel density estimation, a wind power interval prediction method based on hybrid semi-cloud model and nonparametric kernel density estimation is proposed. Firstly, aiming at the asymmetrical distribution of prediction error, a qualitative description method of error concept using hybrid semi-cloud model is proposed. And the conceptual data produced by the hybrid semi-cloud model are filtered through the “3En rule”. Secondly, according to the fitting characteristics of nonparametric fitting, the conceptual cloud droplets are combined with nonparametric kernel density estimation to get the shortest prediction interval under different confidence levels. Finally, in order to verify the effectiveness of the method, the simulation of a wind farm data in the Oklahoma State is carried out, and the relative entropy is introduced to evaluate the fitting effect of the nonparametric fitting method. The comprehensive performance of the prediction interval under different methods is compared by three evaluation indexes: interval coverage probability, prediction interval width and comprehensive evaluation index F value. The experimental results show that compared with the prediction results of the other two methods, the method used in this paper has the best fitting effect, and the relative entropy is reduced by 1.1542. The comprehensive evaluation index F value of the prediction interval under each confidence level has been significantly improved, among which the increase at 80% confidence level is the most, and the F value increases by 0.0063.

Wang J., Wang S., Zeng B., Lu H.

• Developed a novel deep QrBiLStm that can perform probabilistic forecasts. • An ensemble probabilistic forecasting system was proposed. • A pseudo-interval training method for ensemble probabilistic forecast is designed. • Improved the optimizer with three strategies. The quantification of wind speed uncertainty is of great significance for real-time control of wind turbines and power grid dispatching. However, the intermittence and fluctuation of wind energy present great challenges in modeling its uncertainty; research in this field is limited. A quantile regression bi-directional long short-term memory network (QrBiLStm) and a novel ensemble probabilistic forecasting strategy are proposed in this study to explore ensemble probabilistic forecasting. To verify the reliability of the proposed ensemble probabilistic forecasting system, the uncertainties of wind speed at wind farms in China were modeled as a case study. The results of comparative experiments including 15 other models demonstrate the superiority of this ensemble probabilistic forecasting system in terms of sharpness while maintaining high interval coverage. More specifically, it was observed that the prediction interval coverage probability obtained by the proposed system is above 97%, and the sharpness is improved by at least 24.21% as compared with the commonly used single models. The proposed ensemble probabilistic forecasting system can accurately quantify the uncertainty of wind speed, and also reduce the operation cost of power systems by improving the efficiency of wind energy utilization.

Goh H.H., Liao L., Zhang D., Dai W., Lim C.S., Kurniawan T.A., Goh K.C., Cham C.L.

Noise significantly reduces the detection accuracy of transient power quality disturbances. It is critical to denoise the disturbance. The purpose of this research is to present an improved wavelet threshold denoising method and an adaptive parameter selection strategy based on energy optimization to address the issue of unclear parameter values in existing improved wavelet threshold methods. To begin, we introduce the peak-to-sum ratio and combine it with an adaptive correction factor to modify the general threshold. After calculating the energy of each layer of wavelet coefficient, the scale with the lowest energy is chosen as the optimal critical scale, and the correction factor is adaptively adjusted according to the critical scale. Following that, an improved threshold function with a variable factor is proposed, with the variable factor being controlled by the critical scale in order to adapt to different disturbance types’ denoising. The simulation results show that the proposed method outperforms existing methods for denoising various types of power quality disturbance signals, significantly improving SNR and minimizing MSE, while retaining critical information during disturbance mutation. Meanwhile, the effective location of the denoised signal based on the proposed method is realized by singular value decomposition. The minimum location error is 0%, and the maximum is three disturbance points.

Pei N., Wu Y., Su R., Li X., Wu Z., Li R., Yin H.

During long-term geological tectonic processes, multiple fractures are often developed in the rock mass of high-level radioactive waste disposal sites, which provide channels for release of radioactive material or radionuclides. Studies on the permeability of fractured rock masses are essential for the selection and evaluation of geological disposal sites. With traditional methods, observation and operation of fractured rock mass penetration is time-consuming and costly. However, it is possible to improve the process using new methods. Based on the penetration characteristics of fractured rock mass, and using machine learning techniques, this study has created a prediction model of the fractured rock mass permeability based on select physical and mechanical parameters. Using the correlation coefficients developed by Pearson, Spearman, and Kendall, the proposed framework was first used to analyze the correlation between the physical and mechanical parameters and permeability and determine the model input parameters. Then, a comparison model was created for permeability prediction using four different machine-learning algorithms. The algorithm hyper-parameters are determined by a ten-fold cross-validation. Finally, the permeability interval prediction values are obtained by comparing and selecting the prediction results and probability distribution density function. Overall, the computational results indicate the framework proposed in this paper outperforms the other benchmarking machine learning algorithms through case studies in Beishan District, Gansu, China.

Wang J., Wang S., Li Z.

As a renewable, clean and economical energy source, wind energy has rapidly infiltrated into the modern power grid system. Wind speed forecasting, the crucial technology of wind power grid connection, has attracted large amounts of scholars for research and modeling. However, a large number of models only focus on the point forecasts, which are far from meeting the requirements of risk control and evaluation of power system. To fill the gap, a novel forecasting model which combined the modified multi-objective tunicate algorithm, benchmark models, and Quantile regression is proposed for deterministic and probabilistic interval forecasts. Theoretical proof demonstrates that the proposed modified algorithm can combine the merits of all benchmark models and better solve the nonlinear characteristics of wind speed. Comparative experiments which include sixteen relevant models are performed on three datasets to validate the performance of the proposed model. Simulation results show that the proposed model is the most accurate in all datasets, and can also get the interval forecast results with relatively high coverage and the narrowest width. Therefore, this model can provide accurate point forecasting results and uncertainty information, which is beneficial to the real-time control of wind turbine and power grid dispatching.

Lian L., He K.

The accuracy of wind power prediction directly affects the operation cost of power grid and is the result of power grid supply and demand balance. Therefore, how to improve the prediction accuracy of wind power is very important. In order to improve the prediction accuracy of wind power, a prediction model based on wavelet denoising and improved slime mold algorithm optimized support vector machine is proposed. The wavelet denoising algorithm is used to denoise the wind power data, and then the support vector machine is used as the prediction model. Because the prediction results of support vector machine are greatly affected by model parameters, an improved slime mold optimization algorithm with random inertia weight mechanism is used to determine the best penalty factor and kernel function parameters in support vector machine model. The effectiveness of the proposed prediction model is verified by using two groups actually collected wind power data. Seven prediction models are selected as the comparison model. Through the comparison between the predicted value and the actual value, the prediction error and its histogram distribution, the performance indicators, the Pearson’s correlation coefficient, the DM test, box-plot distribution, the results show that the proposed prediction model has high prediction accuracy.

Wang Y., Xu C., Wang Y., Cheng X.

A comprehensive fault diagnosis method of rolling bearing about noise interference, fault feature extraction, and identification was proposed. Based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), detrended fluctuation analysis (DFA), and improved wavelet thresholding, a denoising method of CEEMDAN-DFA-improved wavelet threshold function was presented to reduce the distortion of the noised signal. Based on quantum-behaved particle swarm optimization (QPSO), multiscale permutation entropy (MPE), and support vector machine (SVM), the QPSO-MPE-SVM method was presented to construct the fault-features sets and realize fault identification. Simulation and experimental platform verification showed that the proposed comprehensive diagnosis method not only can better remove the noise interference and maintain the original characteristics of the signal by CEEMDAN-DFA-improved wavelet threshold function, but also overcome overlapping MPE values by the QPSO-optimizing MPE parameters to separate the features of different fault types. The experimental results showed that the fault identification accuracy of the fault diagnosis can reach 95%, which is a great improvement compared with the existing methods.

Wu S., Jia L., Liu Y.

Wind energy is a typical representative of environmentally friendly new energy. However, its huge randomness and suddenness have caused many harms and losses to the actual applied wind power. Therefore, predicting wind energy in advance and improving the prediction accuracy have become the top priority. Since the wind energy data is a kind of time series, LSTM model has excellent performance. Most researches are focused on one-dimensional wind energy data. This paper uses a multivariate LSTM model. In addition to weather conditions, the wind energy at the previous moment is included as one of the variables to establish a prediction model. The model built including the wind power at the last moment performed very well. This paper compares it with the prediction model that does not consider the wind power at the previous moment, and verifies its effectiveness.The innovation of this paper is that firstly, wavelet denoising, a signal processing method, is applied to wind energy data processing, and secondly, a multivariate LSTM model is used to forecast wind energy. Considering that wind energy is a time series, the wind energy at the previous moment is also introduced except wind speed and wind direction. Compared with the prediction model that does not include the wind energy at the previous moment, the accuracy is significantly improved.

Peng X., Wang H., Lang J., Li W., Xu Q., Zhang Z., Cai T., Duan S., Liu F., Li C.

Effective wind-power prediction enhances the adaptability of a wind power system to the instability of wind power, which is beneficial for load and frequency regulation, helping to convert wind power to electricity and connect wind power to the grid safely. Moreover, the use of numerical weather prediction (NWP) to predict the probability results of wind power is a matter of general concern in the field of wind power prediction, and deep neural networks have become an indispensable research tool. In this study, a new neural-network prediction model called EALSTM-QR was developed for wind-power prediction considering the input of NWP and the deep-learning method. In the model, there are four main levels: Encoder, Attention, bidirectional long short-term memory (LSTM), and quantile regression (QR). The combination inputs contain historical wind-power data and the features extracted and obtained from the NWP through the Encoder and Attention levels. The bidirectional LSTM is used to generate wind-power time-series probability prediction results. The QR method and confidence interval limits are used to obtain the final prediction intervals. The proposed method was compared with several interval prediction models and probability prediction models based on neural networks for wind-power prediction by using datasets from wind farms in China. The results indicated that the proposed EALSTM-QR has good accuracy and reliability for the prediction of intervals and probabilities. • A novel model is developed for wind power probabilistic forecasting. • The NWP features are extracted to improve the accuracy. • A new learning method is presented. • The PositionEncoding and MultiHeadAttention have been applied in the model.

Qiao Y., Li Q., Qian H., Song X.

Abstract Seismic data usually contains a lot of noise. In order to effectively remove noise and improve the signal-to-noise ratio of seismic signals, this paper proposes a method of combining complete empirical mode decomposition (CEEMD) with improved wavelet threshold denoising.method. CEEMD has good adaptability to signal decomposition; the new wavelet threshold function can effectively overcome the discontinuity of hard threshold function and the deviation of wavelet coefficients in soft threshold function. The combination of the two methods can obtain better denoising effect. After processing the simulated signal with the method proposed in this paper, the signal-to-noise ratio is significantly better than the traditional single denoising method.

Gan Z., Li C., Zhou J., Tang G.

• A promising wind speed interval prediction method with deep learning technique is proposed. • TCNs are employed to develop a neotype interval prediction method in LUBE framework. • An interval width adaptive adjustment strategy is designed to optimize the model by directly constructing training labels. • Experimental results demonstrate the superiority of the method on prediction accuracy and reliability by comparing with the state of art methods. Wind speed interval prediction is one of the most elusive and long-standing challenges in wind power production. As a data source with intermittent and fluctuant characteristics, wind speed time series require highly nonlinear temporal features for the prediction tasks. In this paper, a novel interval prediction model is proposed based on temporal convolutional networks to forecast wind speed. A temporal convolutional networks architecture layer, multiple fully connected layers using tanh activation function and an end-to-end sorting layer are respectively served as input, hidden and output layers of the temporal convolutional networks interval prediction model which can generate prediction intervals directly. Additionally, an adaptive interval construction optimization strategy is put forward to devise training labels for learning of model. Eight cases from two wind fields are implemented to test and verify the proposed method. Specially, experiments have been designed to compare the prediction accuracy and reliability between the proposed model and the most recent state-of-the-art models. The forecasting results suggest that the proposed model has a significant performance improvement on both prediction interval coverage probability and prediction interval width criteria and thus can be a practical tool for wind speed forecasting.

Zhang Y., Pan G., Zhao Y., Li Q., Wang F.

At present, environmental pollution, climate warming and other problems are becoming more and more serious. And wind energy is pollution-free and never be exhausted, so it can make a major contribution to the global energy transformation. However, its random fluctuations and uncertainties bring adverse effects to the power system and endanger the safety of the power grid. Therefore, this paper combines artificial intelligence methods with statistical knowledge, and proposes a new interval prediction model based on the Fast Correlation Based Filter (FCBF) algorithm, the optimized Radial Basis Function (RBF) model and Fourier distribution for wind speed. Firstly considering environmental factors, this paper studies multi-factor wind speed prediction and applies the FCBF algorithm to filter the factors that affect the wind change. After that, this paper applies the idea of the Extremal Optimization (EO) to improve the Particle Swarm Optimization (PSO) and constructs a new EPSO optimization model for optimizing the RBF model. Next, using the Fourier function to fit the error probability distribution, and the wind speed interval is estimated based on point prediction results. Finally, the actual data of Changma Wind Farm is used for experiments to verify the feasibility and effectiveness of the proposed model. And through experimental results and comparison, it can be concluded: (1) Using the FCBF algorithm to select input variables can reduce redundant variables and lay a good foundation for subsequent prediction; (2) Applying the constructed EPSO-RBF model to predict wind speed, and the maximum and average value of the prediction error are only 0.8430 m/s, 0.1749 m/s, which is significantly better than several other traditional neural network models; (3) Introducing the Fourier function into the wind speed interval prediction, even at the 80% confidence level, the average width of the interval prediction is less than 3 m/s, and the coverage rate is higher than 90%.