Wind Power Dataset Research Articles

Uncertain wind power forecasting using LSTM‐based prediction interval

Estimating prediction intervals (PIs) is an efficient and reliable way of capturing the uncertainties associated with wind power forecasting. In this study, a state of the art recurrent neural network (RNN) known as long short-term memory (LSTM) is used to produce reliable PIs for one-hour ahead wind power uncertainty forecast using the non-parametric lower upper bound estimation framework. Two realistic hourly stamped wind power data sets are obtained and by using mutual information and false nearest neighbours techniques, the data are made suitable for model inputs. A novel comprehensive objective function consisting of the coverage probability, the average width of the PIs, symmetricity and variational synchronicity is developed to train the LSTM model using intelligent optimisation techniques. The standard of the PIs generated for the test set as well as for different seasons are evaluated based on the indices used to design the objective function for model training, with one of them being modified. The performance of the proposed LSTM model is found to outperform typical RNN models like Elman, non-linear auto-regressive with exogenous models and other benchmarking models while tested on the real-world data sets.

Country-wide German hourly wind power dataset mined to provide insight to predictions and forecasts with optimized data-matching machine learning

German wind power (MW) generated on an hourly basis is evaluated in terms of weather-environmental-market related inputs using 8784 data records for the year 2016. An optimized data-matching algorithm achieves high degrees of accuracy predicting MW from a diverse set of inputs across the year and forecasting MW in the short-term using only historical inputs. The developed algorithm demonstrates its value by enabling detailed auditable data mining of all data record predictions and forecasts and providing forensic knowledge of the dataset. The method initially identifies the most closely matching data records using squared differences between each of the inputs included in its feature selection. Subsequently, it applies standard and memetic optimizers to minimize the squared errors for its MW forecasts by tuning a small subset of only ∼150 data records (∼1.7% of all data records). It then selects the best optimum solution based on evaluations with a set of data records held independently from the tuning process. The accuracy realized for the best solution determined, applied using all data records, is RMSE = 791.4 MW and R2 = 0.988. The dataset has a few significant MW prediction outliers. Detailed auditable analysis of the prediction outliers identifies and explains their causes. The method also provides high forecasting accuracy and data insight for short-term time series data. It achieves one hour ahead (t + 1) forecasting accuracies of RMSE = 879.5 MW; R2 = 0.9802, and three hour ahead (t + 3) forecasting accuracies of RMSE = 1880.1 MW; R2 = 0.9095 using just 1000 h of historical data records.

Probabilistic wind power forecasting based on spiking neural network

Accurate and reliable quantification of uncertainty in wind power forecasting is critical to the economic operation and real time control of the electric power and energy system. To this end, this paper proposes a novel direct method for probabilistic wind power forecasting based on spiking neural network. In this method, a new forecasting framework is firstly formulated to simultaneously calculate the coverage probability and sharpness with associated confidence levels. Then, group search optimizer is introduced and re-designed to optimize the parameters of the forecasting framework and directly generate the prediction intervals, so that the prediction reliability and stability are ensured. The main advantage of the proposed probabilistic forecasting method is that it does not involve any distribution assumption of the prediction errors required by most existing forecasting methods. The wind power datasets from real wind farms in Belgium and China are used in the case studies. Traditional back-propagation neural network (BPNN), support vector machine (SVM) and extreme learning machine (ELM) are selected as the benchmarking algorithms. The simulation results show that the average coverage of the proposed method is improved by 72.0%, 54.9% and 51.3% respectively, when compared to BPNN, SVM and ELM. The improvement rates of sharpness index are 43.1%, 28.1% and 21.0%, respectively. These statistical results show that the proposed method outperforms BPNN, SVM and ELM in terms of forecasting accuracy, demonstrating that this method has high practical applications in real power systems.

Short‐term wind power forecasting based on two‐stage attention mechanism

Wind power is usually closely related to the meteorological information around the wind farm, which leads to the fluctuation of wind power and makes it difficult to predict precisely. In this study, a wind power forecasting model based on long–short-term memory network two-stage attention mechanism is established. The attention mechanism is extensively employed to weigh the input feature and strengthen the trend characteristic of wind power. The intermittency and volatility feature of the wind are efficiently mitigated, and the prediction accuracy is improved significantly. Besides, quantile regression and kernel density estimation are combined with the proposed model to predict the wind power interval and the probability density. These two parameters are important information for ensuring security and stability while accessing to the electricity grid. The simulation results on the actual wind power dataset verify the higher prediction accuracy of the proposed model compared with other machine learning methods.

Corrected multi-resolution ensemble model for wind power forecasting with real-time decomposition and Bivariate Kernel density estimation

The power integration is a challenge for the power system because of the fluctuation of the wind power. Wind power forecasting can estimate the future fluctuation of the wind power, and enhance the safety of the power integration. In this study, a corrected multi-resolution forecasting model is proposed to improve current wind power forecasting performance. The proposed model contains three stages, including multi-resolution ensemble, adaptive multiple error corrections and uncertainty estimation. Four real-time wind power data sets are applied to verify the effectiveness of the proposed model. The results are shown as follows: (a) the proposed model is effective for wind power forecasting, the 1-step index of agreement and coverage width-based criterion with 99% confidence level of the proposed model on the dataset #1 are 0.9432 and 0.6951 respectively; (b) the proposed model outperforms the previous models. Through techno-economic analysis, it can be concluded that the proposed model has the potential to be applied to improve the power integration performance.

A parameter and resolution adaptive algorithm for rapid detection of ramp events in different timescale databases of the power system

The flexibility of the power system gets more attention of power system operators due to the high penetration of renewable energy sources. This leads to the need of the ramp detection. Based on the detection results of ramp events (REs) in historical and forecast databases, the assessment of flexibility requirements in the power system can be significantly improved. Thus the system operator is able to procure minimum ancillary services and design flexible ramping products. Furthermore, the accuracy of the ramp forecasting can be increased. In the mean time, the rapid development of measuring facilities in the power system, such as the WAMS and the SCADA system, requires a robust detecting method to detect ramp events in the databases with different data properties and resolutions of the power system. This paper proposes a Parameter and Resolution Adaptive Algorithm (PRAA) to identify REs fast and accurately. A fast search method that can process raw databases with tens of millions of data points in seconds is proposed to merge adjacent REs with the same ramp direction. The bad data processing method is directly integrated into the proposed algorithm. A post-processing method is developed to increase the accuracy and robustness of the proposed algorithm. The raw wind power data sets from Bonneville Power Authority (BPA) and a real-world power system in China are utilized to validate the performance of the proposed PRAA.

A novel two-stage forecasting model based on error factor and ensemble method for multi-step wind power forecasting

With the fast growth of wind power penetration into the electric grid, wind power forecasting plays an increasingly significant role in the secure and economic operation of power systems. Although there have been numerous studies concerning wind power forecasting, most of them have failed to make the best of the information implied in the error value, focused only on simple error correction, adopted a simple ensemble method to aggregate the predictions of each component, and considered improving only forecasting accuracy. Recognizing these issues, a novel two-stage forecasting model based on the error factor, a nonlinear ensemble method and the multi-objective grey wolf optimizer algorithm is proposed for wind power forecasting. More specially, in stage I, the extreme learning machine optimized by the multi-objective grey wolf optimizer is used to forecast the components decomposed by variational mode decomposition, and an error prediction model based on the extreme learning machine optimized by the multi-objective grey wolf optimizer is utilized to predict forecast errors; also, a novel nonlinear ensemble method based on the extreme learning machine optimized by the multi-objective grey wolf optimizer is utilized to integrate all the components and forecast error values in stage II. Three real-world wind power datasets collected from Canada and Spain are introduced to demonstrate the forecasting performance of the developed model. The forecasting results reveal that the proposed model is superior to all the other considered models in terms of both accuracy and stability and thus can be a useful tool for wind power forecasting.

A Fuzzy Adaptive Probabilistic Wind Power Prediction Framework Using Diffusion Kernel Density Estimators

The inherent uncertainty in predicting wind power generation makes the operation and control of power systems very challenging. Probabilistic measurement of wind power uncertainty in the form of a reliable and sharp interval is of utmost importance, but construction of such high-quality prediction intervals (PIs) is difficult because wind power time series are nonstationary. In this paper, a framework based on the concept of bandwidth selection for a new and flexible kernel density estimator is proposed. Unlike previous related works, the proposed framework uses neither a cost function-based optimization problem nor point prediction results; rather, a diffusion-based kernel density estimator (DiE) is utilized to achieve high-quality PIs for nonstationary wind power time series. Moreover, to adaptively capture the uncertainties of both the prediction model and wind power time series in different seasons, the DiE is equipped with a fuzzy inference system and a tri-level adaptation function. The proposed framework is also founded based on a parallel computing procedure to promote the computational efficiency for practical applications in power systems. Simulation results demonstrate the efficiency of the proposed framework compared to well-known conventional benchmarks using real wind power datasets from Canada and Spain.

Direct Interval Forecast of Uncertain Wind Power Based on Recurrent Neural Networks

Interval forecast is an efficient method to quantify the uncertainties in renewable energy production. In this paper, the idea of prediction intervals (PIs) is employed to capture the uncertainty of wind power generation in power systems. The recurrent neural network (RNN) model is proposed to construct PIs with the lower upper bound estimation method, which is a powerful nonparametric forecast approach. In addition, a novel comprehensive cost function with a new PI evaluation index is designed with the purpose of enhancing the model training. To tune the parameters of RNN prediction model, the dragonfly algorithm with a linearly random weight update method is introduced as the optimization tool. The performance of the proposed prediction model is validated by a case study using a real world wind power dataset, and the comparative results show the superiority of the model.

Research of Wind Power Correlation With Three Different Data Types Based on Mixed Copula

With the increasing integration of large-scale wind farms, the stochastic characteristics of wind speed and the coupling relationship of geographically distributed wind farms become ineligible. Correlation investigation of wind farms based on copula theory is able to lay good foundation for further optimization and schedule in power systems. In this paper, three data types, including wind speed, calculated wind power, and actual wind power, are applied to explore the correlation of two geographically close wind farms. After graphical analysis and comparison of the time series, the structural and compositional characteristics of correlation by different data types are investigated based on mixed copula. Moreover, a two-stage filtration method is proposed to evaluate different types of copulas. Finally, taking into account the practical conditions after careful examination of actual wind power dataset, the practical conditions are applied to the calculated wind power dataset. Correlation research based on the adjusted calculated wind power dataset is further explored and revealed that it is more practical and prospective to provide reference for further power system operation with high penetration of wind farms.

Wind power research in Wikipedia: Does Wikipedia demonstrate direct influence of research publications and can it be used as adequate source in research evaluation?

This paper is a result of the WOW project (Wind power On Wikipedia) which forms part of the SAPIENS (Scientometric Analyses of the Productivity and Impact of Eco-economy of Spain) project (Sanz-Casado et al. in Scientometrics 95(1):197–224, 2013). WOW is designed to observe the relationship between scholarly publications and societal impact or visibility through the mentions of scholarly papers (journal articles, books and conference proceedings papers) in the Wikipedia, English version. We determine (1) the share of scientific papers from a specific set defined by Wind Power research in Web of Science (WoS) 2006–2015 that are included in Wikipedia entries, named data set A; (2) the distribution of scientific papers in Wikipedia entries on Wind Power, named data set B, captured via the three categories for the topic Wind Power in the Wikipedia Portal: Wind Power, Wind turbines and Wind farms; (3) the distributions of document types in the two wiki entry data sets’ reference lists. In parallel the paper aims at designing and test indicators that measure societal impact and RD and a density measure of those types in wiki entries. The study is based on Web mining techniques and a developed software that extracts a range of different types of Wikipedia references from the data sets A and B. Findings show that in data set A 25.4% of the wiki references are academic, with a density of 17.62 academic records detected per wiki entry. However, only 0.62% of the original WoS records on Wind Power are also found as wiki references, implying that the direct societal impact through the Wikipedia is extremely small for Wind Power research. In the second Wikipedia set on Wind Power (data set B), the presence of scientific papers is even more insignificant (10.6%; density: 3.08; WoS paper percentage: 0.26%). Notwithstanding, the Wikipedia can be used as a tool informing about the transfer from scholarly publications to popular and non-peer reviewed publications, such as Web pages (news, blogs), popular magazines (science/technology) and research reports. Non-scholarly wiki reference types stand for 74.6% of the wiki references (data set A) and almost 90% in data set B. Interestingly, the few WoS articles in wiki entries on Wind Power present on average 34.3 citations received during the same period (2006–2015) as WoS Wind Power publications not mentioned in wiki entries only receives on average 5.9 citations. Owing to the scarcity of Wind Power research papers in the Wikipedia, it cannot be applied as a direct source in evaluation of Wind Power research. This is in line with other recent studies regarding other subject areas. However, our analysis presents and discusses six supplementary indirect indicators for research evaluation, based on publication types found in the wiki entry reference lists: share of (WoS) records; density; and reference focus, plus popular science knowledge export, non-scholarly knowledge export and academic knowledge export. The same indicators are direct measures of the Wikipedia reference properties.

Harvested Energy Prediction Schemes for Wireless Sensor Networks: Performance Evaluation and Enhancements

We review harvested energy prediction schemes to be used in wireless sensor networks and explore the relative merits of landmark solutions. We propose enhancements to the well-known Profile-Energy (Pro-Energy) model, the so-called Improved Profile-Energy (IPro-Energy), and compare its performance with Accurate Solar Irradiance Prediction Model (ASIM), Pro-Energy, and Weather Conditioned Moving Average (WCMA). The performance metrics considered are the prediction accuracy and the execution time which measure the implementation complexity. In addition, the effectiveness of the considered models, when integrated in an energy management scheme, is also investigated in terms of the achieved throughput and the energy consumption. Both solar irradiance and wind power datasets are used for the evaluation study. Our results indicate that the proposed IPro-Energy scheme outperforms the other candidate models in terms of the prediction accuracy achieved by up to 78% for short term predictions and 50% for medium term prediction horizons. For long term predictions, its prediction accuracy is comparable to the Pro-Energy model but outperforms the other models by up to 64%. In addition, the IPro scheme is able to achieve the highest throughput when integrated in the developed energy management scheme. Finally, the ASIM scheme reports the smallest implementation complexity.

Half-spectral space–time covariance models

We develop a new class of space–time Gaussian process models by specifying covariance functions using what we call a half-spectral representation. We establish general properties of half-spectral models and appeal to results on a screening effect analysis to claim more natural space–time interaction properties in the class of models we develop. To test this claim and to more generally test this class of models, we fit models we develop in this paper to a wind power dataset. We show our models fit these data better than other common separable and non-separable space–time models.

A hybrid wind power forecasting model based on data mining and wavelets analysis

Accurate forecasting of wind power plays a key role in energy balancing and wind power integration into the grid. This paper proposes a novel time-series based K-means clustering method, named T.S.B K-means, and a cluster selection algorithm to better extract features of wind time-series data. A hybrid of T.S.B K-means, discrete wavelet transform (DWT) and harmonic analysis time series (HANTS) methods, and a multilayer perceptron neural network (MLPNN) is developed for wind power forecasting. The proposed T.S.B K-means classifies data into separate groups and leads to more appropriate learning for neural networks by identifying anomalies and irregular patterns. This improves the accuracy of the forecast results. A cluster selection method is developed to determine the cluster that provides the best training for the MLPNN. This significantly accelerates the forecast process as the most appropriate portion of the data rather than the whole data is used for the NN training. The wind power data is decomposed by the Daubechies D4 wavelet transform, filtered by the HANTS, and pre-processed to provide the most appropriate inputs for the MLPNN. Time-series analysis is used to pre-process the historical wind-power generation data and structure it into input-output series. Wind power datasets with diverse characteristics, from different wind farms located in the United States, are used to evaluate the accuracy of the hybrid forecasting method through various performance measures and different experiments. A comparative analysis with well-established forecasting models shows the superior performance of the proposed forecasting method.

One-hour-ahead wind power forecast using hybrid grey models

This paper proposes two hybrid grey-based short-term wind power prediction techniques: GM(1,1)-ARMA and GM(1,1)-NARnet. These techniques are combined with ARMA models and nonlinear autoregressive neural network (NARnet) models, respectively. The efficiency of these algorithms is examined using a recorded wind power dataset. The performance of these predictors is compared with classical ARMA models as well as the traditional grey model GM(1,1). Unlike the classical predictors, the proposed hybrid algorithms are not affected by the inherent uncertainty in the wind power. Therefore, the results obtained using the proposed hybrid algorithms outperform those obtained using classical predictors. In contrast to the GM(1,1)-ARMA model, the GM(1,1)-NARnet model utilises the nonlinear components of wind power in the forecasting procedure. Consequently, the obtained results from the GM(1,1)-NARnet outperform those obtained by the GM(1,1)-ARMA.

Machine Learning based short term wind power prediction using a hybrid learning model

Depletion of conventional resources has led to the exploration of renewable energy resources. In this regard, wind power is taking significant importance, worldwide. However, to acquire consistent power generation from wind, the expected wind power is required in advance. Consequently, various prediction models have been reported for wind power prediction. However, we observe that Support Vector Regression (SVR), and specially, a hybrid learning model based on SVR offer better performance and generalization compared to multiple linear regression (MLR) and is thus quite suitable for the development of short-term wind power prediction system. To this end, a new methodology ML-STWP namely Machine Learning based Short Term Wind Power Prediction is proposed for short-term wind power prediction. This approach utilizes a combination of machine learning (ML) techniques for feature selection and regression. The proposed methodology is thus a hybrid ML model, which makes use of feature selection through irrelevancy and redundancy filters, and then employs SVR for auxiliary prediction. Finally, the wind power is predicted using enhanced particle swarm optimization and a hybrid neural network.The wind power dataset on which the model is tuned and tested consists of real-time daily values of wind speed, relative humidity, temperature, and wind power. The obtained results demonstrate that the proposed prediction model performs better as compared to the existing methods and demonstrates the efficacy of the proposed intelligent system in accurately predicting wind power on daily basis.