Probabilistic Wind Power Forecasting Research Articles

A review of short‐term wind power probabilistic forecasting and a taxonomy focused on input data

A review of state-of-the-art short-term wind power probabilistic forecasting models is the focus here. The improvement of the accuracy and efficiency of probabilistic forecasting models has been in the centre of attention of researchers in recent years, since the need to further comprehend and efficiently use the uncertainty of forecasts is increasing. Since the optimal operation and control of energy systems and electricity markets is one of the important aspects of performing wind power forecasts, this review focuses in short-term probabilistic forecasting models, which could prove to be useful in the daily planning and operation of power systems. The short-term concept of forecasts is analysed in detail, along with the case studies and examples proposed by the reviewed literature. The key advantages and disadvantages of the reviewed probabilistic forecasting methodologies are identified. Furthermore, different classifications of the reviewed works according to the data that are used to provide an accurate forecasting model are also provided. Future directions in the field of short-term wind power probabilistic forecasting are also proposed.

Regional wind power probabilistic forecasting based on an improved kernel density estimation, regular vine copulas, and ensemble learning

Reliable wind energy forecasting is crucial for the stable operation of power grids. This paper proposes a regional wind power probabilistic forecasting model comprising an improved kernel density estimation (IKDE), regular vine copulas, and ensemble learning. The IKDE is firstly used to generate the margin probability density function (PDF) of each wind farm and the KDE bandwidth is optimized via the golden-section search algorithm to obtain the best possible prediction. Then, several dependence structures are formulated by building different regular vine copulas based on multiple criteria, and all the dependence structures work together with marginal PDF to generate respective joint distribution functions. Finally, ensemble learning is applied to combine all the joint distribution functions and establish an ultimate distribution function. Furthermore, a novel multi-distribution mega-trend-diffusion (MD-MTD) with parametric optimization is proposed to improve the prediction when the data are insufficient. The results of comparative evaluations conducted on datasets from eight wind farms indicate that the proposed model outperforms existing models in wind power generation prediction. Specifically, the proposed model can reliably forecast power generation for an entire region for the next 24 h with only three months of historical data. In contrast, most benchmark models require a year of data.

Spatio-temporal probabilistic forecasting of wind power for multiple farms: A copula-based hybrid model

Accurate probabilistic forecasting of wind power output is critical to maximizing network integration of this clean energy source. There is a large literature on temporal modeling of wind power forecasting, but considerably less work combining spatial dependence into the forecasting framework. Through the careful consideration of the temporal modeling component, complemented by support vector regression of the temporal model residuals, this work demonstrates that a DVINE copula model most accurately represents the residual spatial dependence. Additionally, this work proposes a complete set of validation mechanisms for multi-h-step forecasts that, when considered together, comprehensively evaluate accuracy. The model and validation mechanisms are demonstrated in two case studies, totaling ten wind farms in the Texas electric grid. The proposed method outperforms baseline and competitive models, with an average Continuous Ranked Probability Score of less than 0.15 for individual farms, and an average Energy Score of less than 0.35 for multiple farms, over the 24-hour-ahead horizon. Results show the model’s ability to replicate the power output dynamics through calibrated and sharp predictive densities.

Probabilistic Forecasting Based Sizing and Control of Hybrid Energy Storage for Wind Power Smoothing

With the increasing wind power integration, the security and economy of the power system operations are greatly influenced by the intermittency and fluctuation of wind power. Due to the flexible operational modes for charging/discharging, the hybrid energy storage system (HESS) is composed of battery energy storage system and super-capacitor can effectively mitigate the wind power uncertainty. This paper proposes a probabilistic forecasting-based HESS sizing and control scheme to cost-effectively smooth wind power fluctuations. First, probabilistic wind power forecasting is combined with multivariate Gaussian copula to generate temporally correlated wind power scenarios. Then, an adaptive variational mode decomposition (VMD) method is proposed to extract the frequency components of each wind scenario and determine the pre-scheduled power of wind-HESS system adaptively. Finally, a two-stage stochastic optimization model is constructed to determine the capacity and correct the pre-scheduled power of HESS with the objective of minimizing total cost. Case studies based on actual data from a Danish wind farm demonstrate that the proposed HESS sizing and control scheme can significantly reduce the installation cost and operation cost of HESS and prominently smooth wind power fluctuations.

Review of Deterministic and Probabilistic Wind Power Forecasting: Models, Methods, and Future Research

The need to turn to more environmentally friendly sources of energy has led energy systems to focus on renewable sources of energy. Wind power has been a widely used source of green energy. However, the wind’s stochastic and unpredictable behavior has created several challenges to the operation and stability of energy systems. Forecasting models have been developed and excessively used in recent decades in order to deal with these challenges. Deterministic forecasting models have been the main focus of researchers and are still being developed in order to improve their accuracy. Furthermore, in recent years, in order to observe and study the uncertainty of forecasts, probabilistic forecasting models have been developed in order to give a wider view of the possible prediction outcomes. Advanced probabilistic and deterministic forecasting models could be used in order to facilitate the energy systems operation and energy markets management. This paper introduces an overview of state-of-the-art wind power deterministic and probabilistic models, developing a comparative evaluation between the different models reviewed, identifying their advantages and disadvantages, classifying and analyzing current and future research directions in this area.

Regional Wind Power Probabilistic Forecasting Based on an Improved Kernel Density Estimation, Regular Vine Copulas, and Ensemble Learning

Reliable wind energy forecasting is crucial for the stable operation of power grids. This paper proposes a regional wind power probabilistic forecasting model comprising an improved kernel density estimation (IKDE), regular vine copulas, and ensemble learning. The IKDE is firstly used to generate the margin probability density function (PDF) of each wind farm and the KDE bandwidth is optimized via the golden-section search algorithm to obtain the best possible prediction. Then, several dependence structures are formulated by building different regular vine copulas based on multiple criteria, and all the dependence structures work together with marginal PDF to generate respective joint distribution functions. Finally, ensemble learning is applied to combine all the joint distribution functions and establish an ultimate distribution function. Furthermore, a novel multi-distribution mega-trend-diffusion (MD-MTD) with parametric optimization is proposed to improve the prediction when the data are insufficient. The results of comparative evaluations conducted on datasets from eight wind farms indicate that the proposed model outperforms existing models in wind power generation prediction. Specifically, the proposed model can reliably forecast power generation for an entire region for the next 24 h with only three months of historical data. In contrast, most benchmark models require a year of data.

A new hybrid model for point and probabilistic forecasting of wind power

The accurate and reliable forecasting of wind power is of great importance for electrical systems’ control and operation. However, the intermittent nature of wind power generation implies a complicated forecasting framework. In this paper, a new hybrid model including three steps is proposed for point and probabilistic forecasting of wind power. Within the first step, by using data preprocessing methods, proposed weighted Extreme Learning Machine (ELM) by Mutual Information, and bootstrap approach, point forecasting and variance of the model uncertainties are estimated. In the second step, by employing ELM, bootstrap approach, and an ensemble structure, the noise variance is calculated. During the final step, to improve the results of the probabilistic forecasting, methods consisting of ELM, bootstrap, improved particle swarm optimization based on information feedback models and a new proposed prediction interval based objective function are used. Effectiveness of the proposed hybrid model is verified by employing real data of Australian wind farms for 1-h ahead and day ahead forecasting.

A novel probabilistic short-term wind energy forecasting model based on an improved kernel density estimation

The evolution of renewable energy especially wind energy over the past decade has surpassed all expectations. Short-term probabilistic wind power forecasting is a good option to increase the reliability of power system. But short-term forecasting of power generated by wind turbine has high error due to uncertainty parameter such as wind speed so, it is very important to find a way to increase the accuracy of forecasting. Therefore, in this paper, a novel forecasting method that has high reliability compared to other methods is presented. In this paper, in order to benefit the superiority of various prediction models, a new Improved Kernel Density Estimation (IKDE) method is exploited to estimate the wind energy possibility. The combination of various prediction models and the suggested method might develop the function of probabilistic prediction by supplying divergent types of compactness performance. KDE method is a powerful method to analyze background and foreground characteristic. In order to increase the efficiency of KDE method some of predicting model are combined and a detection algorithm based on an Improved Kernel Density Estimation (IKDE) model is defined. The appropriate bandwidths, comparative threshold, comparative background sample learning array, and an enhanced sample updating model for sample learning array are proposed as the basics of the IKDE model. Two levels of optimization are used to simplify the IKDE model parameters. Finally, in order to prove the superiority of the proposed method over other methods, this method and 4 other methods have been implemented on 10 wind farms. The simulation results show that the prediction accuracy of the proposed method is about 3.8% higher than other methods due to the improved structure.

Leveraging Turbine-Level Data for Improved Probabilistic Wind Power Forecasting

This paper describes two methods for creating improved probabilistic wind power forecasts through the use of turbine-level data. The first is a feature engineering approach whereby deterministic power forecasts from the turbine level are used as explanatory variables in a wind farm level forecasting model. The second is a novel bottom-up hierarchical approach where the wind farm forecast is inferred from the joint predictive distribution of the power output from individual turbines. Notably, the latter produces probabilistic forecasts that are coherent across both turbine and farm levels, which the former does not. The methods are tested at two utility scale wind farms and are shown to provide consistent improvements of up to 5%, in terms of continuous ranked probability score compared to the best performing state-of-the-art benchmark model. The bottom-up hierarchical approach provides greater improvement at the site characterized by a complex layout and terrain, while both approaches perform similarly at the second location. We show that there is a clear benefit in leveraging readily available turbine-level information for wind power forecasting.

The Schaake Shuffle Technique to Combine Solar and Wind Power Probabilistic Forecasting

One way to mitigate the variability of wind and solar power generation is to install the corresponding plants in nearby locations. For example, in Kuwait, the facility at Shagaya Renewable Energy Park is located in a desert area with both photovoltaic panels and wind turbines that allow the continuous generation of renewable energy throughout the day. The National Center for Atmospheric Research (NCAR) has developed a system to generate probabilistic wind and solar predictions for the Shagaya facility. These predictions are based on the analog ensemble technique that post-processes the wind speed and solar irradiance predictions based on a combination of multiple models including the Weather Research and Forecasting (WRF) numerical model. The ensemble forecasts have 20 members and are generated independently at each wind and solar power production facility. Here we present a method based on the Schaake Shuffle (SS) technique to pair the ensemble members from the independent systems to obtain a unique ensemble prediction of the aggregated wind and solar generation. After reordering through the SS technique, the corresponding paired solar and wind power members can be summed to build a unique ensemble of combined generation that is statistically consistent, as verified by the presented metrics.

Using of improved models of Gaussian Processes in order to Regional wind power forecasting

The fluctuation of the wind speed and direction is due to the stochastic nature of the wind and the enforcing atmospheric pressure. Accordingly, the output power forecasting of the wind farms (WFs) will be difficult. In this paper, a novel method based on Gaussian Processes (GPs) is proposed to improve the probabilistic prediction of the WF levels and regional WFs. The Covariance Functions (CFs) are the key ingredient in using GPs. Thus, different groupings of CFs are investigated comprehensively. The GP includes different types as dynamic, static, direct, indirect, and combined structures, which are investigated in this study. The results of comparison between dynamic and static GP, reveal that the dynamic GP generates keen Prediction Intervals (PIs). In addition, with comparing the accuracy of direct and indirect prediction plan, it shows that indirect prediction strategy brings about wider PIs. The various evaluation metrics have applied to benchmark the different methods performance, and its results show that the indirect–dynamic GP has better performance than other combined structures of GP as well as other methods, in both WF levels and regional WFs, while its maximum error has obtained as about 5% less than others. Moreover, the proposed model provides precise results of forecasted energy in every time steps in both deterministic and probabilistic wind power forecasting. The compared results between indirect–dynamic GP and other structures show the highest average coverage error, about 1% and 2.2% higher in the regional level and WF levels, respectively, the lowest prediction interval nominalized average width, about 5% and 15% lower in the regional level and WF levels, respectively, and the highest interval sharpness, about 2% and 5% higher in the regional level and WF levels, respectively.

Improved Deep Mixture Density Network for Regional Wind Power Probabilistic Forecasting

Unsteady motion of the atmosphere incurs nonlinear and spatiotemporally coupled uncertainties in the wind power prediction (WPP) of multiple wind farms. This brings both opportunities and challenges to wind power probabilistic forecasting (WPPF) of a wind farm cluster or region, particularly when wind power is highly penetrated within the power system. This paper proposes an Improved Deep Mixture Density Network (IDMDN) for short-term WPPF of multiple wind farms and the entire region. In this respect, a deep multi-to-multi (m2m) mapping Neural Network model, which adopts the beta kernel as the mixture component to avoid the density leakage problem, is established to produce probabilistic forecasts in an end-to-end manner. A novel modified activation function and several general training procedures are then introduced to overcome the unstable behavior and NaN (Not a Number) loss issues of the beta kernel function. Verification of IDMDN is based on an open-source dataset collected from seven wind farms, and comparison results show that the proposed model improves the WPPF performance at both wind farm and regional levels. Furthermore, a laconic and accurate probabilistic expression of predicted power at each time step is produced by the proposed model.

A Novel Framework of Reservoir Computing for Deterministic and Probabilistic Wind Power Forecasting

The path towards wind power forecasting has yielded huge socio–economic benefits at a global scale. However, most of the previous studies tend to emphasize the improvement of deterministic forecasting, usually losing sight of the significance of probabilistic forecasting. In this paper, a novel forecasting system that can perform deterministic and probabilistic forecasting of wind power simultaneously, composed by the modules of feature selection, forecasting, system optimization, and system evaluation is presented to further supplement the existing studies in this field. Concretely, a hybrid feature selection strategy is proposed in the feature selection module to determine optimal system input; superior to traditional gradient descent algorithm, a dynamic reservoir theory-based recurrent neural network is developed in the forecasting module; an enhanced multi-objective optimization algorithm with the objectives of accuracy and stability is proposed in the system optimization module to provide an optimal scenario for system parameters; the effectiveness and feasibility of the proposed system is then validated in the evaluation module. Moreover, the comprehensive performance analysis of the proposed system is investigated in depth. Finally, the experimental results demonstrate that the proposed system has a significant advantage over the benchmarks considered, further verifying its tremendous potential to be used in a practical wind power system.

Multi-Objective Optimal Dispatching for a Grid-Connected Micro-Grid Considering Wind Power Forecasting Probability

In recent years, a large number of wind power has been applied in the micro-grid (MG). Influenced by randomness characteristics of wind speed, the uncertainty in the power output of wind turbines imposes some safety and stability problems on the optimal energy management in MG. To address this problem, an expert energy management system (EEMS) considering wind power probability is developed in this study for optimal dispatching of a typical grid-connected MG. The EEMS composes of wind power probabilistic forecasting module, multi-objective optimization module and energy storage system (ESS) module. In the wind power forecasting module, wind power probabilistic forecasting based on complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and Gaussian process regression (GPR) is proposed in this study. To improve the forecasting results, CEEMDAN, an effective signal processing method, is employed to decompose the wind power data, then, the decomposed subseries are utilized as the inputs of GPR for probabilistic forecasting. A two-step solution methodology combining an efficient and effective improved multi-objective bat algorithm (IMOBA) with fuzzy set theory (FST) is put forward to solve the optimal dispatching problems. In the first step, IMOBA is developed to optimize the energy dispatching of EEMS by minimizing both economic cost and pollutant emissions simultaneously, and obtain a well-distributed set of Pareto optimal front (POF), then, FST is employed to identify the best compromise solutions from POF. Six operational scenarios of a typical grid-connected MG based one-POF-one-day and one-POF-one-hour dispatching schemes are constructed to investigate the effectiveness of the proposed strategy and provide more flexibility for decision makers. The results illustrate that EEMS can effectively schedule power generation and energy storage by considering economic cost and pollutant emission objectives simultaneously.

Analytical Model of Day-ahead and Real-time Price Correlation in Strategic Wind Power Offering

In this paper, the model of strategic wind power offering in the day-ahead (DA) market is proposed considering the uncertainties of wind power production, and price forecasting of DA and real-time (RT) market. The wind power deviation in the RT market is settled with the two-price mechanism based on the deviation direction and the relation between the locational marginal prices (LMPs) of DA and RT. Instead of using the point forecasting for the DA and RT LMPs, the uncertainties of LMP forecasting are modeled. In addition, the correlation between the forecasting errors of DA and RT LMP are directly modeled instead of generating the correlated scenarios. Finally, the optimal offered quantity of wind power in the DA market is derived using the probability theory based on the probabilistic wind power forecasting. The case study using the price data of actual DA and RT from Midcontinent Independent System Operator (MISO) validates the effectiveness of the proposed model. It shows that the correlation of the forecasting errors of DA and RT LMP has a significant impact on the wind power quantity offered by DA and revenue results.

Combined model with secondary decomposition-model selection and sample selection for multi-step wind power forecasting

Wind power forecasting plays a significant role to ensure the safe operation of power systems. However, due to the stochastic nature and dynamic uncertainty of wind power, accurate and stable forecasting faces challenges. In this study, a modified combined structure with secondary decomposition-model selection and non-agnostic uncertainty sampling-active learning-sample selection strategy is proposed for multi-step deterministic and probabilistic wind power forecasting. Secondary decomposition-model selection strategy is used before experiments, selecting better forecasting models from the initial model space to the selected model space, and secondary decomposition methods are applied for decreasing the interference in initial data. Non-agnostic uncertainty sampling-active learning-sample selection strategy is employed to accelerate the sample selection process and enhance the testing efficiency, indirectly promoting the final performance. To further promote the forecasting precision and stability simultaneously, an advanced multi-objective optimization is applied, automatically selecting the top best models into the combined models from the selected model space. Except deterministic forecasting, uncertainty estimation is vital, offering different aspects of forecasting information for risk management, as the grid system absorbs large-scale wind power generation, therefore, probabilistic forecasting is considered too. Multi-step forecasting is also considered. Eight datasets from Elia, Belgium, are applied to assess the forecasting performance of secondary decomposition-model selection and the proposed combined models. According to results, compared with comparison models, the proposed combined models obtain higher accuracy and stability of multi-step deterministic forecasting (improving the forecasting skill by more than 90% in accuracy and stability for 1-step to 3-step forecasting) and also offer multi-step probabilistic information.

A novel wind power probabilistic forecasting approach based on joint quantile regression and multi-objective optimization

Probabilistic forecasts of wind power generation play a significant role in operation management and decision-making for wind power producers and power system operators. This study introduces a novel joint quantile model based on the use of Mercer’s kernels in the vector-valued RKHSs (Reproducing Kernel Hilbert Spaces) for the nonparametric probabilistic forecasts of wind power. The proposed approach takes the autocorrelation of the (nonlinearly) transformed input and output into consideration and improves the model flexibility by constructing composite kernels. Besides, the model is optimized by a new meta-heuristics algorithm named Multi-Objective Salp Swarm Optimization Algorithm (MSSA), which mathematically models and mimics the behavior of salp swarms for solving multiple objective optimization problems. Under this model framework, several conditional quantiles of wind power can be estimated and predicted at the same time. Moreover, due to the joint estimation process, the problem of quantile curve crossing can be effectively weakened during the process of model training. The qualitative and quantitative performances of the proposed method are tested and verified on a set of real case studies. The results show that the proposed algorithm provides superior outputs compared to the well-known and recent algorithms in the literature.

Multi-distribution ensemble probabilistic wind power forecasting

Ensemble methods have shown to be able to improve the performance of deterministic wind forecasting. In this paper, an improved multi-distribution ensemble (MDE) probabilistic wind power forecasting framework is developed to explore the advantages of different predictive distributions. Both competitive and cooperative strategies are applied to the developed MDE framework to generate 1–6 h ahead and day-ahead probabilistic wind power forecasts. Three probabilistic forecasting models based on Gaussian, gamma, and laplace predictive distributions are adopted to form the ensemble model. The parameters of the ensemble model (i.e., weights and standard deviations) are optimized by minimizing the pinball loss at the training stage. A set of surrogate models are built to quantify the relationship between the unknown optimal parameters and deterministic forecasts, which can be used for online forecasting. The effectiveness of the proposed MDE framework is validated by using the Wind Integration National Dataset (WIND) Toolkit. Numerical results of case studies at seven locations show that the developed MDE probabilistic forecasting methodology has improved the pinball loss metric score by up to 20.5% compared to the individual-distribution models and benchmark ensemble models.

Day ahead powerful probabilistic wind power forecast using combined intelligent structure and fuzzy clustering algorithm

Nowadays, Operational power forecasts are associated with the one-value deterministic Numerical Weather Forecasting (NWF) simulation in the anticipated wind speed. This article introduced a novel predicting methodology called SSOFC-Apriori-WRP, which presents one-day-ahead wind power and speed forecasting. This methodology relies highly on a Weather Research and Prediction (WRP) simulation, a shark smell optimization (SSO), enhanced fuzzy clustering (EFC), and an Apriori association procedure. The wind speed prediction with the help of shaped WRP model was produced. Then by dividing wind speed predictions series into different parts, definite conditions were met and were introduced. Next the suggested methodology by the combination of SSO-optimized fuzzy clustering and Apriori algorithm withdraws the association rules, which are dominated among the anticipating errors in the divided waves and the shape features. The suggested methodology could be implemented to the other compared models and decrease the unreliability of the WRP simulation, if the association rules are used in the ultimate optimization process.

Conditional aggregated probabilistic wind power forecasting based on spatio-temporal correlation

Aggregated probabilistic wind power forecasting is important for power system operations. In this paper, an improved aggregated probabilistic wind power forecasting framework based on spatio-temporal correlation is developed. A Q-learning enhanced deterministic wind power forecasting method is used to generate deterministic wind power forecasts for individual wind farms. The spatio-temporal correlation between the member wind farms and the aggregated wind power is modeled by a joint distribution model based on the copula theory. The marginal distributions of actual aggregated wind power and forecasted power of member wind farms are built with Gaussian mixture models. Then, a conditional distribution of the aggregated wind power is deduced through the Bayesian theory, which is used for aggregated probabilistic forecasts. The effectiveness of the proposed aggregated probabilistic wind power forecasting framework is validated by using the Wind Integration National Dataset Toolkit. Numerical results of case studies at nine locations show that the developed aggregated probabilistic forecasting methodology has improved the pinball loss metric score by up to 54% compared to three benchmark models.