Probabilistic wind power forecasting: A multi-scheme ensemble technique with gradual coopetitive soft gating

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.

[Departement_IRSTEA]Eaux [TR1_IRSTEA]ARCEAU

Bayesian bootstrapping in real-time probabilistic photovoltaic power forecasting

Probabilistic wind power forecasting is an important input in the decision-making process in future electric power grids with large penetrations of renewable generation. Traditional probabilistic wind power forecasting models are trained offline and are then used to make predictions online. However, this strategy cannot make full use of the most recent information during the prediction process. In addition, ensemble learning is recognized as an effective approach for further improving forecasting performance by combining multiple forecasting models. This paper studies an online ensemble approach for probabilistic wind power forecasting by taking full advantage of the most recent information and leveraging the strengths of multiple forecasting models. The online ensemble approach is first formulated as an online convex optimization problem. On this basis, a quantile passive-aggressive regression model is proposed to solve the online convex optimization problem. Case studies and comparisons with other online learning methods are conducted on an open wind power data set from Belgium. Results show that the proposed method outperforms competing methods in terms of pinball loss and Winkler score with high reliability.

The fluctuation and uncertainty of wind power generation bring severe challenges to secure and economic operation of power systems. Because wind power forecasting error is unavoidable, probabilistic forecasting becomes critical to accurately quantifying the uncertainty involved in traditional point forecasts of wind power and to providing meaningful information to conduct risk management in power system operation. This paper proposes a novel direct quantile regression approach to efficiently generate nonparametric probabilistic forecasting of wind power generation combining extreme learning machine and quantile regression. Quantiles with different proportions can be directly produced via an innovatively formulated linear programming optimization model, without dependency on point forecasts. Multistep probabilistic forecasting of 10-min wind power is newly carried out based on real wind farm data from Bornholm Island in Denmark. The superiority of the proposed approach is verified through comparisons with other well-established benchmarks. The proposed approach forms a new artificial neural network-based nonparametric forecasting framework for wind power with high efficiency, reliability, and flexibility, which can be beneficial to various decision-making activities in power systems.

K-nearest neighbors and a kernel density estimator for GEFCom2014 probabilistic wind power forecasting

This paper proposes a privacy-preserving framework that couples probabilistic wind power forecasting with decentralized anomaly detection in complex power networks. We first design an adaptive federated learning (FL) scheme to produce probabilistic forecasts for multiple geographically distributed wind farms while keeping their raw data local. In this scheme, an artificial neural network with quantile regression is trained collaboratively across sites to provide calibrated prediction intervals for wind power outputs. These forecasts are then embedded into an alternating direction method of multipliers (ADMM)-based load-side dispatch and anomaly detection model for decentralized power systems with plug-and-play industrial users. Each monitoring node uses local measurements and neighbor communication to solve a distributed economic dispatch problem, detect abnormal load behaviors, and maintain network consistency without a central coordinator. Experiments on the GEFCom 2014 wind power dataset show that the proposed FL-based probabilistic forecasting method outperforms persistence, local training, and standard FL in RMSE and MAE across multiple horizons. Simulations on IEEE 14-bus and 30-bus systems further verify fast convergence, accurate anomaly localization, and robust operation, indicating the effectiveness of the integrated forecasting–dispatch framework for smart industrial grids with high wind penetration.

In recent years, alternating drought and extreme precipitation events have highlighted the need for subseasonal to seasonal forecasts of the terrestrial water cycle. In particular, predictions of the impacts of dry and wet extremes on subsurface water resources are crucial to provide stakeholders in agriculture, forestry, the water sector, and other fields with information supporting the sustainable use of these resources.In this context, we release an experimental Water Resources Bulletin (https://adapter-projekt.de/bulletin/index.html) four times per year, offering probabilistic forecasts of the total subsurface water storage (TSS) anomaly at a 0.6 km resolution, from the surface down to 60 m depth, for the upcoming seven months across Germany. These seasonal forecasts are generated using the integrated, physics-based hydrological model ParFlow/CLM, forced by 50 ensemble members of the SEAS5 seasonal forecast from the European Centre for Medium-Range Weather Forecasts (ECMWF).To evaluate our forecasts, we evaluated six 7-months probabilistic forecasts covering the vegetation period (March to September) for the years 2018 to 2023 with a reference long-term historical time series based on the same ParFlow/CLM setup. The forecast skill was assessed by comparing these seasonal forecasts to a climatology-based 10-member pseudo-forecast over the 2013–2023 period (using the leave-one-out method), extracted from the reference time series.The monthly Continuous Ranked Probability Skill Score (CRPSS), which evaluates the ensemble distribution based on daily TSS data, indicates that the probabilistic forecast outperforms the climatology-based pseudo-forecast in most regions, except in 2018 and, to a lesser extent, in 2020 and 2022. This can be attributed to an under-representation of extremely dry members in the ensemble, combined with the memory effect of the initial conditions at increasing soil depths. For example, while March 2018 started with a slightly above-average TSS and experienced a strong meteorological drought leading to an agricultural drought, the initial TSS anomaly in March 2019 was already negative, with a less pronounced precipitation deficit during the vegetation period. This resulted in a much higher forecast skill, because of the memory effect accurately simulated with the physics-based model. Notably, the forecast skill only slightly decreases with increasing lead time, both for precipitation and TSS.The analysis of the Relative Operating Characteristic Skill Score (ROCSS) for the lower quintile of the TSS distribution assesses whether negative TSS anomalies (i.e., droughts) are adequately represented within the probabilistic forecast ensemble. The results are consistent with those of the CRPSS, showing lower skill in 2018. Nevertheless, the ROCSS analysis overall indicates moderate to high skill for the probabilistic forecast, while the climatology-based pseudo-forecast demonstrates no skill. This confirms that the dry conditions experienced in central Europe in recent years were captured within the probabilistic forecast, underlining the added value of these forecasts and their usefulness in the experimental Water Resources Bulletin.

Machine learning techniques are increasingly used in weather forecasting, to either improve results from numerical weather prediction (NWP) models or emulate some aspects of the forward integration part of NWP models using a purely data-driven approach. So far most of the techniques presented have focused on producing deterministic forecasts, as many standard machine learning techniques lack the ability to express uncertainty.   Whilst methods that attempt to fully emulate NWP models are both very complex and computationally expensive, in this talk I will show how much simpler and computationally cheaper techniques can be used to create a probabilistic forecast from a single global high-resolution forecast by post-processing. Specifically, I will focus on the relatively novel technique of a Bayesian Neural Network and show how it can predict the distribution of the forecast error relative to its own analysis. By adding these error distributions to the original forecast, we can create a probabilistic forecast. Time permitting, I will discuss how more advanced techniques can be applied to take these distributions and generate spatially consistent ensemble members.  This methodology is particularly useful for NWP models at very high resolutions where running an ensemble is too computationally expensive and for machine learning approaches where no uncertainty information is available. In this talk, I will show how this methodology can be successfully applied to both ECMWF’s high-resolution forecast and a purely data-driven weather forecast model being run at ECMWF, for both the surface variable of 2m temperature and the atmospheric variable of Z500. These probabilistic forecasts have been verified using standard metrics and, in the case of the high-resolution numerical forecast, have as good as or better CRPS scores than the ECMWF ensemble forecast for the lead times tested. Moreover, these probabilistic forecasts are reliable with spread-skill ratios close to one. Hence, this novel machine learning post-processing technique has the potential to produce probabilistic forecasts that are valuable and useful to forecasters. 

Heat stress can reduce crop yield or even cause total crop failure. The ability to predict heat stress in advance would allow growers time to implement protective measures, helping to avoid such losses. This study presents a strategy for producing probabilistic heat stress forecasts for well‐watered cotton (Gossypium hirsutum L.) in the Camilla, GA, region. Multiple linear regression was used to develop a cotton canopy temperature model based on predicted air temperature, humidity, solar radiation, and wind speed. The European Centre for Medium‐Range Weather Forecasts Ensemble Prediction System was used to predict the meteorological variables used in the canopy temperature model, which produced 10‐d probabilistic canopy temperature forecasts for each day of observations during 2014. A statistical mean bias correction was applied to improve on the raw model forecasts. The forecasts were found to be skillful, with relative operating characteristic areas greater than 0.5. The bias‐corrected forecasts were found to increase skill. A heat stress warning system was then created using the forecasts. Additionally, an economic analysis was performed as an example of how probabilistic forecasts can be used to aid producers with financial decisions pertaining to weather‐related risks. Creating heat stress advisories for crops provide growers time to protect assets. Coupling crop and atmospheric models translates weather data into usable information for growers. Probabilistic canopy temperature forecasts can be used to evaluate on‐farm weather‐related risks.

Winter road maintenance is one of the main tasks for the Washington State Department of Transportation. Anti-icing, that is, the preemptive application of chemicals, is often used to keep the roadways free of ice. Given the preventive nature of anti-icing, accurate predictions of road ice are needed. Currently, anti-icing decisions are usually based on deterministic weather forecasts. However, the costs of the two kinds of errors are highly asymmetric because the cost of a road closure due to ice is much greater than that of taking anti-icing measures. As a result, probabilistic forecasts are needed to optimize decision making.We propose two methods for forecasting the probability of ice formation. Starting with deterministic numerical weather predictions, we model temperature and precipitation using distributions centered around the bias-corrected forecasts. This produces a joint predictive probability distribution of temperature and precipitation, which then yields the probability of ice formation, defined here as the occurrence of precipitation when the temperature is below freezing. The first method assumes that temperatures, as well as precipitation, at different spatial locations are conditionally independent given the numerical weather predictions. The second method models the spatial dependence between forecast errors at different locations. The model parameters are estimated using a Bayesian approach via Markov chain Monte Carlo.We evaluate both methods by comparing their probabilistic forecasts with observations of ice formation for Interstate Highway 90 in Washington State for the 2003–2004 and 2004–2005 winter seasons. The use of the probabilistic forecasts reduces costs by about 50% when compared to deterministic forecasts. The spatial method improves the reliability of the forecasts, but does not result in further cost reduction when compared to the first method.

This study applied machine learning methods to perform the probabilistic forecasting of coastal wave height during the typhoon warning period. The probabilistic forecasts comprise a deterministic forecast and the probability distribution of a forecast error. A support vector machine was used to develop a real-time forecasting model for generating deterministic wave height forecasts. The forecast errors of deterministic forecasting were then used as a database to generate probabilistic forecasts by using the modified fuzzy inference model. The innovation of the modified fuzzy inference model includes calculating the similarity of the data by performing fuzzy implication and resampling the potential data from the fuzzy database for probability distribution. The probabilistic forecasting method was applied to the east coast of Taiwan, where typhoons frequently cause large waves. Hourly wave height data from an offshore buoy and various typhoon characteristics were used as inputs of the probabilistic forecasting model. Validation results from real typhoon events verified that the proposed probabilistic forecasting model can generate the predicted confidence interval, which can properly enclose the observed wave height data, excluding some cases with extreme wave heights. Moreover, an objective measure was used to validate the proposed probabilistic forecasting method.

/ A method adapted from the National Weather Service's Extended Streamflow Prediction technique is applied retrospectively to three Great Lakes case studies to show how risk assessment using probabilistic monthly water level forecasts could have contributed to the decision-mak-ing process. The first case study examines the 1985 International Joint Commission (IJC) decision to store water in Lake Superior to reduce high levels on the downstream lakes. Probabilistic forecasts are generated for Lake Superior and Lakes Michigan-Huron and used with riparian inundation value functions to assess the relative impacts of the IJC's decision on riparian interests for both lakes. The second case study evaluates the risk of flooding at Milwaukee, Wisconsin, and the need to implement flood-control projects if Lake Michigan levels were to continue to rise above the October 1986 record. The third case study quantifies the risks of impaired municipal water works operation during the 1964-1965 period of extreme low water levels on Lakes Huron, St. Clair, Erie, and Ontario. Further refinements and other potential applications of the probabilistic forecast technique are discussed.KEY WORDS: Great Lakes; Water levels; Forecasting; Risk; Decision making

Subseasonal weather forecasts are becoming increasingly important for a range of socioeconomic activities. However, the predictive ability of physical weather models is very limited on these time scales. We propose four postprocessing methods based on convolutional neural networks to improve subseasonal forecasts by correcting systematic errors of numerical weather prediction models. Our postprocessing models operate directly on spatial input fields and are therefore able to retain spatial relationships and to generate spatially homogeneous predictions. They produce global probabilistic tercile forecasts for biweekly aggregates of temperature and precipitation for weeks 3–4 and 5–6. In a case study based on a public forecasting challenge organized by the World Meteorological Organization, our postprocessing models outperform the bias-corrected forecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF), and achieve improvements over climatological forecasts for all considered variables and lead times. We compare several model architectures and training modes and demonstrate that all approaches lead to skillful and well-calibrated probabilistic forecasts. The good calibration of the postprocessed forecasts emphasizes that our postprocessing models reliably quantify the forecast uncertainty based on deterministic input information in the form of ECMWF ensemble mean forecast fields only.

A seamless prediction of convective precipitation for a continuous range of lead times from 0–8 h requires the application of different approaches. Here, a nowcasting method and a high‐resolution numerical weather prediction ensemble are combined to provide probabilistic precipitation forecasts. For the nowcast, an existing deterministic extrapolation technique was modified by the local Lagrangian method to calculate the probability of exceeding a threshold value in radar reflectivity. Numerical forecasts were obtained from an experimental high‐resolution ensemble that provides 20 different deterministic forecasts of synthetic radar reflectivity. Probabilistic information was calculated by different approaches from the ensemble output. The probabilistic forecasts based on the ensemble were calibrated with the reliability diagram statistics method. The skill of the probabilistic nowcasts and forecasts was evaluated using three quality measures. Finally, a seamless probabilistic forecast was generated as an additive combination of nowcast and forecast, using a weighting function based on their relative skills. The skill of the seamless forecast was greater than or equal to that of the nowcast or ensemble forecast in all quality measures and at all lead times. Copyright © 2011 Royal Meteorological Society