Physics-Informed AI Surrogates for Day-Ahead Wind Power Probabilistic Forecasting with Incomplete Data for Smart Grid in Smart Cities

Due to the high inherent uncertainty of renewable energy,probabilistic day-ahead wind power forecasting is crucial for modeling and controlling the uncertainty of renewable energy smart grids in smart cities.However,the accuracy and reliability of high-resolution day-ahead wind power forecasting are constrained by unreliable local weather prediction and incomplete power generation data.This article proposes a physics-informed artificial intelligence(AI)surrogates method to augment the incomplete dataset and quantify its uncertainty to improve wind power forecasting performance.The incomplete dataset,built with numerical weather prediction data,historical wind power generation,and weather factors data,is augmented based on generative adversarial networks.After augmentation,the enriched data is then fed into a multiple AI surrogates model constructed by two extreme learning machine networks to train the forecasting model for wind power.Therefore,the forecasting models’accuracy and generalization ability are improved by mining the implicit physics information from the incomplete dataset.An incomplete dataset gathered from a wind farm in North China,containing only 15 days of weather and wind power generation data withmissing points caused by occasional shutdowns,is utilized to verify the proposed method’s performance.Compared with other probabilistic forecastingmethods,the proposed method shows better accuracy and probabilistic performance on the same incomplete dataset,which highlights its potential for more flexible and sensitive maintenance of smart grids in smart cities.

Probabilistic modeling of multifunction radars with autoregressive kernel mixture network

The task of modeling and analyzing intercepted multifunction radars(MFRs)pulse trains is vital for cognitive electronic reconnaissance.Existing methodologies predominantly rely on prior information or heavily constrained models,posing challenges for non-cooperative applications.This paper introduces a novel approach to model MFRs using a Bayesian network,where the conditional probability density function is approximated by an autoregressive kernel mixture network(ARKMN).Utilizing the estimated probability density function,a dynamic programming algorithm is proposed for denoising and detecting change points in the intercepted MFRs pulse trains.Simulation results affirm the proposed method's efficacy in modeling MFRs,outperforming the state-of-the-art in pulse train denoising and change point detection.

Nash Model Parameter Uncertainty Analysis by AM-MCMC Based on BFS and Probabilistic Flood Forecasting

A hydrologic model consists of several parameters which are usually calibrated based on observed hy-drologic processes. Due to the uncertainty of the hydrologic processes, model parameters are also uncertain, which further leads to the uncertainty of forecast results of a hydrologic model. Working with the Bayesian Forecasting System (BFS), Markov Chain Monte Carlo simulation based Adaptive Metropolis method (AM-MCMC) was used to study parameter uncertainty of Nash model, while the probabilistic flood forecasting was made with the simu-lated samples of parameters of Nash model. The results of a case study shows that the AM-MCMC based on BFS proposed in this paper is suitable to obtain the posterior distribution of the parameters of Nash model according to the known information of the parameters. The use of Nash model and AM-MCMC based on BFS was able to make the probabilistic flood forecast as well as to find the mean and variance of flood discharge, which may be useful to estimate the risk of flood control decision.

Probabilistic Residential Load Forecasting with Sequence-to-sequence Adversarial Domain Adaptation Networks

Lately,the power demand of consumers is increasing in distribution networks,while renewable power generation keeps penetrating into the distribution networks.Insufficient data make it hard to accurately predict the new residential load or newly built apartments with volatile and changing time-series characteristics in terms of frequency and magnitude.Hence,this paper proposes a short-term probabilistic residential load forecasting scheme based on transfer learning and deep learning techniques.First,we formulate the short-term probabilistic residential load forecasting problem.Then,we propose a sequence-to-sequence(Seq2Seq)adversarial domain adaptation network and its joint training strategy to transfer generic features from the source domain(with massive consumption records of regular loads)to the target domain(with limited observations of new residential loads)and simultaneously minimize the domain difference and forecasting errors when solving the forecasting problem.For implementation,the dominant techniques or elements are used as the submodules of the Seq2Seq adversarial domain adaptation network,including the Seq2Seq recurrent neural networks(RNNs)composed of a long short-term memory(LSTM)encoder and an LSTM decoder,and quantile loss.Finally,this study conducts the case studies via multiple evaluation indices,comparative methods of classic machine learning and advanced deep learning,and various available data of the new residentical loads and other regular loads.The experimental results validate the effectiveness and stability of the proposed scheme.

Monthly and seasonal streamflow forecasting of large dryland catchments in Brazil

Streamflow forecasting in drylands is challenging.Data are scarce,catchments are highly humanmodified and streamflow exhibits strong nonlinear responses to rainfall.The goal of this study was to evaluate the monthly and seasonal streamflow forecasting in two large catchments in the Jaguaribe River Basin in the Brazilian semi-arid area.We adopted four different lead times:one month ahead for monthly scale and two,three and four months ahead for seasonal scale.The gaps of the historic streamflow series were filled up by using rainfall-runoff modelling.Then,time series model techniques were applied,i.e.,the locally constant,the locally averaged,the k-nearest-neighbours algorithm(k-NN)and the autoregressive(AR)model.The criterion of reliability of the validation results is that the forecast is more skillful than streamflow climatology.Our approach outperformed the streamflow climatology for all monthly streamflows.On average,the former was 25%better than the latter.The seasonal streamflow forecasting(SSF)was also reliable(on average,20%better than the climatology),failing slightly only for the high flow season of one catchment(6%worse than the climatology).Considering an uncertainty envelope(probabilistic forecasting),which was considerably narrower than the data standard deviation,the streamflow forecasting performance increased by about 50%at both scales.The forecast errors were mainly driven by the streamflow intra-seasonality at monthly scale,while they were by the forecast lead time at seasonal scale.The best-fit and worst-fit time series model were the k-NN approach and the AR model,respectively.The rainfall-runoff modelling outputs played an important role in improving streamflow forecasting for one streamgauge that showed 35%of data gaps.The developed data-driven approach is mathematical and computationally very simple,demands few resources to accomplish its operational implementation and is applicable to other dryland watersheds.Our findings may be part of drought forecasting systems and potentially help allocating water months in advance.Moreover,the developed strategy can serve as a baseline for more complex streamflow forecast systems.

Probabilistic Wind Gust Forecasting during the 2022 Beijing Winter Olympics

The probabilistic forecast of wind gusts poses a significant challenge during the post-processing of numerical model outputs.Comparative analysis of probabilistic forecasting methods plays a crucial role in enhancing forecast accuracy.Within the context of meteorological services for alpine skiing at the 2022 Beijing Winter Olympics,The ECMWF ensemble products were used to evaluate six post-processing methods.These methods include ensemble model output statistics(EMOS),backpropagation neural networks(BP),particle swarm optimization algorithms with backpropagation neural networks(PSO),truncated normal distributions,truncated logarithmic distributions,and generalized extreme value(GEV) distributions.The performance of these methods in predicting gust probabilities at five observation points along a ski track was compared.All six methods exhibited a substantial reduction in forecast errors compared to the original ECMWF products;however,the ability to correct the model forecast results varied significantly across different wind speed ranges.Specifically,the EMOS,truncated normal distribution,truncated logarithmic distribution,and GEV distribution demonstrated advantages in low wind-speed ranges,whereas the BP and PSO methods exhibit lower forecast errors for high wind-speed events.Furthermore,this study affirms the rationality of utilizing the statistical characteristics derived from ensemble forecasts as probabilistic forecast factors.The application of probability integral transform(PIT) and quantile–quantile(QQ) plots demonstrates that gust variations at the majority of observation sites conform to the GEV distribution,thereby indicating the potential for further enhanced forecast accuracy.The results also underscore the significant utility of the PSO hybrid model,which amalgamates particle swarm optimization with a BP neural network,in the probabilistic forecasting of strong winds within the field of meteorology.

Forecasting tropical cyclone coastal and marine hazards and impacts

The report highlights the significant progress over various regions with respect to understanding of coastal hazards,numerical modeling techniques and the generation&dissemination of coastal hazard warnings and products.The developments over various regions in the globe during 2014–18 have been discussed in this report as presented during 10th Session of International Workshop on Tropical Cyclones(IWTC-X)at Bali,Indonesia.More specifically,various regions have started to confront the uncertainty that cannot be removed from TC analyses and forecasts and further communicate those hazards within the context of risk[probabilistic]based information.Progress also includes impact-based forecasts such as communicating coastal inundation information relative to total water level instead of storm surge,specifically(i.e.,anomaly from as-tronomical tide and waves).Lastly,updates to model grid configuration,model resolution,and coupled dynamical systems continue to resolve the costal hazards more effectively.Those approaches have likely helped reduce loss of life relative to historical standards.However,regions agree that the generation and dissemination of coastal hazard information still need to be improved in view of growing population along the coast and thus increased exposure of life to coastal hazard.

Bayesian bootstrap quantile regression for probabilistic photovoltaic power forecasting

Photovoltaic(PV)systems are widely spread across MV and LV distribution systems and the penetration of PV generation is solidly growing.Because of the uncertain nature of the solar energy resource,PV power forecasting models are crucial in any energy management system for smart distribution networks.Although point forecasts can suit many scopes,probabilistic forecasts add further flexibility to an energy management system and are recommended to enable a wider range of decision making and optimization strategies.This paper proposes methodology towards probabilistic PV power forecasting based on a Bayesian bootstrap quantile regression model,in which a Bayesian bootstrap is applied to estimate the parameters of a quantile regression model.A novel procedure is presented to optimize the extraction of the predictive quantiles from the bootstrapped estimation of the related coefficients,raising the predictive ability of the final forecasts.Numerical experiments based on actual data quantify an enhancement of the performance of up to 2.2%when compared to relevant benchmarks.

A hybrid ensemble deep reinforcement learning model for locomotive axle temperature using the deterministic and probabilistic strategy

This paper proposes a hybrid deep reinforcement learning framework for locomotive axle temperature by combining the wavelet packet decomposition(WPD),long short-term memory(LSTM),gated recurrent unit(GRU)reinforcement learning and generalized autoregressive conditional heteroskedasticity(GARCH)algorithms.The WPD is utilized to decompose the raw nonlinear series into subseries.Then the deep learning predictors LSTM and GRU are established to predict the future axle temperatures in each subseries.The Q-learning could generate optimal ensembleweights to integrate the predictors to finish the deterministic forecasting and GARCH is used to conduct the deterministic forecasting based on the deterministic forecasting residual.These parts of the hybrid ensemble structure contributed to optimal modelling accuracy and provided effective support in the real-time monitoring and fault diagnosis of transportation.

Machinery Condition Prediction Based on Support Vector Machine Model with Wavelet Transform

Soft failure of mechanical equipment makes its performance drop gradually,which occupies a large proportion and has certain regularity. The performance can be evaluated and predicted through early state monitoring and data analysis. The vibration signal was modeled from the double row bearing,and wavelet transform and support vector machine model( WT-SVM model) was constructed and trained for bearing degradation process prediction. Besides Hazen plotting position relationships was applied to describing the degradation trend distribution and a 95%confidence level based on t-distribution was given. The single SVM model and neural network( NN) approach were also investigated as a comparison. Results indicate that the WT-SVM model outperforms the NN and single SVM models,and is feasible and effective in machinery condition prediction.

An Effective Configuration of Ensemble Size and Horizontal Resolution for the NCEP GEFS

Two important questions are addressed in this paper using the Global Ensemble Forecast System （GEFS） from the National Centers for Environmental Prediction （NCEP）： （1） How many ensemble members are needed to better represent forecast uncertainties with limited computational resources？ （2） What is tile relative impact on forecast skill of increasing model resolution and ensemble size？ Two-month experiments at T126L28 resolution were used to test the impact of varying the ensemble size from 5 to 80 members at the 500- hPa geopotential height. Results indicate that increasing the ensemble size leads to significant improvements in the performance for all forecast ranges when measured by probabilistic metrics, but these improvements are not significant beyond 20 members for long forecast ranges when measured by deterministic metrics. An ensemble of 20 to 30 members is the most effective configuration of ensemble sizes by quantifying the tradeoff between ensemble performance and the cost of computational resources. Two representative configurations of the GEFS the T126L28 model with 70 members and the T190L28 model with 20 members, which have equivalent computing costs--were compared. Results confirm that, for the NCEP GEFS, increasing the model resolution is more （less） beneficial than increasing the ensemble size for a short （long） forecast range.

A Regional Ensemble Forecast System for Stratiform Precipitation Events in the Northern China Region.Part Ⅱ:Seasonal Evaluation for Summer 2010

In this study, the Institute of Atmospheric Physics, Chinese Academy of Sciences - regional ensemble forecast system （IAP-REFS） described in Part I was further validated through a 65-day experiment using the summer season of 2010. The verification results show that IAP-REFS is skillful for quantitative precipitation forecasts （QPF） and probabilistic QPF, but it has a systematic bias in forecasting near-surface variables. Applying a 7-day running mean bias correction to the forecasts of near-surface variables remarkably improved the reliability of the forecasts. In this study, the perturbation extraction and inflation method （proposed with the single case study in Part I） was further applied to the full season with different inflation factors. This method increased the ensemble spread and improved the accuracy of forecasts of precipitation and near-surface variables. The seasonal mean profiles of the IAP-REFS ensemble indicate good spread among ensemble members and some model biases at certain vertical levels.

Probabilistic Precipitation Forecasting Based on Ensemble Output Using Generalized Additive Models and Bayesian Model Averaging

A probabilistic precipitation forecasting model using generalized additive models （GAMs） and Bayesian model averaging （BMA） was proposed in this paper. GAMs were used to fit the spatial-temporal precipitation models to individual ensemble member forecasts. The distributions of the precipitation occurrence and the cumulative precipitation amount were represented simultaneously by a single Tweedie distribution. BMA was then used as a post-processing method to combine the individual models to form a more skillful probabilistic forecasting model. The mixing weights were estimated using the expectation-maximization algorithm. The residual diagnostics was used to examine if the fitted BMA forecasting model had fully captured the spatial and temporal variations of precipitation. The proposed method was applied to daily observations at the Yishusi River basin for July 2007 using the National Centers for Environmental Prediction ensemble forecasts. By applying scoring rules, the BMA forecasts were verified and showed better performances compared with the empirical probabilistic ensemble forecasts, particularly for extreme precipitation. Finally, possible improvements and a^plication of this method to the downscaling of climate change scenarios were discussed.

A Regional Ensemble Forecast System for Stratiform Precipitation Events in Northern China.Part I:A Case Study

A single-model, short-range, ensemble forecasting system （Institute of Atmospheric Physics, Regional Ensemble Forecast System, IAP REFS） with 15-km grid spacing, configured with multiple initial conditions, multiple lateral boundary conditions, and multiple physics parameterizations with 11 ensemble members, was developed using the Weather and Research Forecasting Model Advanced Research modeling system for prediction of stratiform precipitation events in northern China. This is the first part of a broader research project to develop a novel cloud-seeding operational system in a probabilistic framework. The ensemble perturbations were extracted from selected members of the National Center for Environmental Prediction Global Ensemble Forecasting System （NCEP GEFS） forecasts, and an inflation factor of two was applied to compensate for the lack of spread in the GEFS forecasts over the research region. Experiments on an actual stratiform precipitation case that occurred on 5-7 June 2009 in northern China were conducted to validate the ensemble system. The IAP REFS system had reasonably good performance in predicting the observed stratiform precipitation system. The perturbation inflation enlarged the ensemble spread and alleviated the underdispersion caused by parent forecasts. Centering the extracted perturbations on higher-resolution NCEP Global Forecast System forecasts resulted in less ensemble mean root-mean-square error and better accuracy in probabilistic quantitative precipitation forecasts （PQPF）. However, the perturbation inflation and recentering had less effect on near-surface-level variables compared to the mid-level variables, and its influence on PQPF resolution was limited as well.

An Assessment of the Predictability of the East Asian Subtropical Westerly Jet Based on TIGGE Data

The predictability of the position,spatial coverage and intensity of the East Asian subtropical westerly jet （EASWJ） in the summers of 2010 to 2012 was examined for ensemble prediction systems （EPSs） from four representative TIGGE centers,including the ECMWF,the NCEP,the CMA,and the JMA.Results showed that each EPS predicted all EASWJ properties well,while the levels of skill of all EPSs declined as the lead time extended.Overall,improvements from the control to the ensemble mean forecasts for predicting the EASWJ were apparent.For the deterministic forecasts of all EPSs,the prediction of the average axis was better than the prediction of the spatial coverage and intensity of the EASWJ.ECMWF performed best,with a lead of approximately 0.5-1 day in predictability over the second-best EPS for all EASWJ properties throughout the forecast range.For probabilistic forecasts,differences in skills among the different EPSs were more evident in the earlier part of the forecast for the EASWJ axis and spatial coverage,while they departed obviously throughout the forecast range for the intensity.ECMWF led JMA by about 0.5-1 day for the EASWJ axis,and by about 1-2 days for the spatial coverage and intensity at almost all lead times.The largest lead of ECMWF over the relatively worse EPSs,such as NCEP and CMA,was approximately 3-4 days for all EASWJ properties.In summary,ECMWF showed the highest level of skill for predicting the EASWJ,followed by JMA.

Impact of Soil Moisture Uncertainty on Summertime Short-range Ensemble Forecasts

To investigate the impact of soil moisture uncertainty on summertime short-range ensemble forecasts（SREFs）, a fivemember SREF experiment with perturbed initial soil moisture（ISM） was performed over a northern China domain in summertime from July to August 2014. Five soil moisture analyses from three different operational/research centers were used as the ISM for the ensemble. The ISM perturbation produced notable ensemble spread in near-surface variables and atmospheric variables below 800 h Pa, and produced skillful ensemble-mean 24-h accumulated precipitation（APCP24） forecasts that outperformed any single ensemble member. Compared with a second SREF experiment with mixed microphysics parameterization options, the ISM-perturbed ensemble produced comparable ensemble spread in APCP24 forecasts, and had better Brier scores and resolution in probabilistic APCP24 forecasts for 10-mm, 25-mm and 50-mm thresholds. The ISM-perturbed ensemble produced obviously larger ensemble spread in near-surface variables. It was, however, still under-dispersed, indicating that perturbing ISM alone may not be adequate in representing all the uncertainty at the near-surface level, indicating further SREF studies are needed to better represent the uncertainties in land surface processes and their coupling with the atmosphere.

Improving the real-time probabilistic channel flood forecasting by incorporating the uncertainty of inflow using the particle filter

An accurate and reliable real-time flood forecast is crucial for mitigating flood disasters. The errors associated with the inflow boundary forcing data are considered as an important source of uncertainties in hydraulic model. In this paper, a real-time probabilistic channel flood forecasting model is developed with a novel function to incorporate the uncertainty of the forcing inflow. This new approach couples a hydraulic model with the particle filter（PF） data assimilation algorithm, a sequential Bayesian Monte Carlo method. The stage observations at hydrological stations are assimilated at each time step to update the model states in order to improve the next time step＇s forecasting. This new approach is tested against a real flood event occurred in the upper Yangtze River. As compared with the open loop simulations, the evaluations of model performance with several deterministic and probabilistic metrics indicate that the accuracy of the ensemble mean prediction and the reliability of the uncertainty quantification are improved pronouncedly as a result of the PF assimilation. Further assessment of the prediction results at different lead times shows that the improvement of model performance deteriorates with the increase of the lead time due to the gradual diminishing of the updating effect for the initial conditions. Based on the analyses of the number of particles and the assimilation frequency, we find that the optimal number of particles can be determined by balancing the model performance and the computation cost, while a high assimilation frequency is preferred to incorporate the emerging observations to update the model states to match the real conditions.

Analyses and Forecasts of a Tornadic Supercell Outbreak Using a 3DVAR System Ensemble

As part of NOAA＇s ＂Warn-On-Forecast＂ initiative, a convective-scale data assimilation and prediction system was developed using the WRF-ARW model and ARPS 3DVAR data assimilation technique. The system was then evaluated using retrospective short-range ensemble analyses and probabilistic forecasts of the tornadic supercell outbreak event that occurred on 24 May 2011 in Oklahoma, USA. A 36-member multi-physics ensemble system provided the initial and boundary conditions for a 3-km convective-scale ensemble system. Radial velocity and reflectivity observations from four WSR-88 Ds were assimilated into the ensemble using the ARPS 3DVAR technique. Five data assimilation and forecast experiments were conducted to evaluate the sensitivity of the system to data assimilation frequencies, in-cloud temperature adjustment schemes, and fixed- and mixed-microphysics ensembles. The results indicated that the experiment with 5-min assimilation frequency quickly built up the storm and produced a more accurate analysis compared with the 10-min assimilation frequency experiment. The predicted vertical vorticity from the moist-adiabatic in-cloud temperature adjustment scheme was larger in magnitude than that from the latent heat scheme. Cycled data assimilation yielded good forecasts, where the ensemble probability of high vertical vorticity matched reasonably well with the observed tornado damage path. Overall, the results of the study suggest that the 3DVAR analysis and forecast system can provide reasonable forecasts of tornadic supercell storms.

Sub-Seasonal Predictability of the Northeast China Cold Vortex in BCC and ECMWF S2S Model Forecasts for 2006-2021

As an important atmospheric circulation system in the mid-high latitudes of East Asia,the Northeast China cold vortex(NCCV)substantially influences weather and climate in this region.So far,systematic assessment on the performance of numerical prediction of the NCCVs has not been carried out.Based on the Beijing Climate Centre(BCC)and the ECMWF model hindcast and forecast data that participated in the Sub-seasonal to Seasonal(S2S)Prediction Project,this study systematically examines the performance of both models in simulating and forecasting the NCCVs at the sub-seasonal timescale.The results demonstrate that the two models can effectively capture the seasonal variations in the intensity,active days,and spatial distribution of NCCVs;however,the duration of NCCVs is shorter and the intensity is weaker in the models than in the observations.Diagnostic analysis shows that the differences in the intensity and location of the East Asian subtropical westerly jet and the wave train pattern from North Atlantic to East Asia may be responsible for the deficient simulation of NCCV events in the S2S models.Nonetheless,in the deterministic forecasts,BCC and ECMWF provide skillful prediction on the anomalous numbers of NCCV days and intensity at a lead time of 4-5(5-6)pentads,and the skill limit of the ensemble mean is 1-2 pentads longer than that of individual members.In the probabilistic forecasts of daily NCCV activities,BCC and ECMWF exhibit a forecasting skill of approximately 7 and 11 days,respectively;both models show seasonal dependency in the simulation performance and forecast skills of NCCV events,with better performance in winter than in summer.The results from this study provide helpful references for further improvement of the S2S prediction of NCCVs.

BMA probability quantitative precipitation forecasting of land-falling typhoons in south-east China

The probability of quantitative precipitation forecast(PQPF)of three Bayesian Model Averaging(BMA)models based on three raw super ensemble prediction schemes(i.e.,A,B,and C)are established,which through calibration of their parameters using 1-3 day precipitation ensemble prediction systems(EPSs)from the China Meteorological Administration(CMA),the European Centre for Medium-Range Weather Forecasts(ECMWF)and the National Centers for Environmental Prediction(NCEP)and observation during land-falling of three typhoons in south-east China in 2013.The comparison of PQPF shows that the performance is better in the BMA than that in raw ensemble forecasts.On average,the mean absolute error(MAE)of 1 day lead time forecast is reduced by 12.4%,and its continuous ranked probability score(CRPS)of 1-3 day lead time forecast is reduced by 26.2%,respectively.Although the amount of precipitation prediction by the BMA tends to be underestimated,but in view of the perspective of probability prediction,the probability of covering the observed precipitation by the effective forecast ranges of the BMA are increased,which is of great significance for the early warning of torrential rain and secondary disasters induced by it.