scholarly journals Investigation of Forecast Accuracy and its Impact on the Efficiency of Data-Driven Forecast-Based Reservoir Operating Rules

Water ◽  
2021 ◽  
Vol 13 (19) ◽  
pp. 2737
Author(s):  
Ehsan Mostaghimzadeh ◽  
Seyed Mohammad Ashrafi ◽  
Arash Adib ◽  
Zong Woo Geem
Keyword(s):  
Forecast Accuracy ◽  
Forecast Model ◽  
Data Driven ◽  
Operation Performance ◽  
Hedging Rule ◽  
Variable Flow ◽  
Rule Curves ◽  
Operating Rules ◽  
Forecast Models ◽  
Hedging Rule Curves

Today, variable flow pattern, which uses static rule curves, is considered one of the challenges of reservoir operation. One way to overcome this problem is to develop forecast-based rule curves. However, managers must have an estimate of the influence of forecast accuracy on operation performance due to the intrinsic limitations of forecast models. This study attempts to develop a forecast model and investigate the effects of the corresponding accuracy on the operation performance of two conventional rule curves. To develop a forecast model, two methods according to autocorrelation and wrapper-based feature selection models are introduced to deal with the wavelet components of inflow. Finally, the operation performances of two polynomial and hedging rule curves are investigated using forecasted and actual inflows. The results of applying the model to the Dez reservoir in Iran visualized that a 4% improvement in the correlation coefficient of the coupled forecast model could reduce the relative deficit of the polynomial rule curve by 8.1%. Moreover, with 2% and 10% improvement in the Willmott and Nash—Sutcliffe indices, the same 8.1% reduction in the relative deficit can be expected. Similar results are observed for hedging rules where increasing forecast accuracy decreased the relative deficit by 15.5%. In general, it was concluded that hedging rule curves are more sensitive to forecast accuracy than polynomial rule curves are.

Using Local Information to Improve Short-Run Corn Price Forecasts

2018 ◽  
Vol 16 (1) ◽  
Author(s):  
Xiaojie Xu
Keyword(s):  
Forecast Accuracy ◽  
Forecast Model ◽  
Futures Prices ◽  
Forecast Horizon ◽  
Data Set ◽  
Short Run ◽  
Forecast Models ◽  
History Of ◽  
Predictive Content ◽  
Market Density

AbstractWe examine the short-run forecasting problem in a data set of daily prices from 134 corn buying locations from seven states – Iowa, Illinois, Indiana, Ohio, Minnesota, Nebraska, and Kansas. We ask the question: is there useful forecasting information in the cash bids from nearby markets? We use several criteria, including a Granger causality criterion, to specify forecast models that rely on the recent history of a market, the recent histories of nearby markets, and the recent histories of futures prices. For about 65% of the markets studied, the model consisting of futures prices, a market’s own history, and the history of nearby markets forecasts better than a model only incorporating futures prices and the market’s own history. That is, nearby markets have predictive content. But the magnitude varies with the forecast horizon. For short-run forecasts, the forecast accuracy improvement from including nearby markets is modest. As the forecast horizon increases, however, including nearby prices tends to significantly improve forecasts. We also examine the role played by physical market density in determining the value of incorporating nearby prices into a forecast model.

scholarly journals Supervised learning from noisy observations: Combining machine-learning techniques with data assimilation

2021 ◽  
Author(s):  
Georg Gottwald ◽  
Sebastian Reich
Keyword(s):  
Machine Learning ◽  
Data Assimilation ◽  
Forecast Model ◽  
Machine Learning Algorithms ◽  
Data Driven ◽  
Machine Learning Techniques ◽  
Probabilistic Forecasting ◽  
Feature Maps ◽  
Data Driven Approach ◽  
Forecast Models

<p>Data-driven prediction and physics-agnostic machine-learning methods have attracted increased interest in recent years achieving forecast horizons going well beyond those to be expected for chaotic dynamical systems.<span>  </span>In a separate strand of research data-assimilation has been successfully used to optimally combine forecast models and their inherent uncertainty with incoming noisy observations. The key idea in our work here is to achieve increased forecast capabilities by judiciously combining machine-learning algorithms and data assimilation. We combine the physics-agnostic data-driven approach of random feature maps as a forecast model within an ensemble Kalman filter data assimilation procedure. The machine-learning model is learned sequentially by incorporating incoming noisy observations. We show that the obtained forecast model has remarkably good forecast skill while being computationally cheap once trained. Going beyond the task of forecasting, we show that our method can be used to generate reliable ensembles for probabilistic forecasting as well as to learn effective model closure in multi-scale systems.</p>

scholarly journals Sensitivities of the Madden–Julian oscillation forecasts to configurations of physics in the ECMWF global model

2021 ◽  
Vol 21 (6) ◽  
pp. 4759-4778
Author(s):  
Jun-Ichi Yano ◽  
Nils P. Wedi
Keyword(s):  
Forecast Model ◽  
Global Model ◽  
Madden Julian Oscillation ◽  
Forecast Performance ◽  
Theoretical Investigations ◽  
Forecast Models ◽  
Forecasting System ◽  
The Rich ◽  
The Tropics ◽  
Original Motivation

Abstract. The sensitivities of the Madden–Julian oscillation (MJO) forecasts to various different configurations of the parameterized physics are examined with the global model of ECMWF's Integrated Forecasting System (IFS). The motivation for the study was to simulate the MJO as a nonlinear free wave under active interactions with higher-latitude Rossby waves. To emulate free dynamics in the IFS, various momentum-dissipation terms (“friction”) as well as diabatic heating were selectively turned off over the tropics for the range of the latitudes from 20∘ S to 20∘ N. The reduction of friction sometimes improves the MJO forecasts, although without any systematic tendency. Contrary to the original motivation, emulating free dynamics with an operational forecast model turned out to be rather difficult, because forecast performance sensitively depends on the specific type of friction turned off. The result suggests the need for theoretical investigations that much more closely follow the actual formulations of model physics: a naive approach with a dichotomy of with or without friction simply fails to elucidate the rich behaviour of complex operational models. The paper further exposes the importance of physical processes other than convection for simulating the MJO in global forecast models.

Multireanalysis Comparison of Variability in Column Water Vapor and Its Analysis Increment Associated with the Madden–Julian Oscillation

2015 ◽  
Vol 28 (2) ◽  
pp. 793-808 ◽  
Author(s):  
Satoru Yokoi
Keyword(s):  
Water Vapor ◽  
Forecast Model ◽  
Precipitation Anomaly ◽  
Boreal Winter ◽  
Madden Julian Oscillation ◽  
Radiative Feedback ◽  
Feedback Strength ◽  
Advection Term ◽  
Forecast Models ◽  
Cloud Radiative Feedback

Abstract This study conducts a multireanalysis comparison of variability in column water vapor (CWV) represented in three reanalysis products [Japanese 55-year Reanalysis Project (JRA-55), JRA-25, and ECMWF Interim Re-Analysis (ERA-Interim)] associated with the Madden–Julian oscillation (MJO) in boreal winter, with emphasis on CWV tendency simulated by forecast models and analysis increment calculated by data assimilation systems. Analyses of these variables show that, while the JRA-55 forecast model is able to simulate eastward propagation of the CWV anomaly, this model tends to weaken its amplitude. The multireanalysis comparison of the analysis increment further reveals that this weakening bias is related to excessively weak cloud radiative feedback represented by the model. This bias in the feedback strength makes anomalous moisture supply by the vertical advection term in the CWV budget equation too insensitive to precipitation anomaly, resulting in reduction of the amplitude of CWV anomaly. ERA-Interim has a nearly opposite feature: the forecast model represents excessively strong feedback. These results imply the necessity of accurate representation of the cloud radiative feedback strength for a short-term MJO forecast and may be evidence to support the argument that this feedback is essential for the existence of MJO. Furthermore, this study demonstrates that the multireanalysis comparison of the analysis increment will provide useful information for examining model biases and potentially for estimating parameters that are difficult to estimate from observational data, such as gross moist stability.

scholarly journals Validation of Potato Early Blight Disease Forecast Models for Colorado Using Various Sources of Meteorological Data

Plant Disease ◽  
2003 ◽  
Vol 87 (1) ◽  
pp. 78-84 ◽  
Author(s):  
David H. Gent ◽  
Howard F. Schwartz
Keyword(s):  
Early Blight ◽  
Meteorological Data ◽  
Forecast Accuracy ◽  
Weather Data ◽  
Hourly Temperature ◽  
Initial Appearance ◽  
Forecast Models ◽  
Weather Stations ◽  
Disease Forecast ◽  
Meteorological Observations

Disease forecasts from regional or remotely sensed meteorological data free growers from infield weather data monitoring and may improve disease forecast implementation. This study was initiated to validate potato early blight forecast models in Colorado and to determine the influence of sources of meteorological data on forecast accuracy. Hourly temperatures were recorded by Campbell Scientific CR-10, Pessl Instruments μMetos Model MCR300, and Spectrum Technologies Model 450 WatchDog weather stations and data loggers within potato fields, field-specific temperature estimations generated by mPOWER3/EMERGE from off-site weather stations, and regional COAGMET CR-10 weather stations. Mean hourly temperature deviations between mPOWER3/EMERGE or in-field stations and COAGMET varied from 0.93°C greater to 1.11°C less than COAGMET observations. Initial appearance of early blight lesions was predicted using a 300 physiological day threshold in commercial fields in each year from 1998 to 2001 and in experimental plots in each year from 1997 to 2001 as determined by COAGMET meteorological observations. All sources of meteorological data generated early blight forecasts within 6 days of each other across all locations and years. COAGMET weather stations should free potato growers and integrated pest management personnel from collecting in-field microclimatic data and speed the implementation of disease forecasting.

A Method of Improved Support Vector Machine for Network Security Situation Forecasting

2011 ◽  
Vol 187 ◽  
pp. 291-296
Author(s):  
Yuan Cheng Li ◽  
Jing Tao Jing
Keyword(s):  
Network Security ◽  
Forecast Accuracy ◽  
Forecast Model ◽  
Selection Method ◽  
Support Vector ◽  
Swarm Optimization ◽  
Parameters Selection ◽  
Vector Machines ◽  
Good Forecast ◽  
Improved Support Vector Machine

Aiming at the problem that parameters of Support Vector Machines (SVM) are very difficult to confirm, this paper points out a parameter selection method for SVM based on Particle Swarm Optimization (PSO), which can make the SVM more scientific and reasonable in parameters selection; and thus enhance the forecast accuracy of the network security situation. The Simulation results show that the optimized SVR forecast model has good forecast accuracy for the network security situation, and present the future changing at a macro level, then help the network managers control network.

Prediction Model of Least Squares Support Vector Machine of Increased Memory Type Based on GA and Quadratic Renyi-Entropy

2011 ◽  
Vol 219-220 ◽  
pp. 754-761
Author(s):  
Guan Hua Zhao ◽  
Wen Wen Yan
Keyword(s):  
Support Vector Machine ◽  
Prediction Model ◽  
Least Squares ◽  
Financial Distress ◽  
Forecast Accuracy ◽  
Forecast Model ◽  
Renyi Entropy ◽  
Rényi Entropy ◽  
Support Vector ◽  
Memory Type

In order to improve the accuracy of financial achievement, this paper applies a new forecast model of the Increased memory type least squares support vector machine base on neighborhood rough set and quadratic Renyi-entropy on the basis of the traditional support vector machine prediction model. The paper also independently derives the entropy fit for the financial distress prediction which is in discrete sequence, as well as the expression of support vector machine kernel function. The experimental results show that the improved model is significantly superior to the traditional LS-SVM as well as the standard support vector machine prediction model, regardless of the forecast accuracy , training samples number.

Sign in / Sign up

Export Citation Format

Share Document