scholarly journals A Novel Multi-Input Multi-Output Recurrent Neural Network Based on Multimodal Fusion and Spatiotemporal Prediction for 0–4 h Precipitation Nowcasting

Atmosphere ◽  
2021 ◽  
Vol 12 (12) ◽  
pp. 1596
Author(s):  
Fuhan Zhang ◽  
Xiaodong Wang ◽  
Jiping Guan
Keyword(s):  
Neural Network ◽  
Recurrent Neural Network ◽  
Input Data ◽  
Learning Strategy ◽  
Meteorological Data ◽  
Precipitation Amount ◽  
Heavy Precipitation ◽  
Precipitation Intensity ◽  
Multimodal Fusion ◽  
Grid Data

Multi-source meteorological data can reflect the development process of single meteorological elements from different angles. Making full use of multi-source meteorological data is an effective method to improve the performance of weather nowcasting. For precipitation nowcasting, this paper proposes a novel multi-input multi-output recurrent neural network model based on multimodal fusion and spatiotemporal prediction, named MFSP-Net. It uses precipitation grid data, radar echo data, and reanalysis data as input data and simultaneously realizes 0–4 h precipitation amount nowcasting and precipitation intensity nowcasting. MFSP-Net can perform the spatiotemporal-scale fusion of the three sources of input data while retaining the spatiotemporal information flow of them. The multi-task learning strategy is used to train the network. We conduct experiments on the dataset of Southeast China, and the results show that MFSP-Net comprehensively improves the performance of the nowcasting of precipitation amounts. For precipitation intensity nowcasting, MFSP-Net has obvious advantages in heavy precipitation nowcasting and the middle and late stages of nowcasting.

How Often Will It Rain?

2007 ◽  
Vol 20 (19) ◽  
pp. 4801-4818 ◽  
Author(s):  
Ying Sun ◽  
Susan Solomon ◽  
Aiguo Dai ◽  
Robert W. Portmann
Keyword(s):  
Climate Change ◽  
Daily Precipitation ◽  
Precipitation Amount ◽  
Heavy Precipitation ◽  
Precipitation Intensity ◽  
First Century ◽  
Simultaneous Increase ◽  
Percentage Increase ◽  
Precipitation Frequency ◽  
Twenty First Century

Abstract Daily precipitation data from climate change simulations using the latest generation of coupled climate system models are analyzed for potential future changes in precipitation characteristics. For the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) B1 (a low projection), A1B (a medium projection), and A2 (a high projection) during the twenty-first century, all the models consistently show a shift toward more intense and extreme precipitation for the globe as a whole and over various regions. For both SRES B1 and A2, most models show decreased daily precipitation frequency and all the models show increased daily precipitation intensity. The multimodel averaged percentage increase in the precipitation intensity (2.0% K−1) is larger than the magnitude of the precipitation frequency decrease (−0.7% K−1). However, the shift in precipitation frequency distribution toward extremes results in large increases in very heavy precipitation events (>50 mm day−1), so that for very heavy precipitation, the percentage increase in frequency is much larger than the increase in intensity (31.2% versus 2.4%). The climate model projected increases in daily precipitation intensity are, however, smaller than that based on simple thermodynamics (∼7% K−1). Multimodel ensemble means show that precipitation amount increases during the twenty-first century over high latitudes, as well as over currently wet regions in low- and midlatitudes more than other regions. This increase mostly results from a combination of increased frequency and intensity. Over the dry regions in the subtropics, the precipitation amount generally declines because of decreases in both frequency and intensity. This indicates that wet regions may get wetter and dry regions may become drier mostly because of a simultaneous increase (decrease) of precipitation frequency and intensity.

scholarly journals Contrasting Daytime and Nighttime Precipitation Variability between Observations and Eight Reanalysis Products from 1979 to 2014 in China

2017 ◽  
Vol 30 (16) ◽  
pp. 6443-6464 ◽  
Author(s):  
Chunlüe Zhou ◽  
Kaicun Wang
Keyword(s):  
Precipitation Amount ◽  
Heavy Precipitation ◽  
Precipitation Intensity ◽  
Southeastern China ◽  
Precipitation Frequency ◽  
Climate Forecast System Reanalysis ◽  
Negative Center ◽  
Significant Downward Trend ◽  
Light Precipitation ◽  
Positive Center

Daytime (0800–2000 Beijing time) and nighttime (2000–0800 Beijing time) precipitation at approximately 2100 stations in China from 1979 to 2014 was used to evaluate eight current reanalyses. Daytime, nighttime, and nighttime–daytime contrast of precipitation were examined in aspects of climatology, seasonal cycle, interannual variability, and trends. The results show that the ECMWF interim reanalysis (ERA-Interim), ERA-Interim/Land, Japanese 55-year Reanalysis (JRA-55), and NCEP Climate Forecast System Reanalysis (CFSR) can reproduce the observed spatial pattern of nighttime–daytime contrast in precipitation amount, exhibiting a positive center over the eastern Tibetan Plateau and a negative center over southeastern China. All of the reanalyses roughly reproduce seasonal variations of nighttime and daytime precipitation, but not always nighttime–daytime contrast. The reanalyses overestimate drizzle and light precipitation frequencies by greater than 31.5% and underestimate heavy precipitation frequencies by less than −30.8%. The reanalyses successfully reproduce interannual synchronizations of daytime and nighttime precipitation frequencies and amounts with an averaged correlation coefficient r of 0.66 against the observed data but overestimate their year-to-year amplitudes by approximately 64%. The trends in nighttime, daytime, and nighttime–daytime contrast of the observed precipitation amounts are mainly dominated by their frequencies ( r = 0.85). Less than moderate precipitation frequency has exhibited a significant downward trend (−2.5% decade−1 during nighttime and −1.7% decade−1 during daytime) since 1979, which is roughly captured by the reanalyses. However, only JRA-55 captures the observed trend of nighttime precipitation intensity (2.4% decade−1), while the remaining reanalyses show negative trends. Overall, JRA-55 and CFSR provide the best reproductions of the observed nighttime–daytime contrast in precipitation intensity, although they have considerable room for improvement.

scholarly journals Comparison Between Random Forest and Recurrent Neural Network for Photovoltaic Power Forecasting

2021 ◽  
Vol 43 (5) ◽  
pp. 347-355
Author(s):  
Ramek Kim ◽  
Kyungmin Kim ◽  
Johng-Hwa Ahn
Keyword(s):  
Neural Network ◽  
Power Generation ◽  
Random Forest ◽  
Recurrent Neural Network ◽  
Power Plants ◽  
Short Term Memory ◽  
Meteorological Data ◽  
Cloud Amount ◽  
Coefficient Of Determination ◽  
Photovoltaic Power

Objectives : Photovoltaic power generation which significantly depends on meteorological conditions is intermittent and unstable. Therefore, accurate forecasting of photovoltaic power generation is a challenging task. In this research, random forest (RF), recurrent neural network (RNN), long short term memory (LSTM), and gated recurrent unit (GRU) are proposed and we will find an efficient model for forecasting photovoltaic power generation of photovoltaic power plants.Methods : We used photovoltaic power generation data from photovoltaic power plants at Gamcheonhang-ro, Saha-gu, Busan, and meteorological data from Busan Regional Meteorological Administration. We used solar irradiance, temperature, atmospheric pressure, humidity, wind speed, wind direction, duration of sunshine, and cloud amount as input variables. By applying the trial and error method, we optimized hyperparameters such as estimators in RF, and number of hidden layers, number of nodes, epochs, and validation split in RNN, LSTM, and GRU. We compared proposed models by evaluation indexes such as coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE).Results and Discussion : The best RF at 1,000 of number of decision tree achieved test R2=0.865, test RMSE=16.013, and test MAE=9.656. The best choice of RNN was 6 hidden layers and the number of nodes in each layer was 90. We set the epochs at 450. RNN achieved test R2=0.942, test RMSE=10.530, and test MAE=6.390. To find the best result of LSTM, we used 3 hidden layers, and the number of nodes was 600. The epochs were set to 200. LSTM achieved test R2=0.944, test RMSE=10.29, and test MAE=6.360. GRU was set to 3 hidden layer and the number of nodes was 450. The epochs were set to 500. GRU achieved test R2=0.945, test RMSE=10.189, and test MAE=5.968.Conclusions : We found RNN, LSTM, and GRU performed better than RF, and GRU model showed the best performance. Therefore, GRU is the most efficient model to predict photovoltaic power generation in Busan, Korea.

Research on behavior recognition based on feature fusion of automatic coder and recurrent neural network

2020 ◽  
Vol 39 (6) ◽  
pp. 8927-8935
Author(s):  
Bing Zheng ◽  
Dawei Yun ◽  
Yan Liang
Keyword(s):  
Neural Network ◽  
Recurrent Neural Network ◽  
Behavior Pattern ◽  
Rapid Development ◽  
Video Data ◽  
Support Vector ◽  
Behavior Recognition ◽  
Learning Methods ◽  
The Impact ◽  
Internet Of Things Technology

Under the impact of COVID-19, research on behavior recognition are highly needed. In this paper, we combine the algorithm of self-adaptive coder and recurrent neural network to realize the research of behavior pattern recognition. At present, most of the research of human behavior recognition is focused on the video data, which is based on the video number. At the same time, due to the complexity of video image data, it is easy to violate personal privacy. With the rapid development of Internet of things technology, it has attracted the attention of a large number of experts and scholars. Researchers have tried to use many machine learning methods, such as random forest, support vector machine and other shallow learning methods, which perform well in the laboratory environment, but there is still a long way to go from practical application. In this paper, a recursive neural network algorithm based on long and short term memory (LSTM) is proposed to realize the recognition of behavior patterns, so as to improve the accuracy of human activity behavior recognition.

Sign in / Sign up

Export Citation Format

Share Document