دانلود کتاب Wind Forecasting in Railway Engineering

فهرست مطالب :

Front Cover
WIND FORECASTING IN RAILWAY ENGINEERING
WIND FORECASTING IN RAILWAY ENGINEERING
Copyright
Contents
List of figures
List of tables
Preface
Acknowledgments
Nomenclature list
A
B
C
D
E
F
G
H
I
K
L
M
N
O
P
R
S
T
V
W
1 - Introduction
1.1 Overview of wind forecasting in train wind engineering
1.2 Typical scenarios of railway wind engineering
1.2.1 Train overturning caused by wind
1.2.2 Pantograph-catenary vibration caused by wind
1.2.3 Bridge vibration caused by wind
1.2.4 Wind-resistant railway yard design
1.2.5 Wind-break wall design
1.2.6 Other scenarios
1.3 Key technical problems in wind signal processing
1.3.1 Wind measurement technology
1.3.1.1 Anemometers selection
1.3.1.2 Data preprocessing
1.3.2 Wind identification technology
1.3.2.1 Feature recognition
1.3.2.2 Descriptive model construction
1.3.3 Wind forecasting technology
1.3.4 Wind control technology
1.4 Wind forecasting technologies in railway wind engineering
1.4.1 Wind anemometer layout along railways
1.4.2 Single-point wind forecasting along railways
1.4.3 Spatial wind forecasting along railways
1.5 Scope of this book
1.5.1 Chapter 1: Introduction
1.5.2 Chapter 2: Analysis of flow field characteristics along railways
1.5.3 Chapter 3: Description of single-point wind time series along railways
1.5.4 Chapter 4: Single-point wind forecasting methods based on deep learning
1.5.5 Chapter 5: Single-point wind forecasting methods based on reinforcement learning
1.5.6 Chapter 6: Single-point wind forecasting methods based on ensemble modeling
1.5.7 Chapter 7: Description methods of spatial wind along railways
1.5.8 Chapter 8: Data-driven spatial wind forecasting methods along railways
References
2 - Analysis of flow field characteristics along railways
2.1 Introduction
2.2 Analysis of spatial characteristics of railway flow field
2.2.1 Spatial statistical analysis
2.2.1.1 Spatial statistics
2.2.1.1.1 Spatial weight matrix
2.2.1.1.2 Global spatial autocorrelation
2.2.1.1.3 Local spatial autocorrelation
2.2.1.2 Spatial statistical analysis of wind field along railways
2.2.2 Key spatial correlation structure analysis
2.2.2.1 Planar Maximally Filtered Graph
2.2.2.2 Key spatial correlation structure analysis of wind field along railways
2.3 Analysis of seasonal characteristics of railway flow field
2.3.1 Frequency analysis
2.3.1.1 Fast Fourier transform
2.3.1.2 Frequency analysis of wind field along railways
2.3.2 Clustering analysis
2.3.2.1 Bayesian Fuzzy Clustering
2.3.2.2 Clustering analysis of wind field along railways
2.4 Summary and outlook
References
3 - Description of single-point wind time series along railways
3.1 Introduction
3.2 Wind anemometer layout optimization methods along railways
3.2.1 Development progress
3.2.2 Numerical simulation methods
3.2.2.1 Hydrodynamic equations
3.2.2.1.1 Continuity equation
3.2.2.1.2 Momentum equation
3.2.2.1.3 Energy equation
3.2.2.2 Numerical methods in CFD
3.2.2.2.1 Finite difference method
3.2.2.2.2 Finite element method
3.2.2.2.3 Finite volume method
3.2.2.2.4 Particle method
3.2.2.2.5 Lattice Boltzmann method
3.2.2.3 Turbulence model
3.2.3 Anemometer layout optimization
3.3 Single-point wind speed-wind direction seasonal analysis
3.3.1 Seasonal analysis
3.3.1.1 Augmented Dickey Fuller test
3.3.1.2 Hurst exponent
3.3.1.3 Autocorrelation and partial autocorrelation functions
3.3.1.4 Bayesian information criterion
3.3.2 Single-point wind speed seasonal analysis
3.3.2.1 Data description
3.3.2.2 Data difference
3.3.2.3 Seasonal analysis
3.3.2.4 ACF and PACF analysis
3.3.3 Single-point wind direction seasonal analysis
3.3.3.1 Data description
3.3.3.2 Data difference
3.3.3.3 Seasonal analysis
3.3.3.4 ACF and PACF analysis
3.4 Single-point wind speed-wind direction heteroscedasticity analysis
3.4.1 Heteroscedasticity analysis
3.4.1.1 Graphical test
3.4.1.2 Hypothesis tests
3.4.1.2.1 Goldfeld-Quandt test
3.4.1.2.2 Breusch-Pagan test
3.4.1.2.3 White test
3.4.1.2.4 Park test
3.4.1.2.5 Glejser test
3.4.2 Single-point wind speed heteroscedasticity analysis
3.4.2.1 Graphical test
3.4.2.2 Hypothesis tests
3.4.3 Single-point wind direction heteroscedasticity analysis
3.4.3.1 Graphical test
3.4.3.2 Hypothesis tests
3.5 Various single-point wind time series description algorithms
3.5.1 Autoregressive Integrated moving average
3.5.1.1 Theoretical basis
3.5.1.1.1 The autoregressive model
3.5.1.1.2 The moving average model
3.5.1.1.3 The autoregressive moving average model
3.5.1.1.4 The autoregressive integrated moving average model
3.5.1.2 Modeling steps
3.5.1.2.1 Wind speed ARIMA description model
3.5.1.2.2 Wind direction ARIMA description model
3.5.1.3 Description results
3.5.1.3.1 Description results of wind speed ARIMA model
3.5.1.3.2 Description results of wind direction ARIMA model
3.5.2 Seasonal autoregressive integrated moving average
3.5.2.1 Theoretical basis
3.5.2.2 Modeling steps
3.5.2.2.1 Wind speed SARIMA description model
3.5.2.2.2 Wind direction SARIMA description model
3.5.2.3 Description results
3.5.2.3.1 Description results of wind speed SARIMA model
3.5.2.3.2 Description results of wind direction SARIMA model
3.5.3 Autoregressive conditional heteroscedasticity model
3.5.3.1 Theoretical basis
3.5.3.2 Modeling steps
3.5.3.3 Description results
3.5.3.3.1 Description results of wind speed ARCH model
3.5.3.3.2 Description results of wind direction ARCH model
3.5.4 Generalized autoregressive conditionally heteroscedastic model
3.5.4.1 Theoretical basis
3.5.4.2 Modeling steps
3.5.4.3 Description results
3.5.4.3.1 Description results of wind speed GARCH model
3.5.4.3.2 Description results of wind direction GARCH model
3.6 Description accuracy evaluation indicators
3.6.1 Deterministic description accuracy evaluation indicators
3.6.1.1 Deterministic wind speed description results analysis
3.6.1.2 Deterministic wind direction description results analysis
3.6.2 Probabilistic description accuracy evaluation indicators
3.6.2.1 Probabilistic wind speed description results analysis
3.6.2.2 Probabilistic wind direction description results analysis
3.7 Summary and outlook
References
4 - Single-point wind forecasting methods based on deep learning
4.1 Introduction
4.2 Wind data description
4.3 Single-point wind speed forecasting algorithm based on LSTM
4.3.1 Single LSTM wind speed forecasting model
4.3.1.1 Theoretical basis
4.3.1.2 Model structure
4.3.1.3 Modeling steps
4.3.1.4 Result analysis
4.3.1.5 Conclusions
4.3.2 Hybrid WPD-LSTM wind speed forecasting model
4.3.2.1 Theoretical basis
4.3.2.2 Model structure
4.3.2.3 Modeling steps
4.3.2.4 Result analysis
4.3.2.5 Conclusions
4.4 Single-point wind speed forecasting algorithm based on GRU
4.4.1 Single GRU wind speed forecasting model
4.4.1.1 Theoretical basis
4.4.1.2 Model structure
4.4.1.3 Modeling steps
4.4.1.4 Result analysis
4.4.1.5 Conclusions
4.4.2 Hybrid EMD-GRU wind speed forecasting model
4.4.2.1 Theoretical basis
4.4.2.2 Model structure
4.4.2.3 Modeling steps
4.4.2.4 Result analysis
4.4.2.5 Conclusions
4.5 Single-point wind speed direction algorithm based on Seriesnet
4.5.1 Single Seriesnet wind direction forecasting model
4.5.1.1 Theoretical basis
4.5.1.2 Model structure
4.5.1.3 Modeling steps
4.5.1.4 Result analysis
4.5.1.5 Conclusions
4.5.2 Hybrid WPD-SN wind direction forecasting model
4.5.2.1 Theoretical basis
4.5.2.2 Model structure
4.5.2.3 Modeling steps
4.5.2.4 Result analysis
4.5.2.5 Conclusions
4.6 Summary and outlook
References
5 - Single-point wind forecasting methods based on reinforcement learning
5.1 Introduction
5.2 Wind data description
5.3 Single-point wind speed forecasting algorithm based on Q-learning
5.3.1 Q-learning algorithm
5.3.2 Single-point wind speed forecasting algorithm with ensemble weight coefficients optimized by Q-learning
5.3.2.1 Base forecasting models
5.3.2.1.1 Deep belief network
5.3.2.1.2 Long short-term memory
5.3.2.1.3 Gated recurrent units
5.3.2.2 Model abstraction
5.3.2.2.1 State s
5.3.2.2.2 Action a
5.3.2.2.3 Reward r
5.3.2.2.4 Agent
5.3.2.3 Experimental steps
5.3.2.3.1 Training of base forecasting models
5.3.2.3.2 Training of agent
5.3.2.3.3 Testing of model performance
5.3.2.4 Result analysis
5.3.3 Single-point wind speed forecasting algorithm with feature selection based on Q-learning algorithm
5.3.3.1 Forecasting model
5.3.3.2 Model abstraction
5.3.3.2.1 State s
5.3.3.2.2 Action a
5.3.3.2.3 Reward r
5.3.3.3 Experimental steps
5.3.3.3.1 Initialization of candidate feature set
5.3.3.3.2 Training of agent
5.3.3.3.3 Testing of model performance
5.3.3.4 Result analysis
5.4 Single-point wind speed forecasting algorithm based on deep reinforcement learning
5.4.1 Deep Reinforcement Learning algorithm
5.4.2 Single-point wind speed forecasting algorithm based on DQN
5.4.2.1 Multiobjective optimization algorithm
5.4.2.2 Model abstraction
5.4.2.2.1 State s
5.4.2.2.2 Action a
5.4.2.2.3 Reward r
5.4.2.2.4 Agent
5.4.2.3 Experimental steps
5.4.2.3.1 Training of base forecasting models
5.4.2.3.2 Multiobjective optimization of ensemble weight coefficients
5.4.2.3.3 Training of agent
5.4.2.3.4 Testing of model performance
5.4.2.4 Result analysis
5.4.2.4.1 Training and deployment of the DQN agent
5.4.2.4.2 Iteration conditions and optimization results of the NSGA-II algorithm
5.4.2.4.3 Forecasting results and errors of the dynamic ensemble model
5.4.3 Single-point wind speed forecasting algorithm based on DDPG
5.4.3.1 Model abstraction
5.4.3.1.1 State s
5.4.3.1.2 Action a
5.4.3.1.3 Reward r
5.4.3.1.4 Agent
5.4.3.2 Experimental steps
5.4.3.2.1 The training process of the DDPG agent
5.4.3.2.2 Model performance verification
5.4.3.3 Result analysis
5.4.3.3.1 Convergence and reward of the DDPG algorithm
5.4.3.3.2 Forecasting results and errors of the DDPG-based model
5.5 Summary and outlook
References
6 - Single-point wind forecasting methods based on ensemble modeling
6.1 Introduction
6.2 Wind data description
6.3 Single-point wind speed forecasting algorithm based on multi-objective ensemble
6.3.1 Model framework
6.3.2 Theoretical basis
6.3.2.1 Wavelet decomposition
6.3.2.2 Multi-layer perceptron
6.3.2.3 Single-objective optimization algorithm
6.3.2.3.1 Grey wolf optimization algorithm
6.3.2.3.2 Particle swarm optimization algorithm
6.3.2.3.3 Bat algorithm
6.3.2.4 Multi-objective optimization algorithm
6.3.2.4.1 Multi-objective grey wolf optimization algorithm
6.3.2.4.2 Multi-objective particle swarm optimization algorithm
6.3.2.4.3 Multi-objective grasshopper optimization algorithm
6.3.3 Result analysis
6.3.4 Conclusions
6.4 Single-point wind speed forecasting algorithm based on stacking
6.4.1 Model framework
6.4.2 Theoretical basis
6.4.3 Result analysis
6.4.4 Conclusions
6.5 Single-point wind direction forecasting algorithm based on boosting
6.5.1 Model framework
6.5.2 Theoretical basis
6.5.2.1 AdaBoost.RT
6.5.2.2 AdaBoost.MRT
6.5.2.3 Modified AdaBoost.RT
6.5.2.4 Gradient Boosting
6.5.3 Result analysis
6.5.4 Conclusions
6.6 Summary and outlook
References
7 - Description methods of spatial wind along railways
7.1 Introduction
7.2 Spatial wind correlation analysis
7.2.1 Wind analysis methods and data collection
7.2.2 Cross-correlation analysis by MI
7.2.2.1 Theory basis
7.2.2.2 Cross-correlation of the wind locations
7.2.3 Cross-correlation analysis by Pearson coefficient
7.2.3.1 Theory basis
7.2.3.2 Cross-correlation of wind locations
7.2.4 Cross-correlation analysis by Kendall coefficient
7.2.4.1 Theory basis
7.2.4.2 Cross-correlation of wind locations
7.2.5 Cross-correlation analysis by Spearman coefficient
7.2.5.1 Theory basis
7.2.5.2 Cross-correlation of wind locations
7.2.6 Analysis of correlation results
7.3 Spatial wind description based on WRF
7.3.1 Main structures
7.3.2 WRF modeling along the railway
7.3.3 WRF future development trends
7.4 Description accuracy evaluation indicators
7.5 Summary and outlook
References
8 - Data-driven spatial wind forecasting methods along railways
8.1 Introduction
8.2 Wind data description
8.3 Spatial wind forecasting algorithm based on statistical model
8.3.1 Theoretical basis
8.3.1.1 Spatial feature selection based on mutual information
8.3.1.2 Generalized linear regression
8.3.2 Model framework
8.3.3 Analysis of statistical spatial forecasting models
8.3.3.1 Spatial analysis of monitoring sites
8.3.3.2 Results of statistical spatial forecasting models
8.4 Spatial wind forecasting algorithm based on intelligent model
8.4.1 Theoretical basis
8.4.1.1 Spatial feature selection based on binary optimization algorithms
8.4.1.2 Outlier robust extreme learning machine
8.4.2 Model framework
8.4.3 Analysis of intelligent spatial forecasting models
8.4.3.1 Spatial feature selection results
8.4.3.2 Results of intelligent spatial forecasting models
8.5 Spatial wind forecasting algorithm based on deep learning model
8.5.1 The theoretical basis of deep learning spatial forecasting models
8.5.1.1 Spatial feature selection based on sparse autoencoder
8.5.1.2 Deep Echo State Network (DeepESN)
8.5.2 Model framework
8.5.3 Analysis of deep learning spatial forecasting models
8.5.3.1 The convergence of deep learning models
8.5.3.2 Results of deep learning spatial forecasting models
8.6 Summary and outlook
References
Index
A
B
C
D
E
F
G
H
I
K
L
M
N
O
P
Q
R
S
T
U
V
W
Z
Back Cover