The simulation of forest fire spread is a key problem for the management of fire, and Cellular Automata (CA) has been used to simulate the complex mechanism of the fire spread for a long time. The simulation of CA is driven by the rate of fire spread (ROS), which is hard to estimate, because some input parameters of the current ROS model cannot be provided with a high precision, so the CA approach has not been well applied yet in the forest fire management system to date. The forest fire spread simulation model LSTM-CA using CA with LSTM is proposed in this paper. Based on the interaction between wind and fire, S-LSTM is proposed, which takes full advantage of the time dependency of the ROS. The ROS estimated by the S-LSTM is satisfactory, even though the input parameters are not perfect. Fifteen kinds of ROS models with the same structure are trained for different cases of slope direction and wind direction, and the model with the closest case is selected to drive the transmission between the adjacent cells. In order to simulate the actual spread of forest fire, the LSTM-based models are trained based on the data captured, and three correction rules are added to the CA model. Finally, the prediction accuracy of forest fire spread is verified though the KAPPA coefficient, Hausdorff distance, and horizontal comparison experiments based on remote sensing images of wildfires. The LSTM-CA model has good practicality in simulating the spread of forest fires.

Modeling forest fire spread is a very complex problem, and the existing models usually need some input parameters which are hard to get. How to predict the time series of forest fire spread rate based on passed series may be a key problem to break through the current technical bottleneck. In the process of forest fire spreading, spread rate and wind speed would affect each other. In this paper, three kinds of network models based on Long Short-Term Memory (LSTM) are designed to predict fire spread rate, exploring the interaction between fire and wind. In order to train these LSTM-based models and validate their effectiveness of prediction, several outdoor combustion experiments are designed and carried out. Process data sets of forest fire spreading are collected with an infrared camera mounted on a UAV, and wind data sets are recorded using a anemometer simultaneously. According to the close relationship between wind and fire, three progressive LSTM based models are constructed, which are called CSG-LSTM, MDG-LSTM and FNU-LSTM, respectively. A Cross-Entropy Loss equation is employed to measure the model training quality, and then prediction accuracy is computed and analyzed by comparing with the true fire spread rate and wind speed. According to the performance of training and prediction stage, FNU-LSTM is determined as the best model for the general case. The advantage of FNU-LSTM is further demonstrated by doing comparison experiments with the normal LSTM and other LSTM based models which predict both fire spread rate and wind speed separately. The experiment has also demonstrated the ability of the model to the real fire prediction on the basis of two historical wildland fires.

Fire spread prediction is a crucial technology for fighting forest fires. Most existing fire spread models focus on making predictions after a specific time, and their predicted performance decreases rapidly in continuous prediction due to error accumulation when using the recursive method. Given that fire spread is a dynamic spatiotemporal process, this study proposes an expanded neural network of long short-term memory based on self-attention (SA-EX-LSTM) to address this issue. The proposed model predicted the combustion image sequence based on wind characteristics. It had two detailed feature transfer paths, temporal memory flow and spatiotemporal memory flow, which assisted the model in learning complete historical fire features as well as possible. Furthermore, self-attention mechanisms were integrated into the model’s forgetting gates, enabling the model to select the important features associated with the increase in fire spread from massive historical fire features. Datasets for model training and testing were derived from nine experimental ground fires. Compared with the state-of-the-art spatiotemporal prediction models, SA-EX-LSTM consistently exhibited the highest predicted performance and stability throughout the continuous prediction process. The experimental results in this paper have the potential to positively impact the application of spatiotemporal prediction models and UAV-based methods in the field of fire spread prediction.

In fields like healthcare and finance, Multivariate Time Series(MTS) data frequently encounter a growing number and complexity of missing values, hindering the application of advanced analytical methods. Effective imputation of missing sensor data is essential for accurate data analysis. In the MTS, the variable changing with time create the time dependence, and the variables at each moment have a coupling relationship and influenced each other. At the same time, the missing series has the similar representation ability with the complete series. Based on these time series characteristics, this paper establishes a MTS imputation model based on the attention mechanism and contrastive learning. The attention model is used to deeply extract the temporal dependencies as well as the dependencies among the variables, while the contrastive learning is employed to ensure that missing series share similar feature representations with the original series, thereby enhancing the model’s imputation performance for incomplete series. Finally, after conducting extensive experiments on two real-world MTS datasets that contain five missing patterns, and when compared to state-of-the-art (SOTA) methods, the model performed well, with thirteen metrics achieving first place. Our method aims to provide a better database for the tasks such as anomaly detection, classification, and prediction of MTS.