Time Series Kalman Filter Model For Temperature Prediction

Temperature is an important indicator of climate,and changes in temperature are a specific reflection of the climate change.Climate change can lead to a series of problems such as natural disasters and ecological damage,which have a huge negative impact on human society and ecosystems.With the rapid development of industrialization,global temperatures have significantly increased.Once the “temperature threshold” is broken,the frequency and intensity of meteorological disaster events will greatly increase,which will have a negative impact on the natural environment and social environment that humans rely on for survival.Protecting the ecological environment and strengthening ecological governance are the national policies proposed by “The 20 th National Congress of the CPC” to promote sustainable green development.Applying statistical methods to time series data to discover the patterns of data changes is one of the main methods adopted by statisticians.Firstly,this thesis uses the temperature time series data published by the WMO to conduct statistical analysis of the global average temperature situation.Under necessary assumptions,the method of time series analysis is used to preprocess historical data and establish an AR(3)model to study the law of temperature change.Applying some historical data as the test set,the accuracy of the model’s prediction was analyzed,it was found that there was error between the predicted values of the model and the measured values.Secondly,considering that temperature data may be affected by various unstable factors during the measurement process.Kalman proposed filtering method can eliminate interference and noise,and use the observation vector of the system to estimate the state of the system,achieving more accurate state feedback.Considering the existence of white noise in the data series,this thesis transforms the time series model into a discrete state space model.Based on the state space model,on the one hand,the controllability and observability of this model were theoretically studied.And on the other hand,Kalman filtering experiments were designed.By comparing the experimental accuracy of the AR(3)model and the temporal Kalman filtering model,it was found that the latter has significantly improved accuracy and can better predict short-term temperature changes.