Model Construction And Research Of Wind Energy Resource Assessment And Wind Speed Prediction Based On Machine Learning

With the continuous consumption of fossil energy such as coal and oil,the problems of energy shortage,climate warming and ecological environment deterioration are becoming more and more serious.Looking for a clean and renewable energy with development potential has become an urgent problem to be solved by the international community.With the continuous exploration and research,wind power resources have become the most potential reproducible energy with its strengths of large reserves,wide distribution,cleanliness,and pollution-free,low utilization cost and so on.Wind speed is the most representative and important feature of wind energy resources.However,due to the strong intermittency,randomness and uncontrollability of wind speed,the development cost of wind power is high,which is not conducive to the stable operation of wind power grid connected system.Therefore,efficient and accurate wind energy resource assessment and wind speed prediction are the prerequisites to ensure the safety of wind power equipment,maintain the integration of wind power and the safe and stable operation of power system,especially to reduce the cost of wind power development.For wind energy resource assessment and wind speed prediction,the main research work of this paper is the determination of statistical distribution parameters in wind energy resources assessment and the innovation of wind speed prediction technology.It is expected to establish an effective determination method of statistical distribution parameters and a widely used prediction system with good prediction performance.For wind energy resource assessment,there is still a lack of specific methods to determine the most stable statistical distribution in the existing research.In this paper,four statistical distributions commonly used for assessment are introduced,the parameters of the optimal statistical distribution are optimized,and an improved statistical distribution model is established to further improve the accuracy of wind energy resource assessment.For the prediction of wind energy resources,although the existing hybrid prediction methods can generally achieve good prediction results,the hybrid prediction model usually only uses the nonlinear model to analyze the nonlinear information of the data,and ignoring the linear information contained in the data.In this paper,a deterministic and non-deterministic hybrid prediction system based on linear model and nonlinear model is proposed,which makes up the shortcomings of existing research and further improves the prediction accuracy.The established model is applied to the research of different wind farms for empirical research and comparative analysis.Finally,it is verified that the established model can effectively assess wind energy resources and predict wind speed in practical application,which has broad application prospects.This essay is subdivided into six chapters.The primacy chapter presents the research context and topic choosing secundum,theoretical and practical meanings,research route and main research contents,as well as the main innovations and research limitations.The second chapter combs and summarizes the research status of wind energy resource assessment and wind speed prediction,deterministic prediction,and non-deterministic prediction.In addition,the performance evaluation index of the model and hypothesis test method are introduced to verify the performance of the model.In Chapter 3,four statistical distribution models usually used in wind energy assessment and several estimation techniques of parameters of statistic distribution are introduced.Among them,the grey wolf optimizer algorithm is used to optimize the statistical distribution parameters for the first time.In addition,the relevant wind energy indicators are calculated to determine the wind energy resource potential of the study area.In Chapter 4 and Chapter 5,the deterministic and non-deterministic hybrid prediction systems based on linear and nonlinear models are established respectively and applied to the prediction of wind speed in different regions.The demonstration results indicate that the proposed prediction system in this paper can achieve higher prediction precision.The sixth chapter summarizes the research work of the full text and looks forward to the future research direction.The main study working contents of this essay can be summarized into the next three respects:（1）In view of the complex calculation process of statistical distribution parameters in wind resource assessment,the grey wolf optimizer algorithm is applied to optimize the parameters of the selected optimal statistical distribution for the first time,and an improved statistical distribution model is established to assess wind energy resources.Among them,the best statistical distribution is determined by the minimum root mean square error selection strategy.Using the statistical distribution model established in this paper,the wind energy indexes of three observation points in Bohai Bay are calculated,to determine the potential of wind energy resources.The empirical results display that the improved Weibull distribution model on the basis of grey wolf optimizer algorithm has wonderful fitting capacity for wind speed data of three observation points in Bohai Bay,and can accurately and effectively evaluate wind energy resources.（2）In view of the fact that the linear model cannot deal with the nonlinear relationship,and the single nonlinear model cannot deal with the data with linear and nonlinear characteristics at the same time,two deterministic hybrid prediction systems based on linear and nonlinear models are established in this paper.Firstly,the complementary ensemble empirical mode decomposition is introduced to preprocess the data,so as to eliminate the noise and improve system prediction performance;Secondly,the linear model is introduced to analyze the linear characteristics in the data;Then,the improved back-propagation neural network model based on cuckoo search optimization algorithm is used as a nonlinear prediction module to analyze the nonlinear characteristics in the data;Finally,nine wind speed data sets of Penglai wind farm in Shandong Province are used for empirical research.The results show that the hybrid prediction system has significant advantages compared to other models,and can effectively predict the wind speed time series.In addition,the significance of the prediction system is deeply discussed,which further proves the effectiveness and superiority of the established hybrid prediction system in the deterministic prediction of wind speed.（3）In view of the fact that the prediction part of the non-deterministic hybrid prediction model usually only uses the nonlinear model,this paper improves and establishes the non-deterministic hybrid prediction system based on linear and nonlinear models.Firstly,the improved complete ensemble empirical mode decomposition with adaptive noise is introduced to preprocess the wind speed data,so as to effectively identify and extract the main characteristics of the data;Secondly,three exponential smoothing models are introduced as the linear prediction module of the prediction system,and the prediction performance of the three methods is compared.The results show that the prediction effect of quadratic exponential smoothing is better;Third,based on the comparative study of three neural network models,an improved extreme learning machine model based on grey wolf optimizer algorithm is proposed as the benclunark nonlinear prediction module;Then,combined with the interval estimation theory of the best statistical distribution,the upper and lower limits of the prediction interval are estimated at the set level of significance α.Finally,the wind speed data sets of wind energy resources in different regions are used for empirical research,which proves that the hybrid prediction system can effectively predict the determinacy and indeterminacy of wind speed data.The main innovations of this paper are as follows:（1）This paper establishes an improved statistical distribution model based on the wind speed data of Bohai Bay for the first time,compares and analyzes four statistical distribution models,and determines the optimal distribution model according to the minimum root mean square error strategy.In the parameter estimation of statistical distribution,the application of four traditional numerical methods and three swarm intelligent optimization algorithm in parameter estimation is studied for the first time.In this paper,an improved Weibull distribution model based on grey wolf optimizer algorithm is proposed for the first time.The empirical results show that the model can effectively fit the wind speed in the study area and accurately assess the wind energy resources in the area.（2）Based on statistical model,artificial neural network in machine learning method,intelligent optimization algorithm and data preprocessing technology,a wind speed prediction system is established.The system fully integrates the advantages of various methods and shows good prediction performance.（3）Two deterministic wind speed hybrid prediction systems based on linear and nonlinear models are proposed.The system integrates the data preprocessing,hybrid prediction and evaluation modules,and uses the complementary ensemble empirical mode decomposition technology to extract the main features from the data.It successfully uses the advantages of different models to effectively extract the linear and nonlinear features of wind speed data,overcomes the shortcomings of low precision and poor stability of traditional methods.（4）The research on wind speed prediction mostly focuses on deterministic prediction and less on non-deterministic prediction,namely interval prediction.However,uncertainty prediction can quantify the range of wind speed changes caused by many factors and provide more non-deterministic reference information for decision makers.In view of this,a wind speed non-deterministic prediction system based on linear and nonlinear models is proposed in this paper,which realizes the non-deterministic prediction of wind speed time series and obtains effective prediction results.（5）Existing studies only focus on the application of nonlinear model in wind energy resource prediction,ignoring the importance of linear model in improving prediction performance.The hybrid prediction system based on linear and nonlinear models proposed in this paper overcomes the defects of traditional nonlinear prediction models,improves the previous hybrid prediction system,and solves the problem of low prediction accuracy of a single model.The shortcomings of this paper are as follows:（1）The Weibull distribution model based on intelligent optimization algorithm proposed in this paper only considers three optimization algorithms and does not consider other optimization algorithms.If other optimization algorithms,especially multi-objective optimization algorithms,are considered,the research on wind energy resource assessment will be more abundant,and it may further improve the fitting accuracy of the distribution model,which is worthy of further research in the future.（2）The prediction system constructed in this paper only adopts exponential smoothing and autoregressive moving average model,does not consider other statistical linear methods.The neural network which uesed in this paper in the machine learning approach is relatively single.If more statistical linear approaches and other machine learning approaches are tried in future study,the research on wind speed prediction will be richer,which may further improve the prediction performance of the model,which is worthy of further research and discussion.