Solar Radiation Prediction Based On Improved Cuckoo Hybrid Integration Model

With the continuous consumption of fossil fuels,energy and environmental problems have become more and more serious,and there is an urgent need to find solutions to solve energy and environmental problems and achieve sustainable development.Therefore,renewable energy has attracted great attention all over the world and has developed rapidly in recent years.Solar energy is considered to be one of the cleanest and most promising renewable energy sources,which makes solar energy an important direction for exploring renewable energy sources.At present,solar power generation has been widely used in developed countries and quite developed countries,and has partially replaced traditional power generation.In recent years,China has been growing at an annual rate of more than 25% in the development and utilization of renewable energy sources.The prediction of solar radiation intensity plays an important role in solar photovoltaic power generation.With the development of solar technology,the demand for high-precision solar radiation intensity data is increasing.Therefore,the prediction of solar radiation intensity has become one of the core contents of solar photovoltaic power generation.Taking Wuhan City as an example,this thesis selects the solar radiation data of Wuhan City from July 6,1983 to September 21,2019,and performs pre-processing such as anomaly detection on the data.In order to build a prediction model with higher accuracy and confidence,this thesis proposes a hybrid integrated model based on an improved cuckoo search algorithm to predict solar radiation intensity.In the proposed hybrid integrated model method,the ensemble empirical mode decomposition(CEEMDAN)with adaptive white noise is first used to decompose the original solar radiation intensity data into several intrinsic mode functions(IMF)and residual components;thereafter The Lempel-Ziv method divides the prediction results of all components,and further divides the IMF into high-frequency sequences,low-frequency sequences,and trend term sequences of Wuhan's solar radiation.Finally,it uses the improved cuckoo search algorithm to collect the frequency-divided sequences,and the final prediction result is obtained through the corresponding set weights.In order to verify the performance of the proposed hybrid ensemble learning method,solar radiation data from Wuhan were introduced for empirical analysis.At the same time,six other benchmark models were added as comparison models,and the validity of the model was illustrated through two dimensions of error analysis and model testing.The results show that the CEEMDAN-LZC-CS-LSSVR model proposed in this thesis is superior to other benchmark models in both level and level accuracy,and performs best in both DM and robustness tests.The chaos simplex improved cuckoo search algorithm is superior to the basic cuckoo search algorithm in all dimensions.This thesis proposes that the model has improved horizontal accuracy over the CLSCSLSSVR model without decomposition and the CEEMDAN-CLSCS-LSSVR with frequency division without LempZiv complexity algorithm.CEEMDAN decomposition method is significantly superior to the other two decomposition methods.The algorithm after frequency division of Lempel-Ziv complexity algorithm can significantly reduce the model implementation complexity,and the accuracy is slightly higher than that of the hybrid algorithm without frequency division.