Research On Power Load Forecasting Considering Distributed Photovoltaic Power Generation

Accurate power load forecasting can not only play an important role in the economic operation,security and stability of the power grid,but also play a guiding role in the planning and construction of the power grid.At the same time,with the continuous advancement of power market reform,power load forecasting can enable users to better participate in the wave of market-oriented reforms.With the popularization and application of distributed photovoltaic power generation,more and more photovoltaic power sources are connected to the power grid for operation.Due to its intermittent and random characteristics,node shortterm load forecasting of the power grid becomes a need when large-scale photovoltaic power generation is connected for operational and economical consideration.With a large number of distributed photovoltaic power sources connected to the power system,the power consumption pattern of power loads has changed to a certain extent.At the same time,with the advent of the information age,a large number of load data and meteorological data stored by power grid enterprises have been analyzed.Taking into account the changes of the node load pattern of distributed photovoltaic power generation,it is essential to improve the accuracy of node short-term load forecasting under distributed photovoltaic power generation.In this thesis,through the in-depth analysis and summary of the existing traditional power load forecasting methods and distributed photovoltaic power generation load forecasting methods,a CNN-LSTM node short-term load forecasting model based on mathematical morphological decomposition is proposed.The proposed model is validated by simulation.The main research contents of this thesis are as follows:Firstly,starting from the trend and characteristics of the node load curve considering distributed photovoltaic power generation,the characteristics of the load curve of traditional load feeder,distributed photovoltaic power generation load and considering distributed photovoltaic power generation are analyzed in detail respectively,and the distribution of the node load curve considering distributed photovoltaic power generation is obtained.The node load considering photovoltaic power generation is regular and periodic,but also contains certain randomness and fluctuation.At the same time,the Pearson correlation coefficient is used to analyze the node load variation degree with different influencing factors under the consideration of distributed photovoltaic power generation,which identifies the factors that have an effect on the node load.Secondly,in view of the abnormal data problem in the node load data,according to the different characteristics of abnormal data,it is divided into two categories: distorted data and missing data.The distorted data is eliminated by vertical and horizontal processing,and a random-based method is proposed.The missing value completion method of the random forest algorithm performs interpolation and completion on the missing values in the node load data,which enriches the information in the node load data set and reduces the error caused by abnormal data on the prediction accuracy.Then,considering the characteristics of node load under distributed photovoltaic power generation,a CNN-LSTM node load prediction model based on mathematical morphology decomposition is proposed,and the node load data is decomposed into node load trend items through the load decomposition method based on mathematical morphology.The subsequence and the node load random item sub-sequence,and then through the CNN-LSTM network’s load implicit feature mining ability and time feature extraction ability,the two subsequences after decomposition are modeled and predicted,and the prediction results of the two sub-sequences are obtained.The vectors are superimposed to obtain the final node shortterm load forecasting results taking into account distributed photovoltaic power generation.Finally,a simulation experiment is carried out with the load data of a feeder containing distributed photovoltaic power generation in a certain area of Guangdong Province and the meteorological data of the region.The experimental results show that the node load decomposition method based on mathematical morphology proposed in this thesis can effectively improve the consideration and the prediction accuracy of power load under distributed photovoltaic power generation.In the case of using this decomposition method,the prediction error of the CNN-LSTM model used in this thesis is the smallest,which proves the validity and accuracy in node short-term load forecasting of the model proposed in this thesis of considering distributed photovoltaic power generation.