Short-term Load Current Prediction Of Transmission Line Based On Deep Learning

Through the rational utilization of our country’s abundant new energy resources,a huge amount of electricity can be generated.However,in order to fully consume the electricity generated by the utilization of new energy,a considerable scale of transmission lines are required to participate in the transmission work.Although the laying of new transmission lines can relieve the power transmission pressure of the entire grid line,the construction of new lines is not only costly,but also the construction speed is far less than the current growing social demand for electricity.Therefore,reasonably enhancing the current-carrying capacity of existing transmission lines has become the most cost-effective solution.According to the relationship between the line maximum ampacity and the actual load current of the line,the power grid staff can calculate the remaining safe capacity of the transmission line and adjust the generation plan accordingly.Therefore,how to safely determine the residual safety capacity of the line has an important practical significance.In order to determine the remaining safe capacity of transmission lines,firstly,a data driven short-term load current prediction model for transmission lines is proposed in this thesis based on the actual current load and historical meteorological data.In order to reduce the overall prediction difficulty of line load current,the data decomposition method is used to decompose the current data.Secondly,the deep learning model with stronger data information mining ability than machine learning model is selected as the prediction model,and a new model activation function is introduced.At the same time,a dynamic feature weighting method that can participate in model learning synchronously is designed,and it is used to evaluate the importance of diverse input features to the current carrying capacity output.Finally,two swarm intelligence optimization algorithms are improved,and under the guidance of game theory strategy,the four swarm intelligence optimization algorithms before and after improvements are integrated to obtain a brand-new optimization algorithm.It is then applied to the work of optimizing the initial weight and threshold parameters of the model.Experiments are carried out with with the data of an actual transmission line in a certain area in southern our country.The effectiveness of dynamic feature weighting,data decomposition and model parameter optimization is verified in a progressive way.And it is confirmed that compared with other comparative models,the model proposed in this thesis has better forecasting effect,and still has good forecasting performance when dealing with the line operating load current with seasonal fluctuation.According to the characteristics of the heat balance equation,the conductor temperature can be calculated when the corresponding meteorological parameters and line operating current are determined.Therefore,based on the proposed short-term load current prediction model of transmission line,this thesis uses its current prediction results combined with the heat balance equation to predict the line conductor temperature in operation.And the maximum residual safety load allowance of transmission line is calculated considering the risk of line operation.Experiments show that the calculated maximum residual safety load allowance is safe and reliable.