Research On Short-term Load Forecasting Of Offshore Oil Field Cluster Power Grid

Short-term(electric)load forecasting is based on historical load data,combined with other factors such as weather,temperature,season and other influencing factors to forecast the power demand for the next day to one week.Short term load forecasting plays an important role in ensuring the safe and efficient operation of the power system.In recent years,with the continuous increase of human exploration and exploitation in various sea areas,the scale of development and utilization of marine resources has continued to expand.Offshore oilfield power grids have been rapidly developed as the basis for the power supply of offshore platforms.The offshore oil field group grid network is gradually developed from the ship’s power system.Its load is mostly high-power asynchronous motors,and its load changes are relatively large,and it is also vulnerable to environmental changes.In addition,because the platform is powered by diesel engines and gas turbine generator sets,the power generation capacity is very limited.Therefore,improving the accuracy of short-term load forecasting is more important for offshore oil field power grids.This paper first introduces the purpose and significance of short-term load forecasting for offshore oilfield power grids,and describes the current status of short-term load forecasting at home and abroad.The classification and basic flow of power load forecasting are introduced,and the advantages and disadvantages of various load forecasting methods are explained in detail.The structure and particularity of offshore oilfield power grids are discussed,and the characteristics of load forecasting for offshore oilfield power grids are summarized.In view of the characteristics of short-term load forecasting of offshore oilfield power grids,the main influencing factors of daily load,temperature,weather type,etc.were analyzed and finally quantified,laying the foundation for subsequent forecasting work.Then,the basic principle of the support vector machine is fully clarified.The support vector machine has the obvious characteristics of nonlinear fitting,strong generalization ability and fast training convergence.This paper elaborates on a new swarm intelligence optimization algorithm named dragonfly algorithm proposed by American scholars in 2015.The algorithm of dragonfly algorithm is simple and concise,and it is relatively simple to realize,with less algorithm parameters,convenient adjustment,low overall calculation and excellent global optimization ability,fast convergence time and relatively high precision.At last,some existing problems in the application of SVM in various scenes,including data preprocessing,kernel function selection and parameter optimization,are analyzed,and the existing solutions are summarized.On the basis of studying the performance of the support vector regression machine which has a great influence on its performance,combined with the characteristics of the load forecasting of the offshore oil field group,this paper proposes a method based on the dragonfly algorithm optimization support vector machine(DA-SVM)to predict the short-term(electric power)load of the offshore oil field group.This method uses the optimized support vector machine penalty factor C and the kernel parameter a combination as dragonfly dragonfly.The prediction accuracy of the support vector machine is used as the current adjustment value of the dragonfly.The best location of dragonfly is the best C and σ parameters of support vector machine.DA-SVM algorithm is applied to short-term load forecasting of an offshore oilfield group power grid in Bohai,China,and compared with PSO-SVM,GA-SVM and BP neural network prediction results.Experimental results show that DA-SVM algorithm is simple and global search ability,and has higher prediction accuracy and better computing speed.