Ocean Climate prediction using Adaptive Optimization Technique

The ocean parameters including Sea Surface Temperature (SST) and Density are indispensable in predicting the climate of a particular region which are useful in forecasting weather and planning maritime operations. Buoys used in capturing real-time data of ocean parameters will generally be in continuous movement due to dynamic effects in the sea and these buoys need constant monitoring, though they are anchored, they tend to break away during adverse situations. One way to determine the related ocean parameters, as well as the position of the buoys shortly, is by using a model which is data-driven, and it also creates a neural network to solve the prediction problem. This neural network is a time series regression method and is a branch of Recurrent Neural Network (RNN) called Long Short-Term Memory (LSTM). The main objective of our paper is to propose the use of the optimization technique Particle swarm Optimization (PSO) with LSTM for improving the accuracy of predicting the sea surface temperature and locate the position of sensor nodes as these neural network approaches converge and often tend to fall into the optimal local solution.