A Particle Swarm Optimization Based Verhulst Model for Prognostics

In data-driven prognostic methods, autoregressive moving average(ARMA) model requires stationary time series, and grey model(GM) can achieve high prediction accuracy only for exponential increasingly data sequence. To compensate these shortcomings, a novel prognostic method based on the improved Verhulst model optimized by particle swarm optimization (PSO) is proposed. Firstly, the Verhulst model based on the process that has the saturated condition is presented. Secondly, the traditional Verhulst model's inaccuracy cause is analyzed. Thirdly, based on the inaccuracy cause, the background value of the Verhulst model is reconstructed by the PSO algorithm. Moreover, to improve prediction accuracy, at the prediction stage, the information contained in the data is updated through metabolism. Finally, to validate our proposed approach, two experiments are conducted to compare the improved Verhulst model with ARMA model, GM (1, 1) model and grey relevance vector machine (GRVM), respectively. The experiments results show that: (1) the improved Verhulst model is more suitable to describe the process that has the saturated condition than ARMA model, GM (1, 1) model and GRVM model; (2) the improved Verhulst model has smaller prediction errors evaluated by mean absolute error (MAE), mean square error (MSE), mean absolute percentage error (MAPE) and root mean square error (RMSE); (3) the improved Verhulst model has a high efficiency.