A prediction model for piggery ammonia concentration based on least squares support vector regression using fruit fly optimisation algorithm

Abstract

In order to predict the variation trend of ammonia (NH3) concentration accurately in piggery and reduce the risk of livestock breeding, a prediction model is established. Because NH3 has a great influence on the health of pigs, a prediction model can provide an effective way for pig industries to determine the environmental control strategy and take effective measures to evaluate the air quality of piggery. When predicted value of NH3 concentration is above the warning value, farmers can start fans in advance to maintain the health of pigs. The proposed NH3 concentration prediction model is based on Least Squares Support Vector Regression (LSSVR) model with Fruit Fly Optimisation Algorithm (FOA) to search the optimal parameters γ and σ of LSSVR. As the performances of LSSVR are greatly affected by the two parameters, three optimisation algorithms, Particle Swarm Optimisation (PSO) algorithm, Genetic Algorithm (GA) and traditional LSSVR, are used to compare with FOA. The calculated mean absolute percentage errors of the four prediction models are 0.81%, 2.95%, 4.04% and 5.92%, respectively. The prediction model is used in livestock breeding base, Zhenjiang City, China, and it performs well. The FOA-LSSVR prediction model can serve as an effective strategy applied in multivariable and non-linear piggery environmental control system.