Prediction and Early-warning of Water Quality in Aquaculture Based on Computational Intelligence
|School||China Agricultural University|
|Course||Agricultural Information Technology|
|Keywords||non-linear prediction and early-warning computational intelligence water quality inaquaculture least squares support vector regression|
Water quality deterioration is the primary factor that induces the outbreak of aquatic disease and even large quantities of death. The quality of cultured water is affected by many factors, and mechanisms of action between the parameters are complex, all of this lead to the fact that accurate prediction and forecasting and early warning of water quality has been a tough problem that needs to be solved. In this paper, based on computational intelligence, we take the key parameters of crabs in the cultured water, dissolved oxygen and PH to study the aquaculture water quality prediction and warning method by using signal processing technology, swarm intelligence computation and machine learning technology. The detailed explanations are as follows:(1) The analysis of factors and its influence of Aquaculture water quality. In view of problems of too many parameters of aquacultured water quality, complex interaction mechanisms, and also the difficulty in analyzing the relationship between water quality parameters and changing rules of the parameter itself, the water quality parameters interacts upon system dynamics and energy conservation. Moreover, a systematic dynamics model for dissolved oxygen, PH value, temperature and other water quality parameters is also established, the interactive relationship among the key parameters of crabs in cultured water is clarified. The study shows that, this method is the suitable and multi-factored method for determining the nature of the aquatic parameters of aquaculture.(2) The study of water quality data pre-processing method of aquaculture. In view of the data deficiency of the water quality monitored and the performance of method which forecasts and alerts the noise effect, this paper provides a simple and useful mean to repair the cultured water quality data, coming up with the noise reduction method and the feature extraction method. By using linear interpolation and averaging method of horizontal and vertical processing of similar data, which will be repaired. This paper also applies the modified wavelet analysis method to denoise the water quality data and extract the features of water quality data. Compared with other methods under the same condition, the evaluation index SNR of this modified wavelet analysis method has improved18.93%, and BIAS and RMS decreased96.15%and33.76%respectively. The result shows that this method is possibly to meet the requirement of cleansing the data of cultured water quality, and provide a new path for denoise and feature extraction of culture water quality signal.(3) Dissolved oxygen content nonlinear prediction model based on the least squares support vector regression (LSSVR) model with optimal parameters selected by ant colony optimization (ACO) algorithm. In view of the traditional prediction method, which is not suitable for small scale of sampling, high dimension and parameter optimization by artificial subjective factors influence puts forward the prediction model of dissolved oxygen in aquaculture nonlinear basing on ACO-LSSVR algorithm. The method updates the strategy through the local fine "detection" search and dynamic pheromone updating ideas, improved ant colony optimization algorithm (ACO), we construct the nonlinear prediction model of dissolved oxygen based on ACO fusion LSSVR. Compared with BPNN, the RMSE. running time t decreased67.9%and2.3464s respectively. The results show that the model prediction accuracy has been obviously improved, and also has good robustness and generalization ability, which meets the actual needs of the intensive aquaculture water quality management.(4) Dissolved oxygen content nonlinear prediction model based on the least squares support vector regression (LSSVR) model with optimal parameters selected by improved particle swarm optimization (1PSO) algorithm. In view of the traditional prediction method of slow convergence, low accuracy of prediction problems has been put forward the prediction model of dissolved oxygen in aquaculture nonlinear IPSO-LSSVR. The method updates the strategy through the adaptive inertia weight dynamically, improved particle swarm optimization algorithm (IPSO), the fine search LSSVR model parameter optimization process, we construct the prediction model of dissolved oxygen in nonlinear IPSO fusion LSSVR. Compared with the traditional LSSVR, the relative RMSE and MAE are29.36%and67.46%, the results show that, the method has fast convergence speed, good prediction effect, and high precision prediction of dissolved oxygen in aquaculture.(5) Dissolved oxygen content nonlinear forecasting method in aquaculture is based on wavelet analysis (WA) and least squares support vector regression (LSSVR) with an optimal improved Cauchy particle swarm optimization (CPSO) algorithm. In order to solve large noise disturbance, the low prediction accuracy and inefficiency to trap in local extreme of the traditional forecasting methods in water quality, we proposes a hybrid dissolved oxygen content forecasting method based on WA-CPSO-LSSVR algorithm. In the modeling process, the original dissolved oxygen sequences were de-noised and decomposed into several different resolution frequency signal subsets by using the wavelet analysis method. Cauchy mutation and the adaptive weight update operator combination, improved particle swarm optimization algorithm, the combined LSSVR model with adaptive parameter optimal acquisition, construction of the multi-scale analysis of crab aquaculture dissolved oxygen nonlinear combination forecasting model. The experimental results show that, the method can effectively solve problems of the conventional method, multi-scale analysis of the characteristics of water quality, betters the prediction effect, suits for high density aquaculture dissolved oxygen prediction and offers the decision based on quality scientific regulation.(6) Forecasting for pH value of aquaculture water quality based on principal component analysis (PCA) and least squares support vector machine (LSSVR) optimized by modified cultural artificial fish-swarm algorithm (MCAFA). In order to reduce the pH value of the crabs metabolism and physiological function of stress, we puts forward the prediction model of water quality based on PCA-MCAFA-LSSVR, which the hyper-parameters is optimized by MCAFA algorithm. The dimension of aquiculture ecologic environmental data has been reduced by PCA method, a core prediction set based on4factors obtained by using PCA to deduct the redundant and disturbed properties from the initial set based on10factors. Using the double evolutionary mechanism of cultural algorithm for reference, the model takes LSSVR as an artificial fish, using the belief space to guide the shoal evolution step size and global search and the Cauchy mutation to improve the diversity of the artificial fish swarm, which obtains the optimal hyper-parameters nonlinear water quality prediction model automatically. Experimental results show that the PCA-MCAFA-LSSVR prediction model has better prediction effect than the other methods, for example, the absolute error of the93.05%test samples are less than8%. It is obvious that PCA-MCAFA-LSSVR prediction model can eliminate superfluous data, low computational complexity and high forecast accuracy, and open up new approaches for aquaculture pH accurate forecasting.(7) Water quality early-warning model of aquaculture based on support vector machine optimized by rough set algorithm. A new early warning model of water quality, combining rough set (RS) and support vector machine (SVM), is presented to improve the prediction precision affected by mass coupling factors, complex mode and information loss. Firstly, a core warning set based on5factors is obtained by using RS to deduct the redundancy and disturb properties from the initial set based on14factors. Consequently, the early warning model of water quality based on RS-SVM is built up by the core warning set. The experimental results show that our method improves the precision to more than91%in any warning level by using the water quality data. Compared with the other methods, the new model not only has effectiveness of calculation and prediction, but also provides warning results with practicality. This model demonstrates a new thought of early warning on intensive aquaculture water quality.(8) Design and implementation of the prototype system of water quality forecasting warning system for river crab aquaculture. The objective of the prototype system is to test the abovementioned methods. The hardware system of the prototype includes5parts, i.e., water quality sensors, wireless transmitting equipment, the on-site monitoring center and the remote monitoring center. The software system includes5modules, i.e., data acquisition, data preprocessing, water quality forecasting early warning management, data retrieval, Information release, water quality control management and system maintenance. Great deals of experiments indicate that the researched method works effectively and efficiently in aquaculture forecasting early warning based on computational intelligence.