Dissolved Oxygen Prediction Based on SOA-SVM and SOA-BP Models
To improve the accuracy of dissolved oxygen prediction,this paper researches and proposes a prediction method that combines seagull optimization algorithm (SOA) with support vector machine (SVM) and BP neural network,prepares four prediction schemes based on the monthly dissolved oxygen monitoring d...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Pearl River
2021-01-01
|
| Series: | Renmin Zhujiang |
| Subjects: | |
| Online Access: | http://www.renminzhujiang.cn/thesisDetails#10.3969/j.issn.1001-9235.2021.04.015 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | To improve the accuracy of dissolved oxygen prediction,this paper researches and proposes a prediction method that combines seagull optimization algorithm (SOA) with support vector machine (SVM) and BP neural network,prepares four prediction schemes based on the monthly dissolved oxygen monitoring data of the Jinghong Power Station in Xishuangbanna,a national important water supply source in Yunnan Province,from January 2009 to September 2020,optimizes the key parameters of SVM and weight threshold of BP neural network by SOA to construct SOA-SVM and SOA-BP models,predicts the dissolved oxygen of Jinghong Power Station based on the models,and compares the prediction results with those of SVM and BP models.The results show that:The absolute values of the average relative errors of the SOA-SVM and SOA-BP models for the 4 schemes of dissolved oxygen prediction are between 4.07%~4.98% and 3.85%~4.83%,and that of the average absolute errors are 0.309~0.374 mg/L and 0.294~0.371 mg/L,respectively.With better prediction accuracy than SVM and BP models,they have good prediction accuracy and generalization ability.SOA can effectively optimize the key parameters of SVM and weight threshold of BP neural network.SOA-SVM and SOA-BP models are feasible for dissolved oxygen prediction,which can provide references for related prediction research. |
|---|---|
| ISSN: | 1001-9235 |