【摘要】：為解決傳統(tǒng)的水溫小樣本非實(shí)時(shí)預(yù)測(cè)方法預(yù)測(cè)精度低、魯棒性差等問題,基于物聯(lián)網(wǎng)實(shí)時(shí)數(shù)據(jù),提出了遺傳算法(GA)優(yōu)化BP神經(jīng)網(wǎng)絡(luò)的池塘養(yǎng)殖水溫短期預(yù)測(cè)方法,并在此基礎(chǔ)上設(shè)計(jì)開發(fā)了池塘養(yǎng)殖水溫預(yù)測(cè)系統(tǒng),首先采用主成分分析法篩選出影響池塘水溫的關(guān)鍵影響因子,減少輸入元素;然后使用遺傳算法對(duì)初始權(quán)重和閾值進(jìn)行優(yōu)化,獲取最優(yōu)參數(shù)并構(gòu)建了基于BP神經(jīng)網(wǎng)絡(luò)的水溫預(yù)測(cè)模型;最后采用Java語言開發(fā)了基于B/S體系結(jié)構(gòu)的預(yù)測(cè)系統(tǒng)。該系統(tǒng)在江蘇省宜興市河蟹養(yǎng)殖池塘進(jìn)行了預(yù)測(cè)驗(yàn)證。結(jié)果表明:該系統(tǒng)在短期的水溫預(yù)測(cè)中具有準(zhǔn)確的預(yù)測(cè)效果,與傳統(tǒng)的BP神經(jīng)網(wǎng)絡(luò)算法相比,研究內(nèi)容評(píng)價(jià)指標(biāo)平均絕對(duì)誤差(MAE)、平均絕對(duì)百分誤差(MAPE)和誤差均方根(MSE)分別為0.196 8、0.007 9和0.059 2,均優(yōu)于單一BP神經(jīng)網(wǎng)絡(luò)預(yù)測(cè),可滿足實(shí)際的養(yǎng)殖池塘水溫管理需要。
[Abstract]:In order to solve the problems of low prediction accuracy and poor robustness of traditional non-real-time prediction method for small sample water temperature, a short-term prediction method of pond culture water temperature based on real-time data of Internet of things (IoT) was proposed based on genetic algorithm (GA) optimized BP neural network. On this basis, the prediction system of pond culture water temperature is designed and developed. Firstly, the key factors affecting pond water temperature are screened by principal component analysis, and the input elements are reduced, and then the initial weight and threshold are optimized by genetic algorithm. The optimal parameters are obtained and the water temperature prediction model based on BP neural network is constructed. Finally, a prediction system based on B / S architecture is developed by using Java language. The system was predicted and verified in river crab culture pond of Yixing City, Jiangsu Province. The results show that the system has accurate prediction effect in short-term water temperature prediction, compared with the traditional BP neural network algorithm. The average absolute error (MAE),) and mean absolute error (MAPE) and root mean square (RMS) (MSE) of the evaluation index were 0.196 and 0.059 2, respectively, which were superior to the prediction of single BP neural network, and could meet the requirement of water temperature management in culture pond.
【作者單位】： 中國農(nóng)業(yè)大學(xué)信息與電氣工程學(xué)院;農(nóng)業(yè)部農(nóng)業(yè)信息獲取技術(shù)重點(diǎn)實(shí)驗(yàn)室;北京農(nóng)業(yè)物聯(lián)網(wǎng)工程技術(shù)研究中心;
【基金】：山東省重點(diǎn)研發(fā)計(jì)劃項(xiàng)目(2015GGX101041) 上海市科技興農(nóng)重點(diǎn)攻關(guān)項(xiàng)目(滬農(nóng)科攻字(2014)第4-6-2號(hào)) 廣東省海大集團(tuán)基于物聯(lián)網(wǎng)技術(shù)的智慧水產(chǎn)養(yǎng)殖系統(tǒng)院士工作站(2012B090500008)
【分類號(hào)】：S964.3;TP183

【相似文獻(xiàn)】

相關(guān)碩士學(xué)位論文 前2條

1 楊爭光;養(yǎng)殖水質(zhì)數(shù)據(jù)處理與預(yù)測(cè)技術(shù)研究[D];太原科技大學(xué);2015年

2 潘金晶;基于RBF神經(jīng)網(wǎng)絡(luò)的溶解氧預(yù)測(cè)模型研究[D];上海海洋大學(xué);2016年

，

本文編號(hào)：2250336