(1. 中南大學(xué) 信息科學(xué)與工程學(xué)院，長沙 410083；
2. 湖南第一師范學(xué)院 信息科學(xué)與工程系，長沙 410205)
摘 要: 氧化鋁晶種分解過程是一個具有多級串聯(lián)結(jié)構(gòu)和強(qiáng)烈不可測干擾的大規(guī)模復(fù)雜過程，分解溫度是其關(guān)鍵工藝參數(shù)。為精確控制分解溫度，根據(jù)該過程的結(jié)構(gòu)特點，將其分成多個子系統(tǒng)，并綜合機(jī)理分析、參數(shù)辨識和時間序列分析方法建立基于不可測擾動預(yù)測的子系統(tǒng)自適應(yīng)預(yù)測控制模型，并將前級子系統(tǒng)的狀態(tài)作為可測擾動引入本級子系統(tǒng)模型，分別求解各子系統(tǒng)的優(yōu)化控制目標(biāo)，獲取優(yōu)化操作變量。基于實際生產(chǎn)過程數(shù)據(jù)的仿真結(jié)果表明，所提出的分散型自適應(yīng)模型預(yù)測控制方法具有較強(qiáng)的抗干擾能力，能準(zhǔn)確跟蹤分解溫度設(shè)定值，滿足晶種分解生產(chǎn)過程中對分解終止溫度、分解始末溫差和降溫速度的控制要求。本方法對于具有串聯(lián)結(jié)構(gòu)和不可測強(qiáng)干擾的非線性大規(guī)模復(fù)雜過程的模型預(yù)測控制具有顯著的實用價值。
關(guān)鍵字: 氧化鋁；晶種分解；自適應(yīng)預(yù)測模型；分散預(yù)測控制
（1. School of Information Science and Engineering, Central South University, Changsha 410083, China;
2. Department of Information Science and Engineering, Hunan First Normal University, Changsha 410205, China）
Abstract:The alumina seed precipitation process is a complicated large-scale process with multi-stage tandem structure and strong unmeasured disturbances, in which the decomposition temperature is a key technological parameter. In order to control the decomposition temperature precisely, the process was divided into several subsystems according to its structural characteristics, and the adaptive predictive model of each subsystem based on unmeasured disturbance prediction was built by mechanism analysis, parameter estimation and time series analysis method. The front-end subsystem state as a measurable disturbance was introduced into the corresponding subsystem model, and the optimal operational variables were obtained by respectively solving the optimization objectives of each subsystem. The simulation results based on actual process data show that the proposed decentralized adaptive model predictive control (MPC) method has a strong capacity of resisting disturbances and a good following of the set point, and can meet the control requirements of terminal decomposition temperature, decomposition temperature range and the cooling rate for the alumina precipitation process. The proposed method can be applied to the nonlinear complicated large-scale process with multi-stage tandem structure and strong unmeasured disturbances, and is of remarkable practical value.
Key words: alumina; seed precipitation; adaptive predictive model; decentralized model predictive control
