Study On Coagulation Pretreatment Of Oilfield Produced Water And BP Neural Network Prediction Model For Coagulant Dosage

With most of the oilfields in China have been coming to the late development period, oilmoisture content was increased, therefore output of the oilfield produced water constantlyenhanced. The oilfield produced water contained high content of oil, sulfate-reducing bacteria,pigment and mechanical impurities, it should go through a series of processes of purification,disinfection, filtration in order to reuse. Due to the complexity of oilfield produced waterquality and high degree of emulsification, the emulsified oil is difficult to dispose by physicalor biochemical methods separately. Therefore coagulation pretreatment process is commonlyused in treating oilfield produced water at present. Flocculants as the key factor in coagulationtreatment of oilfield produced water, its nature directly determine the flocculation effect,thereby affecting the subsequent water treatment processes. Therefore, the key point is todevelop a high-efficiency flocculant to treat oil field produced water, which can improvetreatment efficiency and reduce process cost. At present, coagulation-sedimentation-filtrationof three sections of the process is a common approach to deal with oilfield produced water.Coagulation is an oil field produced water pretreatment process, the treatment effect dependson the coagulant dosage. Coagulant dosage is usually determined by artificial regular jar tests,but jar tests requires a certain flocculation reaction time, the system settings determined bythis method are only suitable for the approximate optimal value within a certain time, it’sunable to reflect the current water quality conditions, which makes it difficult to adjust thedosage timely and accurately. Coagulation process is a complex physical and chemicalprocess and the factors which affect the coagulant dosage are so much, such as raw waterturbidity, temperature, pH value, flow, conductivity, this makes the system setting that theoptimal economic dosage needs to be constantly revised according to changes with waterquality. Thus, establish appropriate mathematical model which reflects the relation betweenfactors affecting the coagulant dosage and coagulant dose is of great significance for thedetermination of the optimal coagulant dosage and timely online control for coagulationdosage.In this paper, poly aluminum ferric chloride （PAFC） and cationic polyacrylamide（CPAM）,poly（dimethyldiallylammonium chloride）（PDMDAAC） and chitosan （CTS） as raw materialsto prepare inorganic-organic composite flocculants. The optimal preparation processparameters of PAFC-CPAM, PAFC-PDMDAAC and PAFC-CTS were determined byorthogonal experiments and composite flocculants were characterized by means of infraredspectroscopy and scanning electron microscope. The flocculation performance of compositeflocculants was studied preliminarily by dosage, pH value, sedimentation time, stirring timeand other factors in treating simulated oilfield produced water, then to select a compositeflocculant of best flocculation performance. Uniform design experiment was employed andjar tests dose the best composite flocculant to treat simulated oilfield produced water whichcollects sample data for the BP neural network. The data were intended for the BP networkcoagulant dosage prediction model for coagulant dosage training, simulation and prediction ofthe prediction model.（1） Poly aluminum ferric chloride （PAFC） and cationic polyacrylamide （CPAM） as rawmaterial, jar tests were used to treat simulated oil field produced water. The oil removal as the main index, the optimal preparation process parameters of PAFC-CPAM by L₉（3⁴） orthogonaldesign was determined as follows: of CPAM/PAFC composite ratio （mass ratio） of0.20, thereaction temperature of60℃, stirring time of1.0h, the pH of3. PAFC-CPAM, PAFC andCPAM is used to treat simulated oil field produced water, and the factors such as dosage, pHvalue, sedimentation time, stirring time on treatment effects were investigated, the resultsshowed as follows: The advantages of PAFC-CPAM lied in its low dosage, low process cost,high efficiency, rapid settling velocity and the wide pH range compared to PAFC and CPAM.（2） Poly aluminum ferric chloride （PAFC） and poly（dimethyldiallylammonium chloride）（PDMDAAC） as raw material, jar tests were used to treat simulated oil field produced water.The oil removal as the main index, the optimal preparation process parameters ofPAFC-PDMDAAC by L₉（3⁴） orthogonal design was determined as follows: composite ofPAFC/PDMDAAC composite ratio （mass ratio） of20:1, the reaction temperature of50℃, pHof4, the stirring time of45min. PAFC-PDMDAAC, PAFC and PDMDAAC is used to treatsimulated oil field produced water, and the factors such as dosage, pH value, sedimentationtime, stirring time on treatment effects were investigated, the results showed as follows: Theadvantages of PAFC-PDMDAAC lied in its low dosage, low process cost, high efficiency,rapid settling velocity and the wide pH range compared to PAFC and PAMDAAC.（3） Poly aluminum ferric chloride （PAFC） and chitosan （CTS） as raw material, jar testswere used to treat simulated oil field produced water. The oil removal as the main index, theoptimal preparation process parameters of PAFC-CTS by L₉（3⁴） orthogonal design wasdetermined as follows: PAFC/CTS composite ratio （mass ratio） of10:1, the reactiontemperature of50℃, reaction pH of4, the stirring time of1.0h. PAFC-CTS, PAFC and CTSis used to treat simulated oil field produced water, and the factors such as dosage, pH value,sedimentation time, stirring time on treatment effects were investigated, the results showed asfollows: The advantages of PAFC-CTS lied in its low dosage, high efficiency, rapid settlingvelocity and the wide pH range compared to PAFC and CTS.（4） PAFC-CPAM, PAFC-PDMDAAC and PAFC-CTS were respectively employed totreat simulated oil field produced water and the effects of dosage and pH value on thecoagulation effects were investigated. With the dosage of2.0mg/L, the oil removal byPAFC-CPAM, PAFC-PDMDAAC and PAFC-CTS were96.48%,96.23%and94.75%, theturbidity removal were98.08%,97.21%and96.06%, the residual iron content in simulatedoilfield produced water is0.0583mg/L,0.2112mg/L and0.2448mg/L. With the dosage of2.0mg/L, pH value of410, compared with PDMDAAC and PAFC-CTS, the flocculationperformance of PAFC-CPAM with changes in pH value is relatively tiny. The results showedthat the PAFC-CPAM was with a better decontamination ability and a wide pH range （410）.（5） Uniform design experiment was employed and jar tests dose the best compositeflocculant to treat simulated oilfield produced water which collects sample data for the BPneural network. Temperature, pH, influent turbidity, influent oil content, effluent waterturbidity and effluent oil content as the input variables and the coagulant dosage as the outputvariables, BP neural network model is used for coagulant dosage prediction. The BP neuralnetwork with6-15-1type is utilized for predicting coagulant dosage. The results show asfollows: For the training sample, the mean relative error of the predictive value of trainingsamples was2.607%and the sample data that the relative error of the predictive value of≤ 5%accounted for87.69%in the total sample data. Meanwhile, the linear regression analysisof the predictive value of the training samples and the true value revealed the correlationcoefficient R²=0.9987, MSE=0.00373mg/L. The results show that the BP neural networkprediction model has a good predictive performance. Sample data points in the test samplesoutside the training sample is used to test the predictive ability of BP network predictionmodel, the results are as follows: the average relative error of the predictive value of testsamples was2.627%, the sample data that relative error of the predictive value of≤5%accounted for93.33%in the total sample data. Meanwhile, the linear regression analysis ofthe predictive value of testing samples and the true value revealed the correlation coefficientR²=0.9968, MSE=0.00786mg/L. The results show that the BP neural network predictionmodel has a certain degree of generalization ability and has high prediction accuracy.In this article, simulated oil field produced water is disposed by three kinds of inorganic-organic composite flocculants whose treatment efficiency were higher than90%. BP networkprediction model for coagulant dosage has high prediction accuracy of the coagulant dosage;it can achieve the optimum coagulant dosage and can achieve good economic benefits.