Design of a dynamic simulation program for activated sludge systems

Wastewaster treatment processes have dynamic behavior due to changing conditions of flowrate, pollutant concentration and components. Nowadays, mathematical models and computer simulations are widely used to control and estimate complex relationships caused by dynamic behaviors in wastewater treatment processes. In addition to biological organic carbon removal in wastewater treatment plants, the removal of nutrients such as nitrogen from wastewater is important in order to protect the quality of the receiving water bodies. ASM1 and ASM3 models, developed by International Water Association (IWA) to model biochemical processes mathematically, are widely acccepted in applications of activated sludge systems capable of carbon and nitrogen removal. The application of these models is possible with the help of computer programs. Some dynamic simulation programs developed in order to use in WWTP modelling studies are widely available. These programs are used for plant design, process optimization and training purposes. However, as far as is known there is no computer program developed for the dynamic simulation of activated sludge systems in Turkey. In the scope of this thesis in order to contribute to the further studies in this field in Turkey, it is aimed to design a dynamic activated sludge simulator program, which includes WWTP configurations that are able to remove carbon and nitrogen from the wastewater. At the first stage of the study, suitable WWTP configurations for carbon and nitrogen removal were identified. These configurations are classic activated sludge (called aerobic in this study), Modified Ludzack-Ettinger and 4 stage Bardenpho. While the aim of the classic activated sludge configuration, which contains only aerobic reactor, is the removal of organic carbon, it also allows nitrification in proper sludge age conditions. However, this configuration is not proper for the removal of nitrogen, which is also known as denitrification, from the wastewater. Denitrification is possible in MLE configuration because it contains anoxic reactor. The last configuration of the program, which is 4 stage Bardenpho, in addition to the MLE configuration, contains an additional anoxic and oxic reactor. 4 stage Bardenpho configuration is preferred in treatment plants aiming high level nitrogen removal. After the determination of the activated sludge configurations included in the program, biochemical process model building studies were started. For this purpose, ASM1 and ASM3 models, which contain carbon and nitrogen removal processes, were detailedly investigated. As a result of the process model building studies the ASM1 model, which is frequently used worldwide, was decided to be utilized as the biochemical process model structure of the simulator. In addition to this, the concepts of defining endogenous respiration process and endogenous respiration of anoxic heterotrophs separately from normal heterotrophs are adapted to the main biochemical process model. Finally, instead of the one way hydrolysis approach of the biodegradable organic matter defined in the ASM1 model, the parallel hydrolysis approach is added to the process model. As a result of these studies, the final biochemical process model consisting of 14 components and 10 processes was built. Besides the bioreactors, settling units, which remove particulate matters from wastewater, are part of the activated sludge system. In addition, the control of the microorganism concentration in the system is ensured through the return of the biomass separated in the settling units to the bioreactors. For these reasons, the activated sludge simulator design should also include the settling model or models. Within the scope of this thesis, the biochemical model of the carbon and nitrogen removal systems is especially focused on. Therefore, a simple model approach based on acceptance of precipitation efficiency, was determined as the settling model. The dynamic simulation of the activated sludge models is only possible by solving the mass balance equations written for each component in all reactors or calculation points in the system. The mass balance equations in the activated sludge process are classified as nonlinear differential equations. To solve these equations numerical methods are preferred as a solution method because it is not possible to solve these equations analytically. Euler integration, one of the those numerical methods, was applied to solve the differential equations in this study. In addition, the mass balance equations have been written according to the assumption that the reactors are completely mixed. After the formation of all this theoretical concept in the design of activated sludge simulator, implementation stage of this concept was begun. MATLAB ™ general-purpose software has been used for programming needs during the implementation stage. The programming structure has been designed as four parts. These are the scripts of data reading, processor, chart creation and printing results. Moreover, Excel files have been used as the user interface of the program for the data entry and the display of the results by utilizing the interoperability of MATLAB ™ and Excel. Required data for starting and maintaining the simulation is entered to the program through Excel data file. This file consists of time and design, model coefficients, influent wastewater characterization, COD and nitrogen fractionation, model components initial value and graphs selection Excel sheets. In order to display the simulation results of conventional pollutant parameters (COD, TN, etc.) and model components as daily and hourly time series, the results properly edited in MATLAB™ were printed to Excel data output file. After the completion of the designing stage, the biochemical process model was tested by using full scale WWTP data set. The WWTP is operated with the Johannesburg configuration in real life. However, since the Johannesburg configuration was not included in the designed simulator, anoxic volume in the return line of the WWTP was added to the anoxic reactor volume in the main process line of MLE configuration in order to simplfy the WWTP configuration to the MLE configuration. In addition to this, the anaerobic reactor volume, which is part of the WWTP process configuration for phosphorus removal, was not taken into account in the simulation. Model testing simulations were carried out considering all these assumptions. The 220 days portion of the 365 days data of the WWTP was used for calibration and the remaining part of the data was used in the validation phase. The simulation results obtained during the calibration and validation phases were compared with the WWTP output data by using the root mean square error (RMSE) statistical method. The RMSE value was calculated as 1.28 for ammonia nitrogen and 1.54 for nitrate nitrogen in the calibration phase. On the other hand, RMSE value for ammonia nitrogen and nitrate nitrogen was calculated as 1.50 in the verification phase. When the RMSE values are compared with another WWTP modelling study RMSE values, it was determined that the model in this thesis is calibrated and validated within acceptable limits. In addition, Janus coefficients were calculated to analyze whether the error rates of the model components in the calibration phase and those of the model components in the validation phase are different or not. As a result, it is identified that the Janus coefficients calculated as 1.17 and 0.97 for ammonia and nitrate nitrogen, respectively are within reliable limits considering the Janus coefficients calculated in another WWTP modeling study. After the model testing stage was completed, the process model formed in this study was transfered to the AQUASIM program and was run under the same process conditions in order to test calculation accuracy of the designed simulator within the scope of this thesis. Finally, it was detected that the simulation results, which are obtained by using 10-4 day calculation time step, of the designed activated sludge simulator is consistent with the AQUASIM simulation results. The dynamic activated sludge simulator designed within the scope of this thesis was tested in point of run time. An average computer was used during test runs. The computer features are 222 GB SSD hard disk capacity, normal 2.6 GHz, maximum 3.2 GHz processor speed and 8 GB memory capacity. Running time of the 365 days simulations was measured as 1.6 min. for Aerobic configuration with 28 mass balance equations, 5.6 min. for MLE configuration with 42 mass balance equations and 11.0 min. for Bardenpho configuration with 70 mass balance equations. The 365 days simulations contained all of the program steps, which were data reading, processing, printing results as graphical and tabular. Finally, the period of time covered by each script in the total running time of the simulator was also examined. Finally, it has been identified that the processor and printing results (tabular) scripts have significant shares in the total running time of the simulator.