A SIMPLE FUZZY LOGIC MANAGEMENT SUPPORT SYSTEM FOR FARM BIOGAS PLANTS

Transcription

1 Convegno di Medio Termine dell Associazione Italiana di Ingegneria Agraria Belgirate, settembre 2011 memoria n. A SIMPLE FUZZY LOGIC MANAGEMENT SUPPORT SYSTEM FOR FARM BIOGAS PLANTS A. Finzi 1, G. Cocolo 1, F. Perazzolo 1 (1) Dipartimento di Ingegneria Agraria, Università degli Studi di Milano ABSTRACT This paper presents a model based on fuzzy logic that is able to support feeding management of farm biogas plants. The input parameters for the model (the feeding rate, the quantity and quality of the organic matter of the feeding, the production of biogas and the percentage methane in the biogas) have been selected among those that are simple and easily obtained for full-scale plants. The model output represents the feeding adjustment required to maintain or restore a situation of biological stability and productivity. The model can be applied to any plant in terms of the general approach (the input variables, output, and rules) but it requires calibration for the values of the membership functions that are typical of each plant. The model has been successfully applied to a case study (a 330-kW el biogas plant monitored for nine months). The results demonstrate the possibility of using the developed support system to manage farm biogas plants. Keywords: Fuzzy logic, anaerobic digestion, management support system, biogas plant. 1. INTRODUCTION In agriculture, energy recovery from biogas produced during the anaerobic digestion of biomass is an environmental and economic opportunity that should be enhanced through optimal management of the plant. The fundamental objective of any process control strategy is to maintain optimum operating conditions. In the case of anaerobic digestion, this concept becomes particularly significant because the phase that controls the entire process, methanogenesis, is very sensitive to environmental changes Cecchi et al. (2005). The monitoring of anaerobic digestion is a difficult, complex, multi-variate process, with few reliable on-line sensors (i.e., sensors that monitor the process constantly) available for the measurement of important parameters. It should be noted that these parameters must be analyzed comprehensively: the observed variation of a single parameter, unless accompanied by the complete monitoring of all other parameters, is difficult to interpret Cecchi et al. (2005). Thus, the control of a biological process through the detection and interpretation of appropriate parameters provides basic guidelines on how to manage and adapt the feed compared to the stability and productivity of the plant. In fact, feeding is the most important means by which the operator of the plant can control productivity. Agricultural-scale anaerobic plants are

2 A. Finzi, G. Cocolo, F. Perazzolo generally managed on the basis of general directions and periodic analyses. Management support systems based on simple and easily collected information are then able to indicate the feeding strategy to adopt and maintain adequate organic loads and stability, which could improve the performance of the energy production plants. One possible approach to developing such a tool is the use of fuzzy logic, which has already been widely used and validated as a control strategy in the same context (Carrasco et al. 2002, 2004; Guwy et al. 1997; Muller et al. 1997; Murnleitner et al. 2002; Polit et al. 2002; Pullammanappallil et al. 1998; Punal et al ; Scherer et al. 2009). For this reason, we have developed a management support system based on fuzzy logic. This system inputs information regarding the plant that is widely available (i.e., feeding, biogas yield and the percentage of methane in the biogas). The outputs of the system support the operator in adjusting the feeding strategy to maintain the stability and productivity of the digester. This paper describes a simple management support system based on fuzzy logic and the results of its assessment in a full-scale operation. 2. MATERIALS AND METHODS 2.1 Plant monitoring The model, reported in the next paragraph, has been assessed in a full-scale biogas plant monitored for nine months (january- september). The plant under study belongs to the CSTR (continuous stirred tank reactor) category and operates in the mesophilic range. It is fed with silage (lolium, triticale, maize and sorghum) and rice chaff. The plant was monitored by filling out a daily form containing the most important parameters registered by the power board such daily power production, biogas consumption, digester temperature, percentage of methane contained in the biogas, feeding rate (m³*d ¹, the daily amount of biomass put in the digester), composition of the daily feed (quality and quantity of each biomass used). 2.2 Implementation of the model No complex mathematical relationships are required to develop fuzzy logic applications. These application are also conceptually easy to understand, flexible and tolerant of imprecise data, and allow the modeling of complex non-linear functions. Application based on fuzzy logic may be developed based on the experience of experts and may also be mixed with conventional control techniques. The use of a rule base in the frame of fuzzy logic allows the translation of the values of the variables measured into linguistic labels given by membership functions. The membership functions give a degree of membership to a fuzzy set for a given input numerical value and set the output value. The rules represent expert knowledge in a computable way Punal et al. (2001) Model calibration To realize the membership functions, we used data that were collected during plant monitoring and determined the mean of each parameter; minimum and maximum values were defined, and a confidence interval with a significance level of 5% was calculated. Values resulting from abnormal situations caused by technical problems that were independent of biological stability were removed from the data set. For each parameter,

3 A simple fuzzy logic management support system for farm biogas plants a moving average of seven days was calculated to absorb insignificant shock and maintain a daily memory of operations that had been previously undertaken. The same values has been used for the defuzzification of the model outputs Model Assessment The model was developed using the Toolbox Fuzzy Logic from Matlab software. The model returns the daily changes, to be made to the feeding rate and to the quantity and quality of the organic matter of feed. Results provided by the model were compared with the plant condition. For each day of monitoring, we evaluated the indications supplied by the model as indicated by the three output parameters. To do so, we compared each output parameter with its the corresponding input parameter. By doing so, it was possible to gain a general overview of the digester and the necessary interventions to maintain or re-establish the optimal production levels. 3. MODELING The model was implemented using as input, five parameters: OLR organic loading rate (kgom*m ³digester*d ¹) to quantify the organic matter contained in the feed; ABP - anaerobic biogasification potential (m³potential biogas*kg ¹ OM) to specify the quality of the organic matter introduced into the digester; Feeding Rate (m ³feed*d ¹) completes the overview of the feeding dilution; GPR gas production rate (m³biogas*m - ³digester*d ¹) supplies information regarding the production conditions of the digester; %CH 4 - percentage of methane contained in the biogas, is both a stability indicator and a production indicator. The output parameters are the OLR, ABP and feeding rate, but their values will be provided by the model as a limit to be respected to maintain or re-establish the stability and productivity of the digester. After choosing, each input parameters must be fuzzified through the construction of membership functions. The fuzzification phase of the inputs allows the real values assumed by the parameters to be converted to linguistic fuzzy values. Three triangular membership functions have been used for each input, representing low, medium and high values (Figure 1) and for each output (increase, not change, decrease). Figure 1. An example of membership functions of the input and output parameters After fuzzyfication the rules are defined. A fuzzy rule base system is constructed that connects the input variables to the output variables by means of if then rules. Given particular values of the input variables, the degree of fulfillment of a rule is obtained by aggregating the membership degrees of these input values into the respective fuzzy sets.

4 A. Finzi, G. Cocolo, F. Perazzolo The fuzzy output is determined by the degrees of fulfillment and the consequent parts of the rules Adriaenssens et al. (2004). An example of a rule designed for model is as follows: IF OLR (high) & ABP (high) & feeding rate (low) & GPR (not low) & CH4% (not high) THEN OLR (decrease) & ABP (decrease) & feeding rate (increase). Finally with defuzzification, it is possible to obtain from the output parameters a resulting value that indicates necessary feeding changes and their amount, in terms of OLR, ABP and Feeding Rate. 4. RESULTS AND DISCUSSION 4.1 Plant monitoring The data collected during monitoring have allowed to assess the functioning of the plant. The average ABP of the feeding was calculated to be about m³ biogas*kg OM ¹, with a feeding rate of 21.5 m 3. The values assumed by the process management parameters OLR (3.62 kgom*m -3 *d -1 ), GPR (2.52 m3biogas*m -3 digester) and Percentage of methane in the biogas (53.2%) given plant performance agreed with literature values (Bauer et al. 2009; Cavinato et al. 2010; Cecchi et al. 2005). During the monitored period, it is possible to identify two different run critical times. First, from late January to mid-may. During this period, the process slowed down as shown by a moderate increase in the percentage of methane in biogas and the ph of the digester contents. At the same time, there was a decrease in the OLR, ABP and Feeding Rate. The second digestion regime occurred from mid-may to late June and was marked by a slight overload evidenced by a decrease of methane in the biogas, an increase in VFA and a decrease in the ph of the digester contents. At the same time, the OLR and Feeding Rate increased and the ABP decreased due to the increased use of silage with the considerable decrease in the availability of rice chaff. 4.2 Model calibration and assessment The mean, minimum and maximum values used for calibration of the membership functions are reported in Figure 1. The output values (OLR/output, ABP/output and Feeding Rate/output) provided from the model are shown in Figures 2-4. This values were verified by comparing them with the behavior of the plant, in particular, the input values of the model (OLR/input, ABP/input and Feeding Rate/input, i.e., the values derived from the monitoring of the plant). Overall, we observed that every time the input parameters assumed values that were not close to the average and therefore potentially problematic, the model either responds by appropriately changing the output parameters or does not make changes if the situation appears to be stable. The following are some examples that confirm the potential of the model in managing anaerobic digestion. On March 10, the model produced output values of OLR (Figure 2) and Feeding Rate (Figure 4) far exceeding those of the input. This was justified by the low values of input

5 A simple fuzzy logic management support system for farm biogas plants from the methane content of the biogas, indicative of a slowing of the anaerobic digestion. On June 7 the output values of OLR (Figure 2) and Feeding Rate (Figure 4) was characterized by input values of OLR and Feeding Rate that were too high and accompanied by concentrations of methane just below average, and so symptoms of instability and a tendency to overload were observed. On May 25, the output of OLR indicates that the model corrected the value of the input slightly, despite the low values of GPR and methane content in the biogas. This is explained by the fact that the input values of ABP (Figure 3) and Feeding Rate are low, because on this day the ABP rose slightly while the extent of the output proposed by the model indicated a condition of instability. During the first days of July, the output of the ABP decreased, apparently incomprehensibly (Figure 3). This was due primarily to the high value of the input, the low concentrations of methane and high value of GPR, but also due to the fact that the low OLR and the Feeding Rate were just above the average, which did not give the model alternative ways of handling the overload. Figure 2. Model indications related to OLR output parameter

6 A. Finzi, G. Cocolo, F. Perazzolo Figure 3. Model indications related to ABP output parameter Figure 4. Model indications related to feeding rate output parameter 5. CONCLUSIONS The main characteristic of the developed model is its simplicity and thus its applicability to full-scale farm biogas plants. Together with the advantages of fuzzy logic, it uses parameters that are easily obtainable in practice. The developed

7 A simple fuzzy logic management support system for farm biogas plants management support system appears to provide the correct and proportionate indications to modify the feeding to restore or maintain optimal digester conditions. At present, the model can effectively support the management of biogas plants, but more test and further validation of the system are required in different plants to implement the model in a process control device. REFERENCES Adriaenssens, V., De Baets, B., Goethals, P.L.M., De Pauw, N., Review: Fuzzy rule-based models for decision support in ecosystem Management. Sci Total Environ 319: Bauer, A., Mayr, H., Hopfner-Sixt, K., Amon, T Detailed monitoring of two biogas plants and mechanical solid liquid separation of fermentation residues. J Biotechnol 142: Carrasco, E.F., Rodriguez, J., Punal, A., Roca, E., Lema, J.M Rule based diagnosis and supervision of a pilot scale wastewater treatment plant using fuzzy logic techniques. Expert syst appl 22: Carrasco, E.F., Rodriguez, J., Punal, A., Roca, E., Lema, J.M Diagnosis of acidification states in an anaerobic wastewater treatment plant using a fuzzy based expert system. Control Eng Pract 12: Cavinato, C., Fatone, F., Bolzonella, D., Pavan, P Thermophilic anaerobic codigestion of cattle manure with agro-wastes and energy crops: Comparison of pilot and full scale experiences. Bioresour Technol 101: Cecchi, F., Battistoni, P., Pavan, P., Bolzonella, D., Innocenti, L Digestione anaerobica della frazione organica dei rifiuti solidi. APAT. Guwy, A.J., Hawkes, F.R., Wilcox, S.J., Hawkes, D.L Neural network and on-off control of bicarbonate alkalinity in a fluidised-bed anaerobic digester. Water Res 31: Muller, A., Marsili-libelli, S., Aivasidis, A., Lloyd, T., Kroner, S., Wandrey, C Fuzzy control of disturbances in a Wastewater treatment process. Wat Sci Tech 31: Murnleitner, E., Becker, T.M., Delgado, A State detection and control of overloads in the anaerobic wastewater treatment using fuzzy logic. Water Res 36: Polit, M., Estaben, M., Labat, P A fuzzy model for an anaerobic digester, comparison with experimental results. Eng Appl Artif Intel 15: Pullammanapillil, P.C., Svoronos, S.A., Chynoweth, D.P., Lyberatos, G Expert system for control of anaerobic digesters. Biotechnol Bioeng 58: Punal, A., Rodriguez, J., Franco, A., Carrasco, E.F., Roca, E., Lema, J.M Advanced monitoring and control of anaerobic wastewater treatment plants: diagnosis and supervision by a fuzzy-based expert system. Wat Sci Tech 43: Punal, A., Palazzotto, L., Bouvier, J.C., Conte, T., Steyer, J.P Automatic control of volatile fatty acids in anaerobic digestion using fuzzy logic based approach. Wat Sci Tech 48: Scherer, P., Lehmann, K., Schmidt, O., Demirel, B Application of a Fuzzy Logic Control System for continuous anaerobic digestion of low buffered acidic energy crops as mono substrate. Biotechnol Bioeng 102: