IntroductionBSM1_LT is a natural follow-up of BSM1. Although a more thorough discussion will follow, we would like to start by emphasizing the differences between BSM1_LT and BSM1:
Evaluation time period
It is known that seasonal effects have a significant impact on the operation of a treatment plant. Also, typical failures, be it equipment or process failures, do not occur more than a few times per year. To include such phenomena in the assessment of a control or monitoring strategy, the required evaluation period is at least one year. It is realistic to assume that during a year of operation, even less probable faults and disturbances will occur and affect the operation significantly. The start and stop of the evaluation period is preferably set to summer, beginning of July and end of June, respectively, since the relatively favourable summer conditions would minimize the risk that effects of a proposed control strategy are pushed forward, beyond the end of the evaluation period.
Since the evaluation period spans one year, temperature will be a determining factor. A common trajectory for the temperature over a year is more or less sinusoidal with its maximum value at the start of August and its minimum value at the start of February (in the southern hemisphere this is reversed). A more complex description of the temperature may be necessary to emulate the changes in temperature due to precipitation or snow melting. Such a description remains to be decided. As a starting point, however, we believe that a sinusoidal as described would suffice as an approximation of the temperature trajectory. The temperature is modelled as °C, where t is the time in days and the shift is 28 days (4 weeks).
The values of the temperature dependent kinetic parameters in the ASM1 model are, consequently, varying during the evaluation period. The BSM1 parameter values are defined at 15°C, but rounded to one or two decimals. Since it is desirable that BSM1 and BSM1_LT have exactly the same parameter values at 15°C, the Arrhenius function should be based on values at 10 and 15°C, using the BSM1 values for 15°C.
It should be noted that the saturation concentration for dissolved oxygen is temperature dependent. This has an impact on the mass transfer rate of oxygen, since it is modeled as KLa·(SO,sat- SO). KLa is also temperature dependent. In BSM1, and also in the proposed BSM1_LT, the oxygen mass transfer rate is expressed as KLa. This means that no temperature compensation is required for KLa. However, as soon as KLa is to be expressed in terms of energy (an important evaluation criterion), the temperature dependency is crucial. Currently, the aeration-energy relationship is defined at 15°C only.
Influent file design
The design of the influent file is based on the BSM2 influent file. However, there is a difference: since the influent file of BSM2 is defined as the influent to the plant, whereas the influent file in BSM1 is defined as the influent to the biological treatment, there is a need to modify the BSM2 influent file for use in the BSM1_LT. This is done by simply letting the influent file of BSM2 be applied to the same type of primary clarifier used in BSM2 and the output of the clarifier becomes the influent file of BSM1_LT.
It is well known that sensors have dynamic properties, they are afflicted by noise, drift is commonly occurring and that sensors need to be calibrated and maintained in a reoccurring fashion. To emulate such sensor behaviour and test the control or monitoring scheme for its robustness and/or sensitivity to this is crucial. Realistic sensor model behaviour requires the dynamic properties and disturbance sources to be represented. This has been discussed in Rieger et al. (2003) and here we have adopted their approach to sensor modelling. This includes modelling of noise, time response, drift, signal saturation and, if not a continuous sensor, the measuring interval. Rieger et al. (2003) also include a procedure for modelling calibration and maintenance of the sensors. This is an important aspect for both control and monitoring. A control or monitoring system must handle periods of no or non-representative data during sensor calibration/maintenance.
Sensor and actuator failures
Sensor and actuator failures occur in all industrial environments. It is therefore important that the impact of failures on the behaviour of a control system as well as a monitoring algorithm can be evaluated. Sensor and actuator failures can be modelled in different ways, but this aspect is not included in the work of Rieger et al. (2003). A straightforward way is to assume that a failure occurs according to a Poisson distribution, with a specified failure rate. A failing sensor may result in an erroneous signal, no signal at all, change in noise level or excessive drift. An actuator failure may manifest itself as, for example, a complete or partial loss of capacity. The time for repair may of course vary. Rather than including all factors that may influence the time for repair, the duration of a malfunction has been assumed to be random as well. See Rosen et al. (2008) for a detailed description of the sensor and actuator fault model.
Process disturbances are defined as disturbances acting only through the model parameters, e.g. maximum specific growth rates, settling parameters, etc. Thus, process disturbances are emulated using time varying process parameters. For instance, temporary nitrification inhibition and toxicity is imposed on the simulation model by reducing the maximum specific growth rate and the decay rates. Variation of the sludge settling properties due to influent changes, seasonal effects or sludge age alternation is another typical process disturbance. Varying the parameters of the Takács settling velocity function can simulate this behaviour. The particular type of process disturbances will be described in detail in later reports.
Additional control handles
The long-term perspective of the BSM1_LT allows for other control handles than those of the BSM1. Equalization tanks and sludge storage tanks have been already discussed within the benchmark community and could be implemented as a result of the proposed extension. Wastage flow rate is another control handle that, even though it is available within the BSM1 definition, does not have a stable effect within a week. However, an evaluation period of a year creates possibilities for, for instance, sludge retention time (SRT) control.