Description
The optimization-based tool of MIKE+ WD aims to allow the schedule of pumps and operation of control valves.
The optimization is available for two algorithms:
- The Shuffled Complex Evolution (SCE) which is a global optimization algorithm that uses competitive evolution, controlled random search, the simplex method and complex shuffling as search strategies.
- The Dynamic Dimensioned Search algorithm (DDS), this algorithm is intended for a large number of parameters. It scales the search to find the best solution within the maximum number of model calls automatically. It initially searches more globally before becoming more local as the maximum number of model calls is reached.
The logic of the optimization setup is simple, it is based on Controls and Targets. Controls are the network structures to be optimized in this case pumps and valves. Targets are the setup goals such as flows, pressure or water levels.
The pump can be switch ON/OFF versus time working on an hourly basis or following a repetitive pattern, the pump speed can be controlled versus time. A valve can be controlled in terms of the opening percentage or flow set point versus time.
The targets, as described, are desired levels or thresholds, in example; water levels in tanks or reservoirs, water quality at a node, total volume or volume difference in a tank, maximum-average or total flow through a valve, pump power and energy costs and finally sources water balance.
The optimization can investigate multiple settings and control strategies at the same time.
The Dynamic Dimensioned Search algorithm method has a series of stop criteria used to define the conclusion of the optimization process. The maximum number of model calls is the first stop criteria if the number of calls is reached the process stops and the best solution found is reported. The Setpoint target objective is another stop criteria, it simply compares the value of the objective function and stops when the value is below the data entry.
Following the criteria "Use Maximum Run Time per Simulation" will stop the simulation if it takes an extreme long time to find a solution, "Use maximum number of invalid solutions per simulation" is another stop criteria but in this case the criteria take place if by a "random" choice of control variables by the optimizer results in unbalanced or hydraulic unstable conditions during the extended period simulation run. With the last two criteria the hydraulic simulation is cancelled, and it will use a high penalty for its settings for the optimizer. (The program automatically generates very high penalty so this particular model alternative will not be used any further).
Example
Following, a simple example on the tool functionality is presented. The model presented is a small WD model consisting of five wells from where the water is pumped to the Water Treatment Plant. The wells are represented by a sump tank and a pump. The purpose of the system optimization is to provide an equal split of flows from each of the 5 wells. Fig. 1 presents the network horizontal layout.
Fig. 1 - Model map layout view
The model consists of 5 water tanks of constant HGL at 5 meters. Fig. 2 presents the properties configuration of these tank series.
Fig. 2 - Tanks properties
A series of junctions are created to define the geometric definition of the network. Thus, these junction nodes have no demand assigned. See Fig. 3.
Fig. 3 - Network junction properties
The flow in the network is pushed by 5 pumps of the constant speed (CSP) type defined by a 1-point curve operating at a head of 50 meters with a flow of 25 liters per second. Fig. 4 presents the overall pumps configuration.
Fig. 4 - Pump properties
When the extended period hydraulic simulation is run, the five Pump Stations (PS) show different flow rates, all within the operation 1 point curve. As shown in the plot depicted in Fig. 5, all pumps discharge within the curve rate.
Fig. 5 - Pumps discharge on extended period simulation without optimization
Further, it is desired to optimize the pump operation with the Optimization tool using the Dynamically Dimensioned Search (DDS) algorithm. The optimization is set to use a maximum number of calls of 500, with a maximum run time per simulation of 900 seconds, see Fig. 6.
Fig. 6. DDS maximum number of calls and run time stop per simulation
A control ID is set for each pump station in the model, following a Control type of Pattern of 24 hours, as presented in Fig. 7. There are two options the pump can be set ON/OFF or relative speed and the valve settings.
Fig. 7 - Optimization Control ID
Accordingly, there is a target defined for each controlled structure, as presented in Fig. 8. Targets are understood as goals for the network hydraulics, these types of goals are the tank water level, water quality, pressure, flow, pump power, pump energy cost and source water balance as the water supplied from a tank reservoir.
Fig. 8 - Optimization targets settings based on the source water balance
Running the tool with these settings leads to almost achieving the optimization goal as shown below in the tool Outputs tab, check at the Target (Computed) column in Fig. 9.
Fig. 9 - Output of the optimization process
Achieving the desired optimization seems possible, thus the optimization setting is updated increasing the maximum number of calls to 1000, as shown in Fig. 10. The optimization result shows the desired distribution of 20% inflow from each water tank.
Fig. 10 - Optimization result with an increased number of maximum calls (1000)
Looking at the Output and Plots tabs it is possible to observe the targets being achieved and the plot results. A 20 percent of the water demand is coming from each of the sources. Fig. 11 presents the optimization output.
Fig. 11- Outputs after increasing the number of maximum calls
The Plots tab presents time series graphs with computed controls and the optimization settings. Fig. 12 presents the optimization setting for the pumps in the example.
Fig. 12 - Optimization plots after achieving the target of 20 percent source distribution
The report option will export the output tables to a HTML file, this report will be structured in Control Data tables and Target Data regarding the optimization. Plots are not included in this HTML report. An example of the HTML Report is presented in Fig. 13.
Fig. 13 - Example of a HTML Report
Conclusion
The WD optimization provides the user a tool to improve their operating structures (pumps and valves), the tool provides a schematic definition of the structure operation to fulfil the desired operation targets. The tool presents time series graphs with the computed controls and their optimized settings.
FURTHER INFORMATION & USEFUL LINKS
Manuals and User Guides
MIKE+ Water Distribution
Release Notes
MIKEPlus Release Notes
[Training options]
MIKE+ WD | Getting started with water distribution modelling