Tuesday, March 1, 2016

Optimizing your utility energy with MINLP in SimSci ROMeo



By: Rajkumar Vedam & Samantha Weaver of Schneider Electric

Utility energy optimization has the potential to provide significant cost savings through good management of your energy resources. Typically, energy optimization involves switching in and switching out different energy sources depending on market conditions. Traditional nonlinear programming (NLP) solvers, while powerful, are not suitable to investigate cases that require this kind of switching. Switching in or switching out energy sources changes the fundamental process equilibrium of the simulation, which will cause problems for traditional NLP solvers. In these cases, mixed integer nonlinear programming (MINLP) provides the means to optimize processes by taking into account the new nonlinear process equilibrium introduced by the changing energy sources.

Mechanics

Schneider Electric Software’s ROMeo from the SimSci brand offers a MINLP solver that can solve the advanced MINLP problems that are inherent to utility energy optimization. The MINLP problem expands the NLP problem to include a set of integer variables representing the on or off state of the different energy sources that must be accounted for. Using the powerful and reliable NLP solver in ROMeo, the MINLP solver solves a series of NLP problems, each with a unique configuration of energy sources. It uses the branch-and-bound algorithm to prune out infeasible or non-optimal solutions to arrive at the most optimal configuration of energy sources.

Figure 1: An example of a MINLP solution tree for a model application that contains 5 integer variables.

The MINLP solver is customizable and includes a set of MINLP solver tuning parameters that allow you to fine tune the MINLP solver to provide for faster and more efficient solutions. You can manage the MINLP solver tuning parameters in the same way that you manage the NLP solver tuning parameters, that is, by using the ROMeo Solver Manager.
The MINLP solver can switch on or switch off only the energy sources that you expose to the MINLP solver. ROMeo includes a non-process unit, the MINLP Switch, that you can attach to your energy sources. This allows the MINLP solver to access the variables and parameter within the energy source and selectively switch on or switch off those energy sources. In this way, you have full control of the optimization processes.

Easy Conversion between NLP and MINLP Operations

ROMeo includes a specialized MINLP calculation mode that allows you to easily switch between MINLP operations and traditional NLP operations within ROMeo. You can manage the MINLP calculation mode in the same way that you manage the other ROMeo calculation modes through ROMeo’s Mode Manager.

Easy Access to MINLP Variables

Because a single model application may have numerous MINLP-switchable energy sources—and thus numerous MINLP variables—ROMeo includes the MINLP Manager. The MINLP Manager consolidates the MINLP Switch information for easy access and configuration. It also provides a means to group the integer variables in the MINLP Switches, which reduces the computational load on the MINLP solver.

Efficient Design: Grouping

Due to the larger number of variables and NLP solutions in question, MINLP solver runs can be costly in terms of processing time. To reduce the computational load on the MINLP solver, ROMeo provides grouping mechanisms for related integer variables.
Specifically, the MINLP Manager allows you to group variables that turn on or turn off depending on the values of other integer variables. For example, if you have a Motor that turns on when another Motor turns on, you can group the integer variables for the two Motors.
You can also specify complements with the grouped variables. For example, if you have a Steam Turbine that must be off when a Motor is on, you can group the integer variables for the Steam Turbine and the Motor and set the Steam Turbine’s integer variable as a complement.
Grouping greatly improves the efficiency of finding an MINLP solutionin terms of both speed and robustnessby restricting the MINLP search to the specified constraint space.

Types of Utility Energy Optimization

The MINLP solver currently allows for three types of utility energy optimization: Utility Units, Fuel Sources, & Parallel Streams. A single ROMeo model application can include all three types. More information on all three types of energy optimization and the associated business values are found below.

Utility Units

The first type of energy optimization allows you to optimize the configuration of utility units in your plant. Your plant model may contain numerous utility units, such as Motors, Generators, and Steam Turbines. You can specify economic data for these units and attach MINLP Switches to them. The MINLP software can then switch on or switch off these units during an MINLP solver run. The MINLP solver finds the optimal set of utility units that should provide the energy requirements for your model application. You can choose to keep this MINLP solution or discard it.

Figure 2: AN MINLP model application that optimizes the utility unit configuration between a Motor, Steam Turbine, and Generator.

Fuel Sources

The second type of energy optimization allows you to optimize the fuel sources that are used in the combustion heating of your plant. Specifically, the MINLP solver can optimize the combustion fuels that are fed to a boiler. A boiler may have multiple combustion fuel sources. The MINLP solver can switch on or switch off the fuel streams based on the economic data provided for those fuel streams. This applies to both the Boiler and the ERTO Boiler models in ROMeo.

For Boilers, you can specify economic data for the fuel streams in their associated Sources and attach MINLP Switches to the streams. The MINLP solver can then switch on or switch off these process streams during an MINLP solver run. The MINLP solver finds the most economical combination of fuels and turns off any fuel streams that are not needed to meet the energy requirements of the Boiler.
 
For ERTO Boilers, if the fuel streams are external to the ERTO Boiler model in the model application, you can specify economic data for the fuel streams in their associated Sources and attach MINLP Switches to the streams. The MINLP solver can then switch on or switch off these process streams during an MINLP solver run. The MINLP solver finds the most economical combination of fuels and turns off any fuel streams that are not needed to meet the energy requirements of the ERTO Boiler.

Figure 4: An MINLP model application that optimizes the fuel sources for an ERTO Boiler with external fuel streams.

If the fuel streams are internal to the ERTO Boiler model in the model application, you can specify economic data for the fuel streams within the ERTO Boiler model and attach MINLP Switches to the variables associated with the fuel stream in the ERTO Boiler model. The MINLP solver can then switch on or switch off these process streams during an MINLP solver run. The MINLP solver finds the most economical combination of fuels and turns off any fuel streams that are not needed to meet the energy requirement of the Boiler.

Figure 5: An MINLP model application that optimizes the fuel sources for a ERTO Boiler with internal fuel streams.

Parallel Streams

The third type of energy optimization allows you to optimize parallel process streams in your plant. The parallel streams can either terminate at a common Mixer or originate from a common Splitter. You can specify economic data as you would for traditional model applications. Typically, the parallel streams are the beginning or end of parallel process trains that are similar but contain different configurations of utility units. In this case, you attach MINLP Switches to each parallel stream. The MINLP solver can switch on or switch off the parallel streams and by extension, the parallel process trains in your plant. The MINLP solver finds the optimal process train that should provide the energy requirements for your model application.


 





Figure 6: Two MINLP model application that optimize parallel process streams within a model application. The first model application (A) uses parallel streams that terminate in a common Mixer. The second model application (B) uses parallel streams that originate from a common Splitter.

Conclusion

Good management of your energy resources through utility energy optimization provides significant cost savings. The MINLP solver in ROMeo software effectively extends the range of utility optimization strategies to include dynamically switching on and off various energy resources depending on the prevailing market condition. Use ROMeo for energy optimization to optimize your utility units, your energy sources, or parallel flows to minimize your energy costs and boost your profitability.