Microgrid Optimization: Buzzword or Must Have?

Share Button

Scott Kessler, microgrid strategy  and sales at Siemens Digital Grid, explores the ins and outs of microgrid optimization and three approaches to help determine the correct optimization strategy. 

microgrid optimization

Scott Kessler, microgrid strategy and sales, Siemens Digital Grid

Just a few years ago, microgrids were simple.  A diesel generator sat for long periods until a facility manager manually disconnected the site from the grid.  The generator ran until the power came back on or the fuel supply ran out.  All in all, managers faced a limited number of decisions when it came to managing a microgrid.

Now, microgrids are much more complicated.  They often have multiple generation sources, battery storage systems, controllable energy consumption devices and have live connections to the grid. It has become essential for microgrids to be intuitive when it comes to knowing what to do and when.  This helps ensure the best outcome for the lowest cost.

Here is where optimization comes in. Granted, some think optimization is a buzzword, created to sound impressive without any specificity. The fact is understanding the values of various types of optimization helps ensure microgrid owners meet their goals. Optimization automates the best results for a system, considering its site-specific limitations. Optimization is made up primarily of four core concepts:

  1. Objective function: It defines the objective of the solution. For example, the objective could be to minimize the cost for supplying load. How the control variables influence that cost must be known
  2. Control Variables: These are the levers you can use to achieve your objective. For microgrid operations, typical control variables are power output of generating/storage units or load connection statuses, etc.
  3. Constraints: Constraints address the laws of physics such as generation output cannot be outside of the operating range of the generating unit or the load must exactly match the generation
  4. Model: The optimization works on a model, where the model defines number of entities (DERs) participating in the optimization, the constraints, the relationship of the control variables to the objective function etc.

A technique that strictly adheres to these principles and method of solving a problem is referred to as “optimization”. The optimization “engine” finds the optimal values for the control variables that result in achieving the objective while respecting the constraints.

Here are some good starting points for anyone interested in optimizing a new microgrid project.

Greedy Optimization

Greedy optimization is best described as “Doing the best thing at this exact moment without consideration for the larger picture.”  In the classic traveling salesman scenario, a greedy algorithm would send the salesman to the nearest city every time, rather than thinking about the best overall route for visiting every city. This solution usually works well for systems that do not have complexity – such as the shortest distance between two cities.  If comparing a grid price against a static cost of operating behind the meter assets is all that is needed, then a greedy algorithm should do the job.

But, as microgrids get more complex, an optimization strategy that is solely reactive, like a greedy optimization, can leave money on the table. For example, a greedy algorithm may decide to discharge a battery when faced with insufficient generation not realizing that in one-hour grid fees would increase. In that moment, meeting the energy deficit with stored energy instead of using other means makes sense.  However, taking a step back, the same amount of kWh would have had a greater cost impact if the grid energy was imported and the stored energy used later.

Rules-Based optimization

True to its name, rules-based optimization is predicated on rules that are defined in advance. It lets a microgrid know in advance “If this, then that.”  While it sounds simple, rules-based optimization enables a range of pre-programmed scenarios, which can yield improved results over a Greedy optimization, especially when combined with future price schedules.  In our previous example, a well-designed, rules-based optimization strategy would inform a controller that if there is an increase in grid fees coming in an hour, then wait to discharge the battery until the opportune time.

Rules-based optimization allows greater flexibility in the design stage but requires knowledgeable engineering during configuration. It also requires programming of prices in advance, so is not well suited for comparisons between multiple factors with uncertain future costs. Things can get even more complex when adding in the ability to predict weather, or when the user is allowed to alternate between modes – allowing optimization changes for resiliency, cost savings and sustainability.   A poorly designed rules-based optimization can result in headaches and change orders. However, if configured properly, this type of optimization is well suited for small projects, such the Princeton microgrid lab example.

While optimization can sometimes seem like an amorphous term, being knowledgeable about optimization at the beginning of a microgrid project, and building in the right strategy during project design, can improve costs and ROI as well as reduce operational pain for a microgrid owner.

Model-based optimization

Model-based optimization is the most powerful form of microgrid optimization.  It requires user-defined system limitations (what is not allowed to happen), a representation of the system (often like a one-line diagram) and live data feeds (usually weather and energy pricing). From this, the model continually considers thousands of potential operating scenarios and picks the one with the best outcomes.  This is most useful when a system is asked to consider optimizing for multiple objectives.

Consider the example of a generator, battery and PV system. Most systems can be designed to optimize for resiliency, which will ensure the battery stays  charged at all times, or to optimize for cost savings, where the battery will try to arbitrage as much as possible between the cost of grid energy and the cost of behind the meter generation. However, other forms of optimization cannot consider both outcomes simultaneously. Model-based optimization excels at this, and will do so without requiring explicit programming instructions.

A powerful example of model-based optimization is found in the system installed at Blue Lake Rancheria. When Siemens designed the system, we were informed the site was not allowed to the export to the grid. Following that, the preference was to reduce cost to the site, primarily through charging a battery at night when grid prices were low.  During expensive afternoon peak pricing, the system made as much use as possible of both PV and battery-stored energy. However, during the COVID-19 pandemic, energy use at the site was drastically decreased, resulting in an excess of solar energy in the afternoon that had to go somewhere. Knowing its prime limitation on site was to prevent grid export, the Siemens controller automatically self-adjusted, discharging the battery at night to ensure there was sufficient capacity to consume excess, unused solar energy.

Model-based optimization is incredibly powerful. While it may be overkill for smaller projects, it can have significant ROI impacts on larger systems.


Knowing what’s right for you

No solution is right for every system. To determine the correct optimization strategy, it is important to understand: How important is upfront cost vs operational costs? Is this microgrid likely to experience changes in DERs or value streams? Do I want to be able to make changes myself or do I have a trusted partner to make adjustments?  Do I need optimization for just my generation, or do I want to optimize for generation, storage and loads? Do I need optimization to occur locally, or am I OK with the calculations requiring internet connectivity?

While optimization can sometimes seem like an amorphous term, being knowledgeable about optimization at the beginning of a microgrid project, and building in the right strategy during project design, can improve costs and ROI as well as reduce operational pain for a microgrid owner.

Scott Kessler works in microgrid strategy and sales at Siemens Digital Grid.

Share Button


  1. Model based, interesting. I believe most solar PV and smart ESS installed at residential and small businesses should be able to monitor loads and control these loads by grid power, a mix of grid and stored energy or smart load shedding, where electricity is supplied by the smart inverter and energy storage system at the most efficient generation rate.