General Description

The test case provides a set of conditions under which a test can determine whether or how well a system, component or one of its aspects is working given its expected function.

Name of test case:

1. Model Predictive Controller characterization/validation
2. Building model verification

Partner: KIT-IAI

Scope and Goal

What is the test objective, i.e. what is the purpose of carrying out the test? Try to formulate the objective within one of the following three categories:

1. Characterization: a measure is given without specific requirements for passing the test. Examples: understanding the behaviour of a system, developing a mathematical model of a component.
2. Validation: a requirement and abstract measure is provided, but the results are subject to interpretation, i.e. passing a test depends on a qualitative evaluation by an expert or user of the system. These tests seek to answer the question are we building the right system? Example: Is the mathematical model good enough?
3. Verification: acceptance of a test result depends on the direct evaluation against fixed and formalized assessment criteria. These criteria can be formulated as quantitative measures with a set/range of acceptable values of these measures, i.e. quantitative tests. These tests seek to answer the question are we building the system right? Example: Testing whether a component conforms to a standard.

Two test cases have to be evaluated:

1. 1. The developed Model Predictive Controller has to be characterized. The interaction with the building models has to be studied and evaluated. After optimizing the MPC algorithm, it is validated with the simulation results for energy consumption reduction and fluctuation minimization.
   2. The building models are verified by comparison of the simulation trajectories and physically measured values.

To which system configuration does this test apply? KIT building 445-449

To which use case does this test apply? UC11

Formulate one paragraph of narrative summarizing the test and its purpose:

In case of UC11, the developed MPC algorithm has to be tested against three test criteria. First of all, whether the MPC interacts properly with the building models and is therefore able to control the system within given boundaries. Next, the total energy consumption (including thermal and electrical energy) should be reduced. Furthermore, the fluctuation in energy supply from the district heating and the electrical grid should be reduced. In these cases, an intermediate step of building modeling is needed. The resulting building model should also be verified by given measurement data.

For the purpose of reducing the energy consumption over a given time period, the reduction of total energy consumption in kWh could serve as measure for success of the underlying control algorithm. In case of fluctuation minimization, there are multiple possible candidates which could serve as measure for fluctuation. Two possibilities are to measure the deviation from the moving average of the supply either in the l1 or in the l2-norm. In the former, the measure would minimize the peak deviation whereas in the latter an average-like deviation would be minimized. The building model can be validated by comparison of resulting simulation results to real measurements.

Identification of Test Components

What is the System Under Test, i.e. which subset of the entire system configuration needs to be simulated in order to achieve the test objective? (Note: In some cases this may be the entire system)

• The System Under Test is in both cases the whole building.

What is the Object Under Investigation, i.e. which are the components of the System Under Test that are to be characterized or validated?

In both test cases:

• Heat pump and AC system dynamics
• Concrete core activation
• Electrical / heat grid
• Thermal dynamics of the building

What is the Function Under Investigation, i.e. which of the system behaviour defined in the use case is to be characterized, validated or verified?

• In case 1) “Model Predictive Controller characterization/validation” the Model Predictive controller, but with different objective function formulations according to the desired goal.
• In case 2) the deviation of simulated and measured load data.

Test Criteria

Formulate the target metrics, i.e. a list of quantities in the context of the System Under Test which can be used to qualify/quantify the test result with respect to the test objective.

1. 1. Model Predictive Controller characterization/validation
      1. Boundaries
         • The MPC can keep the system within the given boundaries (e.g. temperature comfort zones).
      2. Energy consumption reduction
         • Reduction of energy consumption over a given time period (e.g. a month / a year) in kWh.
      3. Fluctuation minimization
         • The maximum peak in electricity / heat demand has been reduced.
         • The derivation around a moving average for fluctuation has decreased.
         • Different metrics could be used. E.g. the deviation from a moving average in different norms (l1-norm or 12-norm for example).
   2. Building model verification
      • • The deviation of the predicted behavior against measured data. Trajectories of the developed model (ordinary differential equation) behave in simulation like the real system (validated by comparison of the simulation trajectories and physically measured values, e.g. heat/electricity load curves.)

Define quality attributes for assessing an acceptable test result. In case of a characteri-

zation test, this may be the remaining model uncertainty. For verification tests, the acceptance threshold (worst case for passing the test) is stated. For validation tests a criterion for ending the test execution is required.

• The quantitative measures for ending the test case have to be developed during the modeling task.