Project Scope

Problem Statement/Goals of Study

  • What are the goals of developing this model?
    • What problems do you plan to solve with this model?
    • What systems do you plan to study with this model?
    • Is it an improvement on an existing model? If so, what specifically is it supposed to improve?
    • Does it model a device that is an improvement on an existing device? If so, what specifically is it supposed to improve?
  • What level of interaction is the model expected to capture?
    • Are system-level impacts desired?
    • Is the system constrained to the local distribution feeder/microgrid?
    • Are any transmission or sub-transmission impacts going to be examined?
    • Are inter-device impacts going to be examined?
  • Will the study be examining system level responses and overall, aggregate impacts on the power system, or something more specific? Specific responses may include but are not limited to
    • post-fault
    • extremely hot day
    • cloud transient
    • change in control state or setpoint
    • change in load

Time Scale

  • Does the model need to model microsecond (electromagnetic, also known as EMTP), millisecond (transient), longer (quasi-steady state), very long-term (capacity planning) effects?
  • Is the model operating in a single timing realm, or multiple?
    • If there are multiple time scales desired, is it covered by one model or multiple?
    • If there are multiple time scales and effectively separate models, how do you transition between the two?
  • What is the duration in time of the effect under study?
    • Does the simulation need to run for a fraction of a second, or as long as minutes, hours, or days?
    • What is then the total number of time steps or calculation cycles?
    • If multiple regimes are to be studied, what are the requirements for each (for eg., a steady state period pre-fault as well as a response period post-fault)?
  • Is this total number of calculation cycles feasible on available computing resources?

Quantities of Interest

  • What quantities will be extracted from the model?
    • Are these quantities directly associated with the model, or an interaction with another component (i.e., internal state like air temperature or an external interaction like the current power draw)
    • Are all of the necessary quantities actually knowable? Are they either
      • able to be calculated?
      • able to be derived?
      • able to be measured?
      • able to be researched or looked up?
      • able to be estimated to an acceptable accuracy?
    • Is the study examining impact on system loading via power or voltage levels, or some internal variable like device temperature?
    • Are these fundamental quantities (voltage, current, temperature), or a derived quantity (power)?
  • What is the desired level of confidence in the output?
    • Is this a system-level study where deviations on the order of kW or MW are acceptable?
    • Is this a device-level study where the output should be within partial Watts of the true output?
    • Does the model need to incorporate two different levels of detail with precise small-signal characteristics and broader large-signal characteristics? How will these two different levels interact?
  • Will this model be aggregated or represent an aggregate population?
    • Does it make sense to do an object aggregation?
    • Should there be a single object representing the aggregation?
    • What are any resolution constraints gained/lost by doing the aggregation?

Decision Point: Goals

  • Can the project goals be met given the
    • time step and time duration?
    • resolution and dynamic range of quantities of interest?
  • If no, can they be adjusted given available resources and capabilities such that project goals can be met?
    • If the answer is no, the project should not move forward beyond this point.

Model Capabilities

  • What quantities will be modeled / plotted / predicted / fitted? Possibilities include but are not limited to
    • Frequency?
    • Voltage and Current (magnitude or angle, rectangular, or point-on-wave)?
    • Real or reactive power (magnitude or angle, rectangular, or point-on-wave)?
    • Internal temperature?
    • Control state
      • What is the source of control signal? Does it also need to be modeled?
    • State of charge?
  • What is the necessary temporal resolution of each of the quantities of interest? Is it the same for all? Can some be modeled on a coarser level?
  • What kinds of perturbations or outside influences/events will the system respond to?
    • Specific response? Improved response? Interaction influences?
    • Step change in set points?
    • Change in control parameters or overall control scheme?
    • Change in network topology?
    • Voltage, Current, phase at point of common coupling?
    • Faults and failures?
  • Will the model simply be a self-contained system, be attached to an infinite bus, or reside in a host environment that has other elements that it may interact with?
    • What will be the interface between model and host environment
    • What parameters will be passed back and forth?
    • At what time scale will the model interact with the host environment, and will that differ from the time step internal to the model?
  • What computing resources are available? Should the specified model be able to run in a reasonable amount of time given available computing resources? Is time for iteration and improvement available?

Test Capabilities

  • What types of restrictions are there on the equipment used to impart the disturbance on the device or system of test?
    • Are there specific types of disturbances that can be represented? Can fault analysis be done, but only for specific types of faults? Can harmonic distortion be injected, but only within a certain band?
    • Are there rating restrictions on the "faulting" device, such as sinking current ratings or durations?
  • What types of restrictions are there on the device under test for the validation testing?
    • Are there types of tests that would physically destroy the machine (unless the validation test is to prove this)?
    • Are there specific electrical regions of operation that should not be excited, such as outside a rated frequency range of operation?
    • Are there physical limitations of the device, such as measurement point access or inertial properties?
  • Are there limitations on the measurement equipment in the validation facility that will limit the model?
    • What is the sample rate of any digital-based measurement equipment and is that sufficient for the desired study?
    • Are any of the measurements "human read" manually? If so, timing limitations will need to be incorporated.
    • What is the quantization step and dynamic operating range of each necessary measurement device?
      • Are the step sizes small enough to capture the level of detail the model hopes to represent?
      • Are the upper and lower limits of the measurement range sufficient that the device and model output will not exceed them?
    • Are there specific measurement types (e.g., voltage, current, power, frequency) that are required for validating the model and are they available?
    • Are there any connection restrictions on where the measurements can be taken (which parts of the system to be tested are accessible to measurement device connections)? If so, will that impact the validation of the model?
  • Are there any restrictions on the overall testing facility or system under test (not just a single device)?
    • Does the facility support the voltage levels, current levels, and power ratings desired for testing the device? If not, can a smaller subset of tests be performed and still validate the model?
    • Are there administrative or economic restrictions that may limit the testing?

Acceptance Criteria

  • What quantities of interest will be compared between simulation results and hardware tests?
    • What is the duration and temporal resolution of the signals that will be compared?
    • What signals will NOT be compared?
    • Is the focus of the validation on response to perturbation, or also steady state?
  • What frequency behavior should be captured? Is there an expected bandwidth in which phenomena will be accurately captured?
  • What deviations between simulation and hardware test do you expect?
    • What do you already know you will observe but are not planning to model because it is beyond the scope of the project goals?
    • What level of noise on measurements will you observe and ignore?
    • What level of DC offset is acceptable?
  • If fitting parameters, how will fit be quantified? What is the cutoff for declaring an acceptable fit?
    • R2? X2? Maximum error? Average error? Engineering judgment?

Decision Point: Capabilities

  • Are simulations and hardware capable of addressing acceptance criteria?
  • Will simulations capture all of the quantities of interest defined in the project scope section, at the necessary temporal resolution and duration?
  • Will hardware tests measure all of the quantities of interest defined in the project scope section, at the necessary temporal resolution and duration?
  • Are the fit criteria compatible with the expected level of process noise? Compatible with expected level of measurement noise?
  • If no, can model capabilities, hardware test capabilities or acceptance criteria be adjusted given available resources and capabilities such that project goals can be met?
    • If the answer is no, the project should not move forward beyond this point.

Structure

  • What are the boundaries of the modeled elements and how does information transfer to larger environment?
    • What are interface points between model and system as a whole?
    • If there is a point of common coupling, what information is passed?
    • How often does model interact with system as a whole?
      • At every time step? Event driven? At some longer regular interval?
  • Are there multiple distinct modes of operation and is it necessary to model them separately with a case statement separation? How does transition between modes happen if necessary?
    • Does the model consist of a single equation? Cascaded blocks? Sub-devices? Are there controls to be modeled? Are there "pluggable" models?
  • Are equations exactly solvable?
    • Which components are linear?
    • Which are non-linear (and cannot be implemented as linear approximation)?
    • Does non-linear solution require iterative method? Or quadratic method?
    • Which require case statements?
  • Which simplifications will be taken and transformations used?
    • Normal Cartesian plane?
    • dq0 transformation?
    • Laplace transformation?
    • Sequence transformation (Keep only positive sequence component?)
    • Symmetrical components?
  • Will calculations be performed with harmonic signals represented as phasors, point-on-wave, with the DC equivalent approximation? Will algebraic or differential equations be used?
  • Which numerical methods will be adopted? Euler, Runge-Kutta, Trapezoidal?
  • Will matrix inversion be necessary? Will a matrix decomposition be used to aid inversion?

Parameters

  • What are all of the physical quantities that appear in all of the equations that go into the model?
  • Do you need to translate from common notation into the form that you decide to use?
  • Categorize all parameters as fixed/dynamic, tunable/not tunable
    • Fixed: Which parameters will the model take as input? What are the sources of these quantities? What are uncertainties? What units will be used?
    • Dynamic: Which parameters will be dynamically adjustable or be fit by the model? What are uncertainties? What units will be used?
    • Tunable: Which parameters can be adjusted to improve goodness of fit? Are there parameters that the model is specifically intended to find through fitting, given hardware test data?
    • Not Tunable: Which parameters cannot be adjusted?
  • Would use of per-unit make calculations more straight-forward or robust?
  • What is the slate of initial conditions that will begin the model run?
    • Will they vary between different simulations?
    • How will they be obtained? Flat start? A period of steady state operation before a perturbation occurs? Supplied by larger system environment in which model is hosted?
    • Are initial conditions compatible with the hardware system initial conditions?
  • What is the time step interval? Will it be constant or dynamically adjusted?
  • Will there be a "steady state" period of time before the phenomena of interest begin? What will determine when the model run concludes? An event, or meeting a previously specified condition or previously specified end time?
  • What limits or ratings will affect model operation? Power ratings? Current limits?
  • Which parameters have largest uncertainty or smallest confidence?
  • Which parameters drive stability? Of simulation? Or of model itself?

Simulation

  • Did the simulation execute as expected?
  • Are model results reproducible? Are they in any way changed by adjusting simulation conditions like timestep, calculation accuracy, floating point precision?

System Staging

  • What is environment? Stand-alone operation, or interactions with other devices?
  • Does the system have the necessary elements, both model elements and an appropriate analogue to the host system?
  • Is the external environment robust enough to not limit testing? Do power supplies have high enough rating to not current limit tests? Does other equipment have the dynamic range / rating to not limit any tests?
  • How is event induced? Is there some change in set point or parameter? How will that be affected? Need any modifications to inputs to device? Need a trigger to start recording data in response to event?
  • What are characteristics of measurement equipment?
    • Are there measurement equipment restrictions (like dynamic range or rating)?
    • How will data be recorded and how will data be transferred? Does the recording computer, oscilloscope, etc. have enough input channels at sufficient frequency? Do you have enough space to store all of the data? Are you gathering a quantity of data that is feasible to plot / analyze?
    • Do measurements have to be synchronous? (Common timebase to all measurement equipment)?
    • Are there any significant measurement delays associated with measurement devices?
    • Will test equipment be invasive or non-invasive? Will you use shunt or Hall measurements? Will you use direct measurement of device quantities, or PMU measurements to see system response?
    • How is instrumentation attached to system?
  • Will you have sufficient time for careful tests, will you be able to repeat tests if necessary to iterate?
  • How will tests be documented? Will there be a lab log?
  • Are you holding non-critical parameters constant between tests to be able to focus on interesting differences?

Test Procedure

  • What is expected output of each test? What will be plotted / computed? What will be recorded? How fast? For how long? How will measurement device be connected? How will data be recorded?
  • Is there explicit sequence of events? Is there an event to trigger measurements from? Is there an explicit order to hook things up (consider safety/grounding issues)?
  • Are events automatically occurring or depend on human to throw switch?
  • Will tests respect all safety rules of the test facility? Will any dangerous situations be created by the test?
  • Do you know what's energized and what's grounded? Are grounds common or are any floating?
  • Has steady state operation been verified, and sufficient steady state operation been recorded to compare response to events? Does it need to be?
  • How many times will each test be repeated?

Carry Out Tests

  • Did anything unexpected happen? Are tests reproducible?
  • Did you get all of the data you need to compare to model results?
  • Are noise levels reasonable and as expected?
  • Did any unexpected events or deviations from test plan get properly documented?

Decision Point: Acceptance

  • Do simulation results adequately match hardware results according to previously specified acceptance criteria?
    • Are features that you cannot explain compromising the acceptance criteria or detrimental to model, or are they incidental?
  • If not an acceptable match, do you need to revise your model or do more tests?
    • Is the behavior completely unexpected, the wrong order of magnitude, key features missing, or are there features that are at different frequencies than anticipated? If so might need to revise model.
    • Are there unexpected features in hardware tests? Can tests be repeated to see if unexpected feature is reproducible? Is anomaly part of hardware system, or something exogenous to the system like the outside environment or the testing equipment? Is there more noise than expected? Has testing equipment been checked?
  • Consider performing a sensitivity analysis?
    • How does model depend on various parameters?
    • Which responses are most strongly dependent on which parameters?
    • Which parameters might affect only steady state, or which are key to the dynamic response?
    • Which cause large changes in model output when tuned?
    • Which parameters have largest uncertainty or smallest confidence?

Return to flowchart