How to Create an Accurate Network Model and Dynamic State Data for an Advanced Distribution Management System (ADMS)
By Vahraz Zamani and Terry Nielsen
Why is a Distribution System State Estimation (DSSE) is Necessary in an Advanced Distribution Management System? There is some debate in the industry as to whether a Distribution System State Estimator (DSSE) is a critical ADMS function. Without the DSSE, an ADMS uses SCADA measurements, load allocation, and a power flow to determine system state. The premise of state estimation is that by using redundant measurements, it is possible to identify the likely error in each measurement and find the system state that is the best fit. Algorithmically, DSSE estimates the bus voltage using power flow equations derived from the network model, represented in an admittance matrix. DSSE takes several measurements and solves a weighted least-squares problem using an iterative approach. If done properly, the ADMS will have a more accurate system state solution to use in all the advanced applications such as Voltage Optimization.
State estimation also provides real-time bad data detection . Redundant measurements also provide the opportunity to implement a data analytics approach to review clean up measurements and model data. In Transmission Systems, State Estimation has been used for decades where the results of State Estimation are used by the full suite of advanced applications in an Energy Management System (EMS). It has also been used in other industries .
If Distribution redundancy in SCADA measurements doesn’t exist, so state estimation doesn’t make sense, right? Unlike Transmission, Distribution systems typically won’t have the necessary redundancy in traditional SCADA measurements. However, redundancy can be created by including pseudo measurements from multiple sources such as load forecasting, DER forecasts, and AMI measurements [2, 3].
So why go to all this trouble? When performing closed-loop control, such as Volt/VAR Optimization (VVO), it is important that there is confidence in the actions taken are based upon good data. Without DSSE, one bad measurement could cause a closed-loop algorithm to do something “stupid.” The goal of an ADMS is to be able to run optimization algorithms that perform closed-loop control using a DSSE running on a validated model and measurements. This is the top rung of the ADMS data integrity ladder.
Step 0: Connectivity for Outage Management At the base of the ladder is the Outage Management System (OMS). An OMS utilizes the connectivity model to identify the most likely device to have opened using the locations of customer calls and AMI power off notifications. Cleaning up the connectivity of a three-phase model that is suitable for outage management is essential for running other applications. The outage prediction process is relatively forgiving of connectivity errors since the data error cost is an inaccurate prediction that can be identified in the field. ADMS applications require the GIS connectivity model plus the operational aspects of a “real-time” network model. Most utilities rely on GIS to create the connectivity model. The ADMS model is considered a dynamic operational model often referred to “as operated” model. This operational model represents the current state of the network.
Step 1: ADMS with Power Flow The next step up the ladder is to add the attributes necessary to support an accurate ADMS power flow. This includes data required to build the impedance matrix. In the ADMS, in addition to the impedance matrix, the power flow solution requires at a minimum, SCADA measurements at all “source” nodes and the estimated load at all load connection points. Power flow solutions from the ADMS and the planning engineering solution are compared at this level, and all inconsistencies need to be investigated and addressed. This step is critical because the DSSE calculated measurement error is based upon how well measurements fit a solution of the flows that occur given the full network connectivity and all the branch impedances. Incorrect branch impedances will distort the state estimator solution producing inaccurate results.
Figure 1 – The Ladder of ADMS Data Integrity
Step 2: DSSE with SCADA measurements and load profiles To make distribution systems observable by DSSE, historical load data from AMI can be utilized to create pseudo measurements that supplement the SCADA measurements. These meter level load profiles can be disaggregated into load and generation components. Furthermore, they need to have unique profiles by day of the year, time of day, temperature and solar irradiance. Large spot loads should be indepdently modeled. And finally, any Distributed Energy Resources (DERs) such as solar PV generation and battery storage need to be in the model.
Step 3: Audited Model for DSSE Finding model errors while deploying the state estimator is time-consuming and expensive. It is important to make attempts at cleaning up the data prior to using it with the DSSE. Multiple approaches are often taken including:
- Use an analytics approach to AMI data and SCADA to automatically validate GIS connectivity and GIS model phasing, and meter-to-transformer mapping, automated phase identification and meter-to-transformer mapping .
- Integrate DER generation profile databases to ADMS with various resolutions and quality .
- Leverage routine work, such as field inspections, to continuously improve data.
Step 4: ADMS/DSSE with AMI and/or PMU and/or line sensor measurement streams. Utilities are placing Power Quality (PQ) meters/ recorders and Phasor Measurement Units (PMU)  to measure the impact of DERs on their networks and quality of their services. Deployment of more meters/sensors on the distribution system improves the quality of DSSE results and chances to detect and identify the unacceptable measurements and statuses .
Measurement Placement To improve the quality of the DSSE solution, more measurements are needed. The issue is where to place the measurements. Running studies show that measurements should be placed at critical locations determined based on the goal of ADMS applications, such as CVR , VVO, or FLISR . For ADMS advanced applications, the measurement placement matters. Proper measurement placement includes not only determining where to deploy additional line sensors and PMUs but also which AMI meters to classify as bellwether meters that report with a higher periodicity.
The final key to the puzzle is the use of confidence factors for each measurement:
- high confidence in measurements from precision substation transducers
- medium confidence in the measurements from less precise feeder sensors
- lower confidence in AMI measurements from the secondary with lower scan rates
- and the lowest confidence in load/DER estimates
Step 5: Tuning the DSSE The first step in tuning the state estimator is to initialize all the measurements used in the DSSE from an ADMS power flow solution and compare the results. Synthetic errors are then introduced to further test the solutions.DSSE solutions can also be compared against planning power flow. Caution must be taken because often minor differences in modeling approach, algorithms used, and other factors. At each step, data errors may be identified and tuning parameters/confidence factors may need to be adjusted. Finally, after all of the above, DSSE results using real SCADA measurements can be compared against field measurements. Following all of these steps ensure that the distribution operator will be able to trust the DSSE solution.
Step 6: ADMS Optimization functions running closed-loop control Finally, after the DSSE is tuned, the optimization functions such as VVO can be incrementally rolled out. The roll-out strategy should be selected based upon DSSE tuning, data availability, measurement readiness, and business.
This article focused on the importance of systematically cleaning up the data model used by the DSSE in an ADMS to support the roll-out of advanced optimization functions. It defines a clear approach to improving ADMS data integrity in a step by step process based upon the authors' experience deploying ADMS solutions.
 A. Abur and A. G. Exposito. “Power System State Estimation: Theory and Implementation.” CRC Press, 2004.
 V. Zamani Farahani, “State Estimation for Volt/VAR Control on Active Power Distribution Systems,” Ph.D. Dissertation, NC State University, 2014
 L. Schenato, G. Barchi, D. Macii, R. Arghandeh, K. Poolla, and A. V. Meier, “Bayesian linear state estimation using smart meters and pmus measurements in distribution grids.” In 2014 IEEE International Conference on Smart Grid Communications (SmartGridComm), Nov 2014, pp. 572–577.
 H. Mirzaee, V. Zamani, and F. Katiraei, "A Methodology for Primary Voltage Assessment on Highly PV Penetrated Distribution Feeders by Using SCADA and AMI," 2019 IEEE PES ISGT, Washington, DC, USA, 2019, pp. 1-5.
 B. Russell and J. Schoene, “Implementation DSSE into Utility Grid Operation,” IEEE PES-GM 2019.
 V. Zamani and M. E. Baran, "Meter Placement for Conservation Voltage Reduction in Distribution Systems," in IEEE Transactions on Power Systems, vol. 33, no. 2, pp. 2109-2116, March 2018.
Vahraz Zamani, Ph.D., is a Principal Engineering Consultant in GridBright, Inc. He is a power system engineer, specializing in distribution system planning and operation for more than 14 years. Vahraz received his Ph.D. from North Carolina State University on “State Estimation for Volt/VAR Control on Active Power Distribution Systems,” in 2014. As a member of IEEE, his interest topics are Distribution System State Estimation (DSSE), DER Integration, distribution system modeling, control, and operation.
Terry Nielsen (SMIEEE) is Executive Vice President of GridBright, Inc. He is an executive consultant and systems integrator specializing in power industry grid management systems. He has served over 50 electric utilities in 5 continents, improving their grid management and emergency response. He has served on the DistribuTech Advisory Board since 2001 and is the chair of the IEEE DSOP Committee’s Working Group on DMS. He is managing https://bettergrids.org/ to maintain and develop reliable power grid models and is an activity supported by the ARPA-E GRID DATA Program.