Some Considerations on IoT for Power System Operations

Some Considerations on IoT  for Power System Operations

Nokhum Markushevich

The Internet of Things (IoT) concept (see, e.g., [1]–[3]) that incorporates cloud analytics and data storage is considered a promising approach for future Information support of power system operations (see, e.g., [4]-[7]). Under “things,” we consider both physical and cyber objects [3].

 The list of data required for the analytics of distribution and transmission operations is extensive [8] – [13].  If any necessary items are omitted from the list, the analysis may be faulty, or even impossible.

As suggested in [7], the “thing” should

1. Announce itself

2. Describe itself

3. Provision itself

4. Commission itself

As a result, the user of this information (e.g., the utility) “knows what it is, where it is, and what capabilities it has”[7].  This information is supposed to be used for the” descriptive, predictive, and prescriptive” analytics [7].

To support the predictive and prescriptive analytics, the “describe itself” component should include information needed for “what-if” studies. Some of this information may be provided in near real time by the “thing” itself, e.g., the settings of the Remedial Action Schemes (RAS) and other controllers and local Energy Management Systems (EMS), assuming communication capabilities. Alternatively, this information can be obtained from engineering databases as conditionally constant data, provided that it does not change frequently [9]. 

However, in Active Distribution Networks, for instance, the settings of some RAS may need to become adaptive to the changing power balances due to near real-time changes in the operating conditions of Distributed Energy Resources (DER) and microgrids [8] – [10].  In these cases, the adapted setups of these objects should be obtained from several “what-if” studies. Then, when a particular situation from the “what-if” conditions appears, the relevant EMS/DMS applications use the appropriate settings adapted to this situation.

There is a significant amount of other data that is not conditionally constant and is not an attribute of the things that can be provided by sensors (or their equivalents in the case of cyber objects). Many of the attributes needed for “what-if” analyses should be derived using analytic tools across different layers of the information support environment, which can also correspond to different layers of cloud computing, for various levels of the power systems [11]-[16].

The following are examples of such attributes:

• Load-to-Voltage dependencies (LTV-factors)

• Energy-to-Voltage dependencies (CVR-factors)

• Load-to-frequency dependencies

• Generation-to-frequency dependencies

• Generation-to-voltage dependencies (mostly for Distributed Energy Resources

  For Active Distributing Networks, the LTV and CVR factors cannot be determined by performing selective field tests at the heads of distribution systems, as it was done for passive distribution networks [17]-[19], due to the changing contribution of the Distributed Energy Resources [20]-[21]. To determine the load/generation-to-frequency dependencies, some near-real-time and conditionally constant data from individual loads, generators, and Microgrids should be available. [8] – [10], [16]

• DER Watt/var operational capability curves that depend on the real power contributions, local voltages, activated functionality, and local and remote power grid constraints.

• Near real-time dispatchable load/generation in distribution

  This information can be derived from sets of power-flow-based studies that become cyber-physical systems, as well as near-real-time data and statistics collected by smart building IoT systems, DER controllers, and customer EMS. [13]-[15], [22]-[30].

• Dependencies of attributes of individual and aggregated loads and DER models on external factors, such as

  • The environment

  • The real-time energy prices

  • The Demand Response signals

  • The requests of Transmission and/or Distribution System Operators (TSO/DSO

    This information can be derived from sets of power-flow based studies, inputs from external systems that become “cyber things”, near real-time data and statistics collected by smart building IoT systems, DER controllers, and customer EMS [13]-[14], [31]-32].

Almost all this information should be provided for the following power system levels:

• House (apartment)

• Building

• Distribution transformer

• Composite customer’s Point of Common Coupling (PCC)

• Microgrid PCC [16], [32]

• Transmission-to-Distribution demarcation bus [13]-[14]

Ultimately, the above components should be aggregated at the transmission-to-distribution demarcation bus, if it is considered the end node for transmission analytics and also an information source for distribution analytics. This aggregated model is a part of the more inclusive Transmission Bus Load Model (TBLM) [13]-[14]. The model components should include the near-real-time and the short-term look-ahead values.

The TBLM should also include the following components:

• Load forecast

• Load shifting capabilities from one bus to another

• Degrees of uncertainty for the model components [33]-[34].

• Some attributes of the transmission, generation, and market systems relevant to the distribution analytics

• Post-factum distribution of event logs [13].

 Envisioning the buildup of the TBLM through cloud infrastructure, the development and updates of the TBLM start from the primary information sources at the edge of the internet: IoT sensors, local system gateways, engineering and field personnel, and external system interfaces. Each item of this primary information is then transmitted to the appropriate level in the cloud infrastructure [35] for clustering by corresponding Data Management Systems (MS) based on relevant associations (e.g., smart homes connected to the same distribution transformer (DT), DTs and other power equipment connected to the same feeders, feeders to buses, etc.). Based on the conditionally constant and near real time data extracted from relevant Data MS, active models of different composite components of the Active Distribution Networks are developed by the corresponding model processors [13] (e.g., the near real time attributes of DT real and reactive loads including their dependencies on different factors, or the near real time attributes of microgrids including the current setup of their RAS [16]). This computing can be performed either in a common cloud of the distribution system or on the edges of the internet (e.g., in a microgrid EMS or a smart home gateway). In the latter case, the results of the computing, the models, are transmitted to the common distribution system cloud.  These models are used in distributed cloud computing to run advanced DMS applications in near real-time. The same models that include dependencies of the attributes on external factors are used in a series of distribution modeling what-if studies under alternative conditions to develop aggregated models for the Transmission/Distribution demarcation buses (TBLMs). These TBLMs are then transmitted to the next level cloud, which includes the attributes of the transmission system and bulk generation, and serves the bulk power system by running EMS applications

Each level in the cloud infrastructure must be fully populated with the data necessary to run the analytics for the relevant level. Missing information should be replaced by adequate equivalents (e.g., the secondary circuit attributes [33], distribution transformer loading, etc.).

Figure 1 illustrates a sample of “bottom-up” information supports for the creation and updating of TBLMs, which starts from the Internet of physical and cyber things and ascends through several expanding clouds. As seen in the illustration, there may be multiple “narrow” clouds on lower levels submitting information to wider clouds of higher levels, up to the most general clouds of the bulk power system and interconnections of systems.

  Figure 1. Sample structure of information support for TBLMs based on IoT and cloud analytics

Let’s consider, for example, the development and updating of two of the TBLM attributes: the load-to-voltage (LTV) and energy-to-voltage (CVR) dependencies. As mentioned above, the field tests of the LTV and CVR factors performed for passive distribution networks do not accurately represent these load model attributes for active distribution networks. Therefore, a bottom-up methodology was introduced in [33]-[34]. It starts with smart home systems that have IoT capabilities. The methodology for determining the composite LTV-factors is based on LTV-factors defined earlier for different types of equipment and appliances, as well as knowledge of the near real-time composition and demand of the equipment for the relevant premises. The LTV factors for different types of equipment were determined through a series of laboratory tests (see, e.g., [36]). More such studies are needed to have a comprehensive list of individual LTV factors for the old and new types of equipment. If these factors are known and the real-time composition and loads of equipment is determined by the smart home system (see e.g. [37]-[38], the composite LTV factor for the home can be defined as a weighted sum of these factors:

LTVj,t = (ΣLTVi x Li,t)/ ΣLi,t                      (1)

Where

LTVj,t – composite LTV-factor for premises j at time t

LTVi  – individual LTV-factor for the i type of equipment

Li,t – Load of equipment i at time t.

The composition of home appliances and their loading change over time. The changes vary from one home to another. That is why the composite LTV factors differ for various times of day and different homes and should be determined in a near-real-time manner.

The following is an illustrative example of a method for determining the LTV factors and CVR factors based on IoT and cloud analytics.

Figure 2 illustrates the difference in load shapes of “similar” customers chosen for our example. The customers are considered “similar” because they are all residential customers with the same set of appliances (electric heating and cooking) and are connected to the same distribution transformer (DT). The difference lies in the lifestyle of residents: some are working professionals, while others are retirees. Therefore, the times of heating and cooking, as well as other activities, are different (see Figure 3).

 Figure 2. Example load shapes of five “similar” customers

Figure 3. Compositions of appliances for two “similar” customers

It is assumed here that a smart home IoT system determines the composition and loading of the appliances, and the LTV factors for these appliances are known and stored either in the analytical device of the smart home system or in a low-level cloud. Then, the composite LTV factors for the home can be calculated based on (1) [33], [34].

When the total load and LTV-factor of every home connected to the same DT are factors are available for the corresponding cloud analytics, the current LTV-factor for the DT can be calculated similarly to (1) as follows:

LTVDT,t = (ΣLTVj,t x Lj,t)/ ΣLj,t          (2)

The LTV factors for individual appliances used in our example are presented in Table 1. Some of these LTV factors were derived from [36] by linearizing the ZIP models, and others were just assumed for this illustrative example.

Table 1. Example LTV factors of appliances

Figure 4 presents the LTV-factors for the five homes of our example and the LTV-factors for their feeding DT. As shown in the figure, the DT LTV factors of the composite residential load are significantly different from those of individual residences and also vary significantly throughout the day.

It is safe to assume that the LTV factors of other “similar” DT loads are also different from each other.

Figure 4. LTV-factors for the loads of five customers and the load of the DT

Figure 5 illustrates the difference in daily DT LTV-factors for summer and winter and loads with air conditioners.

Figure 5. LTV-factors for DT load at winter and summer (with and without air conditioners)

As follows from the above, the LTV factors change in near real-time and may be significantly different for different DTs and at different times. The Internet of Things (IoT) in smart homes and buildings, linked to cloud analytics, can provide the necessary information at the right time.

Further, when the DT loads and their  LTV factors are known for the corresponding distribution feeders together with the models of other relevant components of the distribution circuits connected to the same substation bus the LTV-factors for the TBLM can be determined in a higher level of the cloud by running the distribution operation model under different bus voltages [13] - [15], [21].

It must be noted that when there is a significant contribution from electric resistive (especially heating) load, the LTV-factor determined in the above way can be used only for very short-term analytics, like transient studies. In the case of peak load reduction for a period of several hours or the case of energy conservation (CVR-factors), the increased ON time of the resistive load may result in increased energy consumption under lower voltage and an increased coincidence factor of the DT load, as well as for the upstream circuits.

Specific information about the CVR factors and the long-term LTV factors can also be obtained by utilizing near-real-time and historical information stored by the IoT systems.

The methodology for determining the individual and composite CVR factors differs from that for determining the LTV factors, especially when there is a large component of thermostat-controlled load [39]. There are no constant CVR factors for many types of equipment due to adjustments in voltage over time.

If the smart house system can determine and store the times when the heater is ON and its demand during these times,  and the ambient temperature and voltages for these times can be made available either from the same system, or from another source, then the energy consumption by the heater under different ambient temperatures and voltages can be determined under a given duration of observation time. Based on such information, the kWh dependencies on voltage for different temperatures can be assessed. Learning algorithms can be used for this task.

An illustration of such dependencies for a hypothetical electric heater is presented in Figure 6. The CVR-factors here are the proportion regression coefficients of the linearized dependency (-2.55 in this example). The negative value of the CVR factor means that the energy consumption increases with the reduction of voltage. As shown in the figure, the kWh dependencies on voltage, expressed in p.u. do not differ significantly for different ambient temperatures (at least, for normally sized heaters).

To derive such dependencies, the smart home system should capture measurements for multiple voltages and various ambient temperatures. The more history is collected, the more points for interpolation/extrapolation become available. However, acceptable accuracy of the dependencies can be obtained even with a small number of points.  Figure 7 illustrates the comparison of the sample dependencies derived from three different sets of three voltages selected from different areas of the ±5% range of voltages. As seen in the figure, the CVR factors in all three-point cases are close to the CVR factor derived from ten points.

Figure 6. Heater’s kWh dependencies on voltage

Figure 7. Comparison of kWh dependencies on voltage determined based on different sets of three voltages and on ten voltages

The CVR factors for individual appliances selected for our hypothetical example are presented in Table 2

Table 2. Example CVR factors of appliances

Based on these CVR factors, the DT factors were derived (see Figure 8). As shown in the figure, the CVR factors are negative for most of the day in winter. Even in the summer, negative CVR factors can occur during periods of high electric cooking loads.

When electric loads are ON longer, the probability of several being ON at the same time increases.

Smart home systems can determine the probabilities of operation for individual appliances within a given time interval. Based on these probabilities, the probabilities of overlapping of different numbers of appliances can be determined. Figure 9 illustrates a simplified case of the probabilities of overlapping of different numbers of heaters out of ten heaters for different voltages. It also shows that undersized heaters have higher probabilities of overlapping.

Figure 8.  CVR factors for DT load at winter and summer (with and without air conditioners)

Figure 9. Probability of the number of heaters ON at the same time (out of ten)

As seen in Figure 9, the average number of the sample heaters that are ON at the same time is 5 for the nominal voltage, but is 6.5 for -5 % of voltage deviation and 4 for +5% of voltage deviation from the nominal.

Applying 0.95% of voltage to 6.5 heaters reduces the load of these heaters by ~10% (an equivalent of 0.65 heaters). However, 1.5 heaters more under this voltage increases the load by 0.85 equivalent heaters.

This means that after a certain time interval of reduced voltage, both the energy consumption and the demand of the electric load increase, and the LTV factors become negative as well (lower voltage – higher load). 

The types and composition of loads at different locations and times change in near real time. Such distinctive information can only be obtained from the customer sites. If this information is transmitted to higher layers of the information support system, it can be integrated into higher-level models, such as the TBLM.  Based on such models, both the distribution and transmission management systems can provide time-consistent situational assessments and controls, informing the model, for example, that at certain times and/or locations, voltage reduction may not save energy and may not reduce demand.

The challenge for such an information support system is that, for the foreseeable future, there will not be enough smart homes to provide sufficient data for the energy management systems. However, the mass installation of smart monitoring devices, which employ learning algorithms, as separate units or as embedded in smart meters, and are capable of recognizing major appliances will provide necessary information for determining the LTV and CVR factors if specific appliance factors are known. 

Conclusions.

1. In Active Distribution Networks, many characteristics of distribution operations cannot be determined based on a top-down approach (when high-level information is considered to represent local information).

2. A smart home system with IoT and AI abilities can provide the basic information for a bottom-up approach for determining a number of critical attributes of distribution system operations.

3. Supplementing the information collected by smart sensors or AI-enabled monitoring devices with conditionally constant LTV and CVR factors of major appliances can provide time and location-specific LTV and CVR factors for individual customers.

4. Submitting the attributes of individual customer loads to a higher level of cloud analytics provides the ability to integrate these attributes with attributes of other customer loads and other components of the distribution system (e.g., DERs, microgrids, etc.) up to the Transmission Bus Load Model.

5. Until enough smart homes are integrated with the Internet and approximate methodologies for determining the required attributes of distribution operations are used, a greater uncertainty in distribution operation management should be accounted for.

References.

1. Kevin Ashton, That 'Internet of Things' Thing. Available: https://www.rfidjournal.com/articles/view?4986   

2. A Non-Geek’s A-to-Z Guide to the Internet of Things, SAS Institute Inc. Available: https://www.sas.com/content/dam/SAS/en_us/doc/whitepaper1/non-geek-a-to-z-guide-to-internet-of-things-108846.pdf

3. Eve M. Schooler, David Zage, Jeff Sedayao, Hassnaa Moustafa, Andrew Brown, and Moreno Ambrosin,  An Architectural Vision for a Data-Centric IoT: Rethinking Things, Trust and Clouds. Available: https://www.researchgate.net/publication/317438671_An_Architectural_Vision_for_a_Data-Centric_IoT_Rethinking_Things_Trust_and_Clouds

4. Rob Young, John McCue, Christian Grant, The power is on: How IoT technology is driving energy innovation. Available: https://www2.deloitte.com/us/en/insights/focus/internet-of-things/iot-in-electric-power-industry.html

5. Internet of Things (IOT) and Its Applications in Electrical Power Industry. Available: https://www.electricaltechnology.org/2016/07/internet-of-things-iot-and-its-applications-in-electrical-power-industry.html

6. Arun Ramamurthy and Pramod Jain, The Internet of Things in the Power Sector Opportunities in Asia and the Pacific. Available: https://www.adb.org/sites/default/files/publication/350011/sdwp-48.pdf

7. Stuart McCafferty, Eamonn McCormick, and David Forfia, EnergyIoT Articles 1-6. Available: https://www.energycentral.com/topics/tags/call-action-series

8. Nokhum Markushevich, What will the Microgrids and EPS Talk about? Part 1. Available: http://www.energycentral.com/gridtandd/gridoperations/articles/2858

9. Nokhum Markushevich, Automatic Load Shedding in Active Distribution Networks. Available: http://www.energycentral.com/c/iu/automatic-load-shedding-active-distribution-networks

10. Coordination of EPS and Microgrid Load Shedding Schemes (Based on Under Frequency Load Shedding), Use Case Description. Available: https://smartgrid.epri.com/UseCases/Use%20Case%20IA-1.pdf

11. Functional Requirements for Advanced Distribution Automation with DER (ADA-DER). Available: https://smartgrid.epri.com/UseCases/AdvancedDistributionAutomationwithDERFunction.pdf

12. Distribution Grid Management (Advanced Distribution Automation) Functions. Use Case Description. Available:  https://collaborate.nist.gov/twiki-sggrid/bin/view/SmartGrid/PAP08DistrObjMultispeak

13. Development of Transmission Bus Load Model (TBLM), Available: http://collaborate.nist.gov/twiki-sggrid/pub/SmartGrid/TnD/TBLMUseCase_V14-03-13-13-posted.pdf

14. Nokhum Markushevich, Transmission Bus Load Model for Smart Distribution and Transmission Grids. Available: https://www.energycentral.com/c/iu/transmission-bus-load-model-smart-distribution-and-transmission-grids

15. What will the Microgrids and EPS Talk about? Part 2. Available: http://www.energycentral.com/gridtandd/gridoperations/articles/2864

16. Use cases for Microgrid. Available:  http://smartgrid.epri.com/Repository/Repository.aspx/

17. Alf Dwyer, Ron Nielsen, Joerg Stangl, Nokhum S. Markushevich, Load to Voltage Dependency Tests at B.C. Hydro,; IEEE/PES 1994 Summer Meeting, July 1994

18. Nokhum S. Markushevich, R.E. Nielsen, A.K. Nakamura, J.M. Hall, R.L. Nuelk, Impact Of Automated Voltage/Var Control In Distribution On Power System Operations,; DA/DSM Conference January 1996, Tampa, Florida

19. Nokhum Markushevich; Aleksandr Berman and Ron Nielsen, Methodologies for Assessment of Actual Field Results of Distribution Voltage and Var Optimization, presented at IEEE PES 2012 T and D

20. Nokhum Markushevich, New Aspects of IVVO in Active Distribution Networks, Presented at IEEE PES 2012 T and D

21. Update aggregated at PCC real and reactive load-to-voltage dependencies, Use Case Description. Available: https://smartgrid.epri.com/UseCases/Use%20Case%20IA-3.pdf

22. Nokhum Markushevich, Operations of Smart Inverters in Active Distribution Networks. Available: https://www.scribd.com/document/376627135/Operations-of-Smart-Inverters-in-Active-Distribution-Networks

23. Updates of capability curves of the microgrid’s DERs, Use Case Description. Available:  https://smartgrid.epri.com/UseCases/Use%20Case%20IA-4.pdf

24. Nokhum Markushevich, Reflections on Volt/Var Control by Multiple Smart Inverters of DERs. Available:  https://www.energycentral.com/c/cp/reflections-topic-voltvar-control-smart-inverters-ders      

25. Nokhum Markushevich, Reflections on the Topic of Volt/Var Control by Smart Inverters of DERs. Available: https://www.energycentral.com/c/cp/reflections-topic-voltvar-control-smart-inverters-ders

26. Nokhum Markushevich, Comparison of the effectiveness of var priority mode for reducing and increasing voltage. Available: https://www.energycentral.com/c/iu/comparison-effectiveness-var-priority-mode-reducing-and-increasing-voltage  

27. Nokhum Markushevich, Var-priority Mode of DER Volt/var Control Function. Available: https://www.energycentral.com/c/ee/var-priority-mode-der-voltvar-control-function

28. Vars versus Watts from Distributed Energy Resources. Available: https://www.scribd.com/document/376740872/Vars-versus-Watts-from-Distributed-Energy-Resources  

29. Nokhum Markushevich, Dispatchable Reactive Load in Active Distribution Networks. Available: https://www.scribd.com/document/376759501/Dispatchable-Reactive-Load-in-Active-Distribution-Networks

30.  Updating information on microgrid dispatchable load, Use Case Description. Available: https://smartgrid.epri.com/UseCases/Use%20Case%20IA-5.pdf

31. Nokhum Markushevich, Information Exchange between Advanced Microgrids and Electric Power Systems. Available: https://www.scribd.com/document/376567099/Information-Exchange-between-Advanced-Microgrids-and-Electric-Power-Systems

32. Updating dependencies of the microgrid operational model on external conditions. Available: https://smartgrid.epri.com/UseCases/Use%20Case%20IA-7.pdf

33. Nokhum Markushevich. Uncertainty of Voltage Control in Active Distribution Networks. Available:  https://www.scribd.com/document/376554187/Uncertainty-of-Voltage-Control-in-Active-Distribution-Networks

34. Nokhum Markushevich, Accuracy of Voltage Optimization in Active Distribution Networks. Available:  https://www.energycentral.com/c/iu/accuracy-voltage-optimization-active-distribution-networks

35. Cloud, Fog and Edge Computing – What’s the Difference? Available: https://www.winsystems.com/cloud-fog-and-edge-computing-whats-the-difference/?keyword_session_id=vt~adwords|kt~|mt~b|ta~376277075726&_vsrefdom=wordstream&gclid=EAIaIQobChMIxqW3p62S5AIVDcRkCh3nygr1EAAYAyAAEgKNFPD_BwE

36. Abdullah Bokhari,  Ali Alkan, Rasim Dogan, Marc Diaz-Aguiló, Francisco deLeón, , Dariusz Czarkowski, Zivan Zabar, Leo Birenbaum, , Anthony Noel, and ReskEbrahem Uosef, Experimental Determination of the ZIP Coefficients for Modern Residential, Commercial, and Industrial Loads. IEEE Transactions on Power Delivery, Vol.29, No.3, June, 2014. Available: http://engineering.nyu.edu/power/sites/engineering.nyu.edu.power/files/uploads/Experimental%20Determination%20of%20the%20ZIP%20Coefficients%20for%20Modern%20Residential%2C%20Commercial%2C%20and%20Industrial%20Loads.pdf

37. How the Sense Energy Monitor Works. Available: https://blog.sense.com/how-the-sense-home-energy-monitor-works/

38. Smart Homes Curb Energy Use. Available: https://energycurb.com/blog/smart-homes-curb-energy-use/

39. Nokhum Markushevich,  Analysis of electric heating load dependency on voltage. Available https://www.energycentral.com/c/iu/analysis-electric-heating-load-dependency-voltage