Skip to main content Skip to Footer


March 18, 2015
How to Use AMI Data to Improve Security of Electric Network Operations
By: Zitong Song

With the large deployment of advanced metering infrastructure (AMI) in smart grids, the exact load from end-user electricity consumers is available to system operators. This information enables operators to manage the electricity flow in a more flexible and reliable way within an electric network.

One operational strategy, called postponement, shortens time horizons and thus improves production accuracy. Postponement delays the manufacture of a finished product closer to product purchase. Retailers realize postponing the delivery of the final product to consumer destination, whereas for manufacturers, this postponing the final assembly of the product offers business benefits.

In response to this need, Accenture Technology Labs in Beijing developed an application--the network device overloading elimination solution--based on the power flow results calculated from AMI data, network topology and device parameters.

“Overloading” happens when a network device is continuously loaded beyond its rated capacity for longer than a certain predefined time such as 30 minutes. Under such a circumstance, the life span of the network device quickly diminishes; the device may even fail prematurely, which would jeopardize system security. Overloading can be caused by an abnormal amount of active power or reactive power in the network. When this “power factor” is greater than a threshold value, we call the overloading reason a P reason; otherwise, it is called a Q reason.

Utilities can use AMI data to measure the active power and reactive power flowing in the network. Based on these values, they can acknowledge, analyze and eliminate an overloading situation (see Figure 1).


Figure 1: Accenture network device overloading elimination solution strategy

From Figure 1, we can see that the two overloading reasons require different solutions. When the active power on a device is greater than its rated capacity (P reason), it is necessary to replace it with a larger capacity device. Alternatively in a real-time operation, the extra active power can be transferred to other parts of the network via switching actions. This is called network reconfiguration.

If the overload problem is due to Q reason, the reactive power is required to be compensated by adjusting some electric network compensators. To determine the value to be adjusted, utilities can use a classic optimization technique with the objective of minimizing the reactive power of the entire network.

There are many methods available to implement network reconfiguration. In the Labs, we invented a heuristic method, which is called a “combined depth-first and breadth-first method.” Compared with conventional methods, our method can provide a more balanced load transfer plan with less switching operation actions.

We have implemented a demo based on the solution strategy shown in Figure 1. A case study of this demo can be seen in Figure 2 and Figure 3. There are three lines, indicated by three colors, connecting to each other to compose one network. The overload happened on the middle line, which is indicated by a red block in Figure 2.


Figure 2: Case study (overload)

After applying our combined depth-first and breadth-first method, the overload is eliminated via transferring the load with yellow background to the other neighbor lines in the network, which is displayed in Figure 3.


Figure 3: Case study (load transfer)

To date, we have presented the demo to different clients, including State Grid of China and Hong Kong CLP, published two papers on the solution and filed one patent.

With a real-time electric power calculation every minute (aka state estimation), system operators will have better insight into the entire network and the ability to maintain security across the system. Greater redundancy in the available input data will result in a more precise state estimation. Our future work will focus on combining AMI data with other data acquired by system operators and SCADA data in order to maximize the outcomes of network state estimation.

More blogs on this topic