Topology identification in distribution network with limited measurements

Patent metadata

Field	Value
Publication number	US-9287713-B2
Application number	US-201213558711-A
Country	US
Kind code	B2
Filing date	Jul 26, 2012
Priority date	Aug 4, 2011
Publication date	Mar 15, 2016
Grant date	Mar 15, 2016

Abstract

A statistical technique is used to estimate the status of switching devices (such as circuit breakers, isolator switches and fuses) in distribution networks, using scares (i.e., limited or non-redundant) measurements. Using expected values of power consumption, and their variance, the confidence level of identifying the correct topology, or the current status of switching devices, is calculated using any given configuration of real time measurements. Different topologies are then compared in order to select the most likely topology at the prevailing time. The measurements are assumed as normally distributed random variables, and the maximum likelihood principle or a support vector machine is applied.

First claim

What is claimed is:

1. A method for restoring continuity in a distribution network including a plurality of interconnected buses and switch devices, the method comprising:
deriving a probability model for power injections at each of the plurality of interconnected buses, the power injections comprising power consumption and power generation at individual buses, the probability model being derived from historical measurements of power injections;
deriving a plurality of possible network topologies from a plurality of possible statuses of the switch devices as either open or closed, wherein pairs of buses connected by closed switches in the possible network topologies are treated as connected by zero impedance lines;
receiving real time sensor measurements of electrical quantities in the distribution network;
selecting, from the plurality of network topologies, a network topology most likely to produce the real time sensor measurements, the selecting being based on the probability model;
based on the network topology most likely to produce the real time sensor measurements, making an identification of a normally closed switch device of the switch devices having an open status; and
restoring continuity in the distribution network based on the identification of the normally closed switch device having an open status.

2. The method as in claim 1 , wherein the selecting the network topology most likely to produce the real time sensor measurements further comprises:
constructing a plurality of optimization problems, each optimization problem corresponding to one of the plurality of network topologies, each optimization problem having:
a state variable representing voltages of the plurality of interconnected buses in the network;
a cost function representing an unlikelihood of a set of power injections for the buses computed from the state variable, given the probability model for the power injections; and
constraints requiring that the real time sensor measurements of electrical quantities match corresponding electrical quantities computed from the state variable; and
selecting a network topology corresponding to an optimization problem having a minimum cost computed from the cost function.

3. The method as in claim 1 , further comprising:
generating a set of likely power injections using the probability model for power injections;
for each of the network topologies, computing expected sensor measurements of the electrical quantities using the set of likely power injections; and
using the expected sensor measurements, deriving a probability model for sensor measurements of the electrical quantities for each of the network topologies; and
wherein selecting a network topology most likely to produce the real time sensor measurements is based on the probability model for sensor measurements of the electrical quantities for each of the network topologies.

4. The method as in claim 1 , further comprising:
generating a set of likely power injections using the probability model for power injections;
for each of the network topologies, computing expected sensor measurements of the electrical quantities using the set of likely power injections; and
using the expected sensor measurements, constructing a classifier to classify a set of sensor measurements of the electrical quantities into a class corresponding to a network topology; and
wherein selecting a network topology most likely to produce the real time sensor measurements comprises classifying the real time sensor measurements using the classifier.

5. The method as in claim 4 , wherein the classifier is a support vector machine.

6. The method as in claim 1 , further comprising:
computing, for each of the plurality of network topologies, an approximate linear transformation from the power injections to expected sensor measurements; and
projecting, using the approximate linear transformations for each of the plurality of network topologies, the probability model for the power injections to probability models on the expected sensor measurements; and
wherein selecting a network topology most likely to produce the real time sensor measurements comprises selecting the network topology using the probability models on the expected sensor measurements.

7. The method as in claim 6 , further comprising:
constructing a classifier to distinguish among the probability models on the expected sensor measurements; and
wherein selecting a network topology most likely to produce the real time sensor measurements comprises using the classifier to identify a probability model to which the real time sensor measurements most likely belong.

8. The method as in claim 7 , wherein the classifier is a support vector machine.

9. The method as in claim 1 , wherein the real time sensor measurements of electrical quantities in the distribution network comprise electrical quantities selected from a group consisting of power flow, voltage and current.

10. A method as in claim 1 , wherein the historical measurements of power injections are labeled with at least one time indicator selected from a group consisting of times of day, days of week and seasons of year, and the probability model for power injections includes the time indicator.

11. A method as in claim 1 , further comprising:
placing sensors in the distribution network to measure the real time sensor measurements, according to a sensor placement configuration determined by, for each of a plurality of proposed sensor placement configurations, performing the following:
generating simulated sensor measurements for each of the plurality of network topologies, by sampling the probability model for power injections at each of the plurality of interconnected buses;
selecting a network topology most likely to produce the simulated sensor measurements, the selecting being based on the probability model; and
determining a reliability of the proposed sensor placement configuration by comparing the network topology most likely to produce the simulated sensor measurements with the network topology used in generating that simulated sensor measurement;
identifying a sensor placement configuration having a highest reliability of the proposed sensor placement configurations.

12. A non-transitory computer-usable medium having computer readable instructions stored thereon for execution by the processor to perform a method for restoring continuity in a distribution network including a plurality of interconnected buses and switch devices, the method comprising:
deriving a probability model for power injections at each of the plurality of interconnected buses, the power injections comprising power consumption and power generation at individual buses, the probability model being derived from historical measurements of power injections;
deriving a plurality of possible network topologies from a plurality of possible statuses of the switch devices as either open or closed, wherein pairs of buses connected by closed switches in the possible network topologies are treated as connected by zero impedance lines;
receiving real time sensor measurements of electrical quantities in the distribution network;
selecting, from the plurality of network topologies, a network topology most likely to produce the real time sensor measurements, the selecting being based on the probability model;
based on the network topology most likely to produce the real time sensor measurements, making an identification of a normally closed switch device of the switch devices having an open status; and
restoring continuity in the distribution network based on the identification of the normally closed switch device having an ope

Assignees

• Sharon Yoav
• Annaswamy Anuradha
• Legbedji Motto Alexis
• Chakraborty Amit
• Siemens Ag
• Massachusetts Inst Technology

Inventors

• Sharon Yoav
• Annaswamy Anuradha
• Legbedji Motto Alexis
• Chakraborty Amit

Classifications

• H02J3/003

  Load forecast, e.g. methods or systems for forecasting future load demand · CPC title

• H02J13/1337

  involving the use of Internet protocols · CPC title

• H02J13/333

  the equipment forming part of substations · CPC title

• G06F2111/08

  Probabilistic or stochastic CAD · CPC title

• G06F30/18

  Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling (circuit design at the physical level G06F30/39; network planning tools for wireless communication networks H04W16/18) · CPC title