Estimation of net-impedance passively by perturbations from loads, with application to network fault detection by smart meters

back to project index

Project abstract

Smart meters have greatly increased the visibility at the extremities of the grid, giving more knowledge about faults and loading. Many applications of smart meter data have been proposed, and some implemented. One interest of the utilities is the detection of network faults, to allow rapid correction.

A broken phase conductor is relatively easily detected as a partial outage. A broken neutral conductor in a Swedish low-voltage (LV) network results in poor control of potentials and of L-N voltages: it is hazardous regarding shock and fire, and also has high chance causing of equipment damage that is expensive for customers and thus for the network operator.

Existing broken-neutral detection methods, already used with smart meters, look at opposite voltage-violations on different phases – one or two high, and the other(s) low. These methods are simple, but require good parameter settings and can even so risk confusing strongly unbalanced load and/or generation conditions with network faults. The challenge is that one wants to have very low chance of missing or delaying response to a real case (false negative), but the rate of false positives must be very low to avoid cost and loss of confidence, since there are many customers and relatively rare actual faults. Existing methods will not detect, for example, cases where a broken conductor’s current has some parallel path (ground, pipework) and/or the current trying to flow in the neutral is moderate. However, it would be good to detect and correct faults even when they start with these more benign situations.

A clearly more accurate principle, not currently implemented in Smart Meter systems, is to look at sensitivity of voltage changes to current changes - or, as another way of describing it, at passively estimating the network impedance. For a multi-wire system with different types of load (line-line, line-neutral etc) the estimation can even be done for each conductor of the supply system. For single-phase two-wire connection just the total loop can be estimated. In our group's earlier work this method has already been seen to work well for cases with a clear change caused by significant changes of load. However, lots of other things happen in power systems to cause noisy measurements, which need to be considered in order to make optimal choices in the algorithm. For example, the source voltage can change due to tap-changers and other loads, on the supply side of the system. Loads have starting transients, and/or highly non-sinusoidal waveforms. The changes in voltage are typically small for normal load changes, which raises the further question of measurement quality and how best to compare the voltage before and after a change.

Studies of actual data and simulations will determine the best combinations and thresholds. The current short project will contribute to this. We have months of real data from several LV networks, sampled at 16 kHz, which can be used to produce any Smart-Meter quantity that we want to test (e.g. fundamental voltage over 6 cycles, or third-harmonic current, etc).

Summary of work

-

Event log

-

Project reference-group

Edel Wallin,  Vattenfall
Nathaniel Taylor,  KTH

Publications by this researcher

See alternatively the researcher's full DiVA list of publications, with options for sorting.
Publications in journals and conferences usually will not show until a while after they are published.

An experimental assessment of the intermittent heating and cooling operation effect on power transformer insulation
Daniil Danylov,   Yanick Patrick Frei,   Leonardo Colombo,   Raimundo Montalba Mesa,   Zhanzhan Liu,   Patrick Janus,   Kateryna Morozovska.

Publication list last updated from DiVA on 2024-08-22 22:58.

Page started: 2021-09-01
Last generated: 2024-08-22