Internal/HEM Meetings

From ISLAB/CAISR

Phone Meeting with Eltel and HEM on 2015 Apr 22

The following document had been sent to Eltel and discussed during the meeting:

In order to collect more information about HEM’s distribution network especially from delivery points, we would like to change load profile of the meters without affecting the communication network and available database. This information would allow us to receive more data in shorter intervals about the power quality at the end point of the distribution network. In this case, we have some discussion topics and would be happy if you can help us.

1. We would like to change load profile, time intervals, and alarm threshold in meters and collect more data. At the moment, if we change a meter’s load profile from NES system, it seems that the meter is re-configured correctly but from the “Powel system ELIN” we have problem in receiving data from additional channels. We do not know if it is the problem in NES or ELIN or the way we are re-configuring the meters.

2. How short can we select the intervals without affecting the communication network (because of the increase of data which needed to be transmitted). The same scenario for the load profile data, how much data we can receive? All the 16 channels? How this would affect the long term storage?

3. We would like to change the intervals without affecting the billing system (V-store K+). For example, we would like to receive data every 15 minutes from smart meters but we want to have the billing system works based on daily measurements, according to the customers’ contract. What would be the best way to do that?

4. Can we receive smart meters data directly from NES without affecting other systems?

5. In ELIN system, when we call for the alarm data, it does not show the information of all the meters. For example, in a very recent failure, among more than 6000 customers facing blackout, only 10 of them recorded “Power Outage, start” in alarm data base in ELIN. However, the alarm data in NES system was correct for all the meters.

6. As soon as a meter generates an alarm signal we would like to be able to have access to the information. How can we receive alarm data in real time?

7. Is it possible to ping all the meters and check for acknowledgement? Does it affect other systems?

8. How much data is available in NES that is not accessible in ELIN?

9. Do you have any ideas and/or suggestions of how the smart meters (or other available systems, e.g., NES, ELIN, etc.) can be used to increase reliability? How can we use smart meters for measuring power quality? Our requirements are:

  • To be able to collect additional data from the meters without interfering with other systems
  • To be able to alter the settings on the smart meters as required (we might have to do this a few times)
  • To be able to receive the alarm data as soon as they are generated by the meters


Eltel had suggested two options according to our questions:

1. Perform memory configuration on all generation 3+ meters. One data set for billing purposes and one for analysis purposes.

2. Exclude all generation 3+ meters which are being billed on load profile data.

Outcomes of the meeting:
  • Schedule the next meeting with Eltel
  • Determine the population of meters to be reconfigured
  • Decision for what option to select for the meter configuration
  • Estimate for Custom Development for configuring, processing, parsing and storing analysis data (Eltel should do this part)

Meeting on 2015 Mar 26

We discussed about underground failure rate estimation and the important factors that have to be considered e.g. age and length of the cables. HEM also mentioned that joints are the weakest point in the underground cables and cause lots of failure; therefore, they should also be considered.

The next topic was about setting up a meeting with Eltel company some time in Apr. Before the meeting Hassan needs to prepare some question and send to the company to give them some idea of what we are going to discuss about.

HEM's board (Halmstad Kommun) presentation 2015 Feb 19

The title of the presentation was "Data Mining for Smart Power Grids". The presentation is summarized in the following.

We are employing a number of machine learning and data analysis algorithms to extract relevant information from available data. We investigate three aspects: predictive maintenance, i.e., discovering and anticipating faults before they become serious problems, load forecasting, i.e., estimating future energy consumption, and customer clustering, i.e., finding similarities between energy consumption patterns of different users.

The main focus of the project is predictive maintenance, and we are currently pursuing several directions of data analysis. They are all interesting and important for the goals of the company, but differ in their focus.

First direction is fault analysis, which consists of investigating different types of failures, finding out causes of their occurrences, as well as evaluating their importance and defining which data is required for predicting them. We look at questions such as: “which types of faults are most significant,” and “on which methods should we focus to predict them.” For example, if a fuse in a substation or underground cable fails, it is important to investigate the cause of the failure, its effects on the grid as a whole, the number of disconnected customers, and the possible methods to prevent similar problems in the future.

The second direction is history-based maintenance, which is based on analyzing the historical information of the network as a whole, or of parts of the network characterized by similar components, usage and external conditions. The goal is to determine the expected lifetime of various assets, often using probabilistic models. This can be used for planning replacement or maintenance operations, in order to prevent faults. This type of analysis can also be called preventive maintenance. For example, we can use historical data of fuses or cable faults in a particular location to recommend replacements after 8 years, even though information from the manufacturer suggests an expected lifetime of 10 years.

The third direction is symptom-based maintenance, which is based on analyzing sensor readings measuring the operation of individual grid components. Signal analysis techniques allow us to use available data, collected from smart meters and other sensors distributed on the grid, to distinguish between assets that are in good condition from those that are starting to reveal fault symptoms. This allows us to base maintenance and replacement operations not only on historical predictions, but also on the current status of the network. For example, by monitoring the energy loss in a fuse or a cable, we may discover fault symptoms building up, before the asset stops working.

Meeting on 2015 Feb 13

In this meeting we mainly discussed about the presentation for HEM’s board (Halmstad Kommun) meeting. Hassan received some comments about the slides and the key objective of the project that should be underlined.

Meeting on 2015 Jan 19

In this meeting we discussed the following issues:

  • HEM mentioned the plan of adding 10 quality measurements for 10 sub-stations. They asked us to provide some suggestions on where we would like to have these measuring devices.
  • We have the possibility to setup a meeting with Eltel Company who is responsible for upgrading the smart meters database. At the initial phase the company wanted to do some upgrade on 16 Feb and after that we can have the meeting.
  • HEM mentioned that it is possible for them to measure a suspected to failure cable in order to find the weak spot. They currently do this kind of measurement after a failure happens in a cable to localize the broken part.
  • Hassan needs to prepare a presentation about the current state of the project and present it for the HEM’s board (Halmstad Kommun). The meeting will be at 19th of February.

Meeting on 2014 Nov 18

In this day the meeting was held in two sessions. The first session was the continuous of the regular monthly meeting with HEM, and the second one devoted to meeting with steering committee regarding to Hassan’s PhD project plan.


Session 1:

In this part, Alexander informed us that by the beginning of 2015, ELTEL Company, who is the support company for smart meters, will update the scripts of the managing and modifying software for meters and data concentrators. Also, he said Hassan needs to present his project plan for inner board members of HEM in 19th of Feb. During the rest of the meeting, the content of the project plan was discussed with Alexander and Peter.


Section 2:

In this session, Hassan described his project plan during the next 4.5 years. Antanas Verikas, Stefan Byttner from Halmstad University, Robert Bass from Portland State University, and Ulf Johansson from Borås University were invited as Hassan’s support committee members. The project plan entitled with “Mining Heterogeneous Data to Improve Reliability and Efficiency of Energy Distribution in Smart Grids” and discussed:

  • The motivation of the project (problems from HEM’s perspective),
  • Scientific problems in the field of project,
  • Resources,
  • Methodologies, and
  • Goals

Predictive maintenance considered as the most important issue in the project. Predictive maintenance, which addresses problems before they escalate, contains data mining and signal analysis techniques on available data collected from smart meters and other sensors distributed on the grid to predict when maintenance should be performed. Hassan received very helpful feedbacks from steering committee, and the next update of the project plan was scheduled (May 2015).

Meeting on 2014 Oct 24

This meeting devoted to the historical failure at HEM’s electricity distribution network. The discussions were based on a document which Hassan had provided. In the following this document is attached:

In order to detect failures in the network and especially in the ground cables, we need to learn about some aspects of the available cables at HEM’s distribution network. We are also interested in the current procedures of detecting faults, determining cause of faults, and ways to fix cables. This information will allow us to investigate the methods for detecting weak spots in the cables and determining what needs maintenance. We have four discussion topic ideas but we would be more than happy to hear about anything you think would help us.

Types of cables:
  • Discussion about different types of cables , their quality, strong and weak points
  • Most faulty cables, according to their type and impact of the usage and environment
Failures:
  • Most frequent faults in the system, what can be reasons for their occurrence, what are their symptoms
  • Discussion about how you detect faults in the cables (different ways of detecting faults), and how you specify the exact position of fault in a cable. How long does it take to detect the fault or find the faulty part?
  • How do you decide what should be done with a faulty cable? Fix? Change? Other operations? What does it depends on?
  • Discussion about how you perform the actual fix or change (the procedure of changing or fixing a faulty cable)
  • Which types of cables are more sensitive to changes in temperature? Why?
  • Discussion about your opinion about the possibilities of predicting different kinds of faults, or for lowering their impact
Maintenance:
  • Discussion about the typical maintenance operations you perform: what is done, when, where etc.
  • The type of cables and the time scheduling for maintenance operations
  • Explaining of how you decide what should be done during maintenance
Reporting:
  • Discussion about how you perform reporting to the system.
  • How do you use the historical information e.g. historical faults or historical maintenance?
  • Who at HEM uses these reports? What is their main purpose?
Additional discussion:
  • Discussion about cost of failure in a cable vs cost of maintenance or changing the cable
  • Discussion about expensive faults which happen rarely and cheap faults which happen more often
  • In your opinion, what are the best ways to reduce the cost (e.g. faster detection, faster operation, predicts failure, better maintenance, etc). What should be taken into account?
  • Some examples of historical faults for discussion in terms of: How did you detect what is wrong in the system, specify where the failure is, and decide how to fix it.

Meeting on 2014 Sep 16

In this meeting we planned to invite an expert from HEM, who is responsible for reporting failures, to discuss about his experience in detecting faults, repairing and reporting them. In this case, Hassan needs to prepare some questions and send them to the expert before the next meeting.

Alexander mentioned the importance of old paper insulated cables which are one of the big causes of failure in their network. Some of these cables are installed in 1929 and needs to be refiled by oil almost every year.

Furthermore, we decided to start changing the meters program with small number of customers. For now 10 small customers (fuse limit below 63A) who are connected to one cable box (N66) are selected. In this regard, Hassan attached 2 documents. One related to the historical information about failures in N66, number of customers, geographical position; and the other one related to the list of additional data we need to collect from smart meters.

The power quality measurement device which are planned to be installed in sub-station N66 will be provided in Jan 2015. There are some old paper insulated cables connected to this sub-station and therefore, it would be a suitable place for monitoring the condition of the grid.

Meeting on 2014 Aug 22

Presentation:

In this meeting, Hassan presented results of collected data from smart meters during 01 June 2013 until 01 June 2014. In this presentation:

  • Only customers who are connected to main station H3 are examined.
  • Total active power and reactive power consumption (EL consumption) during the specified interval are calculated and demonstrated by different plots.
  • The historical fault dataset are used to visualize and investigate any pattern between EL consumption peaks and type of faults.
  • Some of anomaly behaviors recorded by customers are specified, i.e. customers with very high, negative, or zero EL consumption.
  • Correlation between active power, reactive power, fuse limitation, and number of faults for each customer calculated and depicted with different plots.
  • Historical temperature data, extracted from internet, are used to investigate any possible relation between faults and temperature.
Preparing for ELFORSK Conference:

In order to be prepared for the conference, some of the issues discussed and planned. “Eleventh Nordic Conference on Electricity Distribution System Management and Development known as NORDAC 2014 will be held at Stockholm in 8 - 9 September by Elforsk.”

Re-Programming Smart Meters in October:

It is planned to start choosing most faulty ‘breaking point’ or ‘sub-station’ and then re-programming the connected smart meters in October. In this case, Hassan needs to investigate on the type of parameters that we can collect from smart meters and could be useful for our purposes.

More Historical Data from Smart Meter:

It is decided to receive at least one more year data, for customers above 63A, from HEM to investigate how we can forecast EL consumption in the network. Furthermore, for investigate the proportion of EL consumption in causing faults such as fuse break or overload, smart meters data for customers below 63A is required. Therefore, in addition to the previous data it is decided that HEM provides the daily EL consumption data for small customers.

Meeting on 2014 June 02

In this meeting, Hassan presented results from Historical Fault Data, and Smart Meters Data. In this presentation:

  • The distribution of faults related to each year, month, season and hour was shown by using Histogram representation.
  • The total active power and reactive power for each customer, connected to the main station H7, during a specific week was shown by scatter plots. In these plots, the maximum current and the number of faults corresponding to each customer was shown by different color and size.
  • The total active power and reactive power consumption (for all the customers connected to the main station H7) was shown in 24 hours-7days in the specific week.
  • Some of the customers with maximum current above 63A were identified without any corresponding active power or reactive power in smart meters data.

Activities after meeting:

Analyzing the Main Station H3:

The main station H3 should be analyzed based on the distribution of faults related to each year, month, season and hour. Same analysis which has been performed on main station H7 needs to be done on H3.

Understanding of Missing Data:

All of the missing customers in active power and reactive power data files should be detected and the cause of why they are not in the data list should be considered.

Smart Meter:

It is important to understand different aspects of smart meter and the types of data it can transfer. In this case, the smart meter datasheet will be investigated and studied. The aim is to be able to evaluate receiving different type of data from smart meters and implementing them at HEM’s laboratory.

Literature Overview:

The articles related to: smart meter characteristics and aspects, aging cables and effects on transmitted data needs to be investigated.

Activities after meeting (2014/04/28):

ELFORSK:

Elforsk Ltd, which began operations in 1993, is owned by the Swedish Energy and the Swedish National Grid. The overall aim is to rationalize the industry-wide research and development. They provide funding for research and development in this field. [1] Eleventh Nordic Conference on Electricity Distribution System Management and Development known as NORDAC 2014 will be held at Stockholm in 8 - 9 September by Elforsk. The aims of the conference is to give the participants to discuss about future challenges and solutions, as well as disseminate experiences regarding technical, economic, and regulatory issues. The topics are very interesting and we decided to participate in the conference. Webinar for Smart Grid will be held in 9th June by Elforsk. The topics of this Webinar are: the use of controllable loads in power system, future customizing of the smart grid, and voltage fluctuations and regular production. We also plan to attend the Webinar.

Paper for SAIS workshop:

Overview of Smart Grid Challenges in Sweden, H. Nemati, A. Sant´Anna, S. Nowaczyk, 28th annual workshop of the Swedish Artificial Intelligence Society (SAIS 2014)- 22,23 May Stockholm. In this paper we present a brief overview of some of the challenges and solutions in the smart grids, focusing especially on the Swedish point of view. We discuss thirty articles, from 2006 until 2013, with the main interest on data related challenges. The paper overview gives a general perspective of the available challenges and solutions which can be employed in order to solve similar problems at HEM.

Working with data:

At the moment, I start getting familiar with the smart grid distribution system and the user interface (DpPower). Registered failures log from 2000 until May 5, 2014 are collected. Statistical analysis on this data using Python programming language is started. The data consists of 30 attributes with different values. The type of these values and relation between the attributes are very important and need to be evaluated. Until the next meeting I plan to continue working with the data to find interesting results.

Retrieved from "https://wiki.hh.se/caisr/index.php?title=Internal/HEM_Meetings&oldid=2412"