Dear All participants,
\n\nwe, as the ENET research institute, are feeling grateful and delighted by progress that has been done and is still performing in this competition. Because the number of winners is drastically small compare to the sum of all participants, especially those who really came up with something interesting and innovative, we decided to organize a special issue in an academic journal and invite you to submit your thoughts, analysis, approaches and results.
\n\nYou have to take into account that you are working on a problem that has been bothering the research for decades (PD pattern detection). On the other hand, until these days, the data obtained from our patented metering device has not been publicly tested on such a large scale. Those are the reasons why not to loose any of your original work, of course if you are willing to share. We are already in negotiations with several journals choosing the most suitable for us.
\n\nI know about some of you who are representing academia, because I sent a few invitations to friends from foreign universities, but this opportunity, as I said, is for everyone.
\n\nPlease contact me on my email address tomas.vantuch@vsb.cz if you would be interested. Also feel free to ask if you would need some support (text polishing, language correction, etc.).
\n\nThank you,\nBest regards,
\n\nTomas
", "messages": [ { "id": "458665", "postDate": "01/20/2019 08:16:41", "content": "Dear All participants,
\n\nwe, as the ENET research institute, are feeling grateful and delighted by progress that has been done and is still performing in this competition. Because the number of winners is drastically small compare to the sum of all participants, especially those who really came up with something interesting and innovative, we decided to organize a special issue in an academic journal and invite you to submit your thoughts, analysis, approaches and results.
\n\nYou have to take into account that you are working on a problem that has been bothering the research for decades (PD pattern detection). On the other hand, until these days, the data obtained from our patented metering device has not been publicly tested on such a large scale. Those are the reasons why not to loose any of your original work, of course if you are willing to share. We are already in negotiations with several journals choosing the most suitable for us.
\n\nI know about some of you who are representing academia, because I sent a few invitations to friends from foreign universities, but this opportunity, as I said, is for everyone.
\n\nPlease contact me on my email address tomas.vantuch@vsb.cz if you would be interested. Also feel free to ask if you would need some support (text polishing, language correction, etc.).
\n\nThank you,\nBest regards,
\n\nTomas
", "rawMarkdown": "Dear All participants,\n\nwe, as the ENET research institute, are feeling grateful and delighted by progress that has been done and is still performing in this competition. Because the number of winners is drastically small compare to the sum of all participants, especially those who really came up with something interesting and innovative, we decided to organize a special issue in an academic journal and invite you to submit your thoughts, analysis, approaches and results. \n\nYou have to take into account that you are working on a problem that has been bothering the research for decades (PD pattern detection). On the other hand, until these days, the data obtained from our patented metering device has not been publicly tested on such a large scale. Those are the reasons why not to loose any of your original work, of course if you are willing to share. We are already in negotiations with several journals choosing the most suitable for us.\n\nI know about some of you who are representing academia, because I sent a few invitations to friends from foreign universities, but this opportunity, as I said, is for everyone.\n\nPlease contact me on my email address tomas.vantuch@vsb.cz if you would be interested. Also feel free to ask if you would need some support (text polishing, language correction, etc.).\n\nThank you,\nBest regards,\n\nTomas", "votes": null }, { "id": "458954", "postDate": "01/20/2019 22:02:01", "content": "That's great! Regardless of where I end up in the final leader board, I personally plan on releasing a public kernel with my approaches to signal processing and modeling once the submission deadline passes.
\n\nI know most will keep their progress and methods private during the competition phase of the challenge but I hope others follow suit and that collectively your team can learn some novel approaches to improve PD fault detection.
", "rawMarkdown": "That's great! Regardless of where I end up in the final leader board, I personally plan on releasing a public kernel with my approaches to signal processing and modeling once the submission deadline passes. \n\nI know most will keep their progress and methods private during the competition phase of the challenge but I hope others follow suit and that collectively your team can learn some novel approaches to improve PD fault detection.", "votes": null }, { "id": "482435", "postDate": "03/03/2019 01:06:24", "content": "Hi Tomas!\nAfter looking at your data and studying a little about PD patterns, I've been thinking about an alternative approach for measurement, that would possibly make the detection task much simpler.
\n\nI'm just wandering, would it be possible to modify your device such that it would track an average of N consecutive cycles instead of a single one (i.e., turn your sample buffer into accumulator). Or, even simpler - just repeat the same measurement N times and combine them together afterwards (even non consecutive out-of sync signals could be easily aligned together using the 50hz phase as a reference).
\n\nThe idea here is to obtain a \"long exposure shot\", which should reveal the traces of reoccurring PD pattern better than a capture of a single cycle. As a bonus, some interference will be suppressed by averaging :)
\n\nJust curious, if you already considered this approach. It may give a subject for another publication. Please, let me know it it makes any sense to you.
", "rawMarkdown": "Hi Tomas!\nAfter looking at your data and studying a little about PD patterns, I've been thinking about an alternative approach for measurement, that would possibly make the detection task much simpler.\n\nI'm just wandering, would it be possible to modify your device such that it would track an average of N consecutive cycles instead of a single one (i.e., turn your sample buffer into accumulator). Or, even simpler - just repeat the same measurement N times and combine them together afterwards (even non consecutive out-of sync signals could be easily aligned together using the 50hz phase as a reference). \n\nThe idea here is to obtain a \"long exposure shot\", which should reveal the traces of reoccurring PD pattern better than a capture of a single cycle. As a bonus, some interference will be suppressed by averaging :)\n\nJust curious, if you already considered this approach. It may give a subject for another publication. Please, let me know it it makes any sense to you.", "votes": null }, { "id": "482814", "postDate": "03/03/2019 17:34:01", "content": "I think this is a great suggestion!
\n\nDoing what you describe would allow more effective identification of persistent effects like partial discharge faults while presumably also allowing for easier cancellation of sporadic environmental effects like corona.
\n\nWithout even changing the hardware or collection methodology, even just making available a feature that points to the sampling equipment's ID or the power line segment's ID would be helpful. This data must also already exist because you would presumably want to know where to perform maintenance once a fault is detected.
", "rawMarkdown": "I think this is a great suggestion!\n\nDoing what you describe would allow more effective identification of persistent effects like partial discharge faults while presumably also allowing for easier cancellation of sporadic environmental effects like corona.\n\nWithout even changing the hardware or collection methodology, even just making available a feature that points to the sampling equipment's ID or the power line segment's ID would be helpful. This data must also already exist because you would presumably want to know where to perform maintenance once a fault is detected.", "votes": null }, { "id": "484134", "postDate": "03/05/2019 16:00:12", "content": "Exactly! However, I'm not sure if PD occurs in any conditions of a faulty line, regardless of its power state (i.e, voltage drop, asymmetry, power factor, harmonics, temperature, etc.). So it would be safer to have measurements as close as possible in time (consecutive cycles would be ideal), to guarantee sampling under the same conditions.
", "rawMarkdown": "Exactly! However, I'm not sure if PD occurs in any conditions of a faulty line, regardless of its power state (i.e, voltage drop, asymmetry, power factor, harmonics, temperature, etc.). So it would be safer to have measurements as close as possible in time (consecutive cycles would be ideal), to guarantee sampling under the same conditions.", "votes": null }, { "id": "487262", "postDate": "03/10/2019 13:52:47", "content": "My approach doesn't appear to be the best (strictly going by leader board score) but perhaps there is still something others may learn from it. True to form, I've made it public:
\n\nhttps://www.kaggle.com/jeffreyegan/vsb-power-line-fault-detection-approach
\n\nMy kernel/notebook isn't incredibly verbose where it comes to the DSP aspects, I've written up more in the documentation directory of my git repository for the project.
", "rawMarkdown": "My approach doesn't appear to be the best (strictly going by leader board score) but perhaps there is still something others may learn from it. True to form, I've made it public: \n\nhttps://www.kaggle.com/jeffreyegan/vsb-power-line-fault-detection-approach\n\nMy kernel/notebook isn't incredibly verbose where it comes to the DSP aspects, I've written up more in the documentation directory of my git repository for the project.", "votes": null }, { "id": "857122", "postDate": "05/22/2020 10:22:28", "content": "Dear Sir, This competition was over for one year. I wonders that are there any papers published from this? If any available, please share. I want to learn more. Thank you very much.
", "rawMarkdown": "Dear Sir, This competition was over for one year. I wonders that are there any papers published from this? If any available, please share. I want to learn more. Thank you very much.", "votes": null }, { "id": "1016457", "postDate": "09/19/2020 01:47:09", "content": "Thank you very much, your algorithm is very helpful to me
", "rawMarkdown": "Thank you very much, your algorithm is very helpful to me", "votes": null } ], "comments": [ { "id": 458954, "author_name": "jeffreyegan", "author_url": "", "post_date": "01/20/2019 22:02:01", "content": "That's great! Regardless of where I end up in the final leader board, I personally plan on releasing a public kernel with my approaches to signal processing and modeling once the submission deadline passes.
\n\nI know most will keep their progress and methods private during the competition phase of the challenge but I hope others follow suit and that collectively your team can learn some novel approaches to improve PD fault detection.
", "votes": null, "replies": [ { "id": 487262, "author_name": "jeffreyegan", "author_url": "", "post_date": "03/10/2019 13:52:47", "content": "My approach doesn't appear to be the best (strictly going by leader board score) but perhaps there is still something others may learn from it. True to form, I've made it public:
\n\nhttps://www.kaggle.com/jeffreyegan/vsb-power-line-fault-detection-approach
\n\nMy kernel/notebook isn't incredibly verbose where it comes to the DSP aspects, I've written up more in the documentation directory of my git repository for the project.
", "votes": null, "replies": [] }, { "id": 1016457, "author_name": "icandoeverything", "author_url": "", "post_date": "09/19/2020 01:47:09", "content": "Thank you very much, your algorithm is very helpful to me
", "votes": null, "replies": [] } ] }, { "id": 482435, "author_name": "m0nzderr", "author_url": "", "post_date": "03/03/2019 01:06:24", "content": "Hi Tomas!\nAfter looking at your data and studying a little about PD patterns, I've been thinking about an alternative approach for measurement, that would possibly make the detection task much simpler.
\n\nI'm just wandering, would it be possible to modify your device such that it would track an average of N consecutive cycles instead of a single one (i.e., turn your sample buffer into accumulator). Or, even simpler - just repeat the same measurement N times and combine them together afterwards (even non consecutive out-of sync signals could be easily aligned together using the 50hz phase as a reference).
\n\nThe idea here is to obtain a \"long exposure shot\", which should reveal the traces of reoccurring PD pattern better than a capture of a single cycle. As a bonus, some interference will be suppressed by averaging :)
\n\nJust curious, if you already considered this approach. It may give a subject for another publication. Please, let me know it it makes any sense to you.
", "votes": null, "replies": [ { "id": 482814, "author_name": "jeffreyegan", "author_url": "", "post_date": "03/03/2019 17:34:01", "content": "I think this is a great suggestion!
\n\nDoing what you describe would allow more effective identification of persistent effects like partial discharge faults while presumably also allowing for easier cancellation of sporadic environmental effects like corona.
\n\nWithout even changing the hardware or collection methodology, even just making available a feature that points to the sampling equipment's ID or the power line segment's ID would be helpful. This data must also already exist because you would presumably want to know where to perform maintenance once a fault is detected.
", "votes": null, "replies": [] }, { "id": 484134, "author_name": "m0nzderr", "author_url": "", "post_date": "03/05/2019 16:00:12", "content": "Exactly! However, I'm not sure if PD occurs in any conditions of a faulty line, regardless of its power state (i.e, voltage drop, asymmetry, power factor, harmonics, temperature, etc.). So it would be safer to have measurements as close as possible in time (consecutive cycles would be ideal), to guarantee sampling under the same conditions.
", "votes": null, "replies": [] } ] }, { "id": 857122, "author_name": "nlebang", "author_url": "", "post_date": "05/22/2020 10:22:28", "content": "Dear Sir, This competition was over for one year. I wonders that are there any papers published from this? If any available, please share. I want to learn more. Thank you very much.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "458665": "Dear All participants,\n\nwe, as the ENET research institute, are feeling grateful and delighted by progress that has been done and is still performing in this competition. Because the number of winners is drastically small compare to the sum of all participants, especially those who really came up with something interesting and innovative, we decided to organize a special issue in an academic journal and invite you to submit your thoughts, analysis, approaches and results. \n\nYou have to take into account that you are working on a problem that has been bothering the research for decades (PD pattern detection). On the other hand, until these days, the data obtained from our patented metering device has not been publicly tested on such a large scale. Those are the reasons why not to loose any of your original work, of course if you are willing to share. We are already in negotiations with several journals choosing the most suitable for us.\n\nI know about some of you who are representing academia, because I sent a few invitations to friends from foreign universities, but this opportunity, as I said, is for everyone.\n\nPlease contact me on my email address tomas.vantuch@vsb.cz if you would be interested. Also feel free to ask if you would need some support (text polishing, language correction, etc.).\n\nThank you,\nBest regards,\n\nTomas", "458954": "That's great! Regardless of where I end up in the final leader board, I personally plan on releasing a public kernel with my approaches to signal processing and modeling once the submission deadline passes. \n\nI know most will keep their progress and methods private during the competition phase of the challenge but I hope others follow suit and that collectively your team can learn some novel approaches to improve PD fault detection.", "482435": "Hi Tomas!\nAfter looking at your data and studying a little about PD patterns, I've been thinking about an alternative approach for measurement, that would possibly make the detection task much simpler.\n\nI'm just wandering, would it be possible to modify your device such that it would track an average of N consecutive cycles instead of a single one (i.e., turn your sample buffer into accumulator). Or, even simpler - just repeat the same measurement N times and combine them together afterwards (even non consecutive out-of sync signals could be easily aligned together using the 50hz phase as a reference). \n\nThe idea here is to obtain a \"long exposure shot\", which should reveal the traces of reoccurring PD pattern better than a capture of a single cycle. As a bonus, some interference will be suppressed by averaging :)\n\nJust curious, if you already considered this approach. It may give a subject for another publication. Please, let me know it it makes any sense to you.", "482814": "I think this is a great suggestion!\n\nDoing what you describe would allow more effective identification of persistent effects like partial discharge faults while presumably also allowing for easier cancellation of sporadic environmental effects like corona.\n\nWithout even changing the hardware or collection methodology, even just making available a feature that points to the sampling equipment's ID or the power line segment's ID would be helpful. This data must also already exist because you would presumably want to know where to perform maintenance once a fault is detected.", "484134": "Exactly! However, I'm not sure if PD occurs in any conditions of a faulty line, regardless of its power state (i.e, voltage drop, asymmetry, power factor, harmonics, temperature, etc.). So it would be safer to have measurements as close as possible in time (consecutive cycles would be ideal), to guarantee sampling under the same conditions.", "487262": "My approach doesn't appear to be the best (strictly going by leader board score) but perhaps there is still something others may learn from it. True to form, I've made it public: \n\nhttps://www.kaggle.com/jeffreyegan/vsb-power-line-fault-detection-approach\n\nMy kernel/notebook isn't incredibly verbose where it comes to the DSP aspects, I've written up more in the documentation directory of my git repository for the project.", "857122": "Dear Sir, This competition was over for one year. I wonders that are there any papers published from this? If any available, please share. I want to learn more. Thank you very much.", "1016457": "Thank you very much, your algorithm is very helpful to me" }, "source": "meta" }