Hey everyone. Late to the party here, so I'm trying to get up to speed. To better understand what we're looking for I plotted out a few signals from the training set that are marked as having partial discharge signatures. They are signals 228, 229, and 230. I've noticed that many of the PD detections come in triplets of corresponding phase differences, but not all. Knowing that, I'm confused as to why signals 228, and 229 are marked as showing PD. Sig230 seems to show the pattern pointed out in the posted example, while the other two seem to only show transient effects. Am I missing something here?
\n\nThanks for the clarification,
\n\nBen
", "messages": [ { "id": "461712", "postDate": "01/26/2019 20:00:08", "content": "Hey everyone. Late to the party here, so I'm trying to get up to speed. To better understand what we're looking for I plotted out a few signals from the training set that are marked as having partial discharge signatures. They are signals 228, 229, and 230. I've noticed that many of the PD detections come in triplets of corresponding phase differences, but not all. Knowing that, I'm confused as to why signals 228, and 229 are marked as showing PD. Sig230 seems to show the pattern pointed out in the posted example, while the other two seem to only show transient effects. Am I missing something here?
\n\nThanks for the clarification,
\n\nBen
", "rawMarkdown": "Hey everyone. Late to the party here, so I'm trying to get up to speed. To better understand what we're looking for I plotted out a few signals from the training set that are marked as having partial discharge signatures. They are signals 228, 229, and 230. I've noticed that many of the PD detections come in triplets of corresponding phase differences, but not all. Knowing that, I'm confused as to why signals 228, and 229 are marked as showing PD. Sig230 seems to show the pattern pointed out in the posted example, while the other two seem to only show transient effects. Am I missing something here?\n\nThanks for the clarification,\n\nBen", "votes": null }, { "id": "461838", "postDate": "01/27/2019 05:30:58", "content": "Welcome to the party (of confused coding monkeys :P)\nThere's a similar post, but without conclusion.\nNevertheless should give you some info.\nhttps://www.kaggle.com/c/vsb-power-line-fault-detection/discussion/77600
\n\nAlso concerning the domain knowledge on PD you can look at the pinned top post at the forum
", "rawMarkdown": "Welcome to the party (of confused coding monkeys :P)\nThere's a similar post, but without conclusion.\nNevertheless should give you some info.\nhttps://www.kaggle.com/c/vsb-power-line-fault-detection/discussion/77600\n\nAlso concerning the domain knowledge on PD you can look at the pinned top post at the forum", "votes": null }, { "id": "462338", "postDate": "01/28/2019 06:31:21", "content": "Coming from a power systems background, my guess would be that case is a single phase fault, ie, only one phase (which would be the phase representing Sig230) suffers from partial discharge. The remaining two are fine. Single phase faults are very common in power systems, rather all the three phases (or multiple phases) being affected simultaneously is much rarer. I am guessing that out of the three phases, if any phase shows signs of a partial discharge, all the three phases seem to be labeled as a PD class, ie, class 1.
", "rawMarkdown": "Coming from a power systems background, my guess would be that case is a single phase fault, ie, only one phase (which would be the phase representing Sig230) suffers from partial discharge. The remaining two are fine. Single phase faults are very common in power systems, rather all the three phases (or multiple phases) being affected simultaneously is much rarer. I am guessing that out of the three phases, if any phase shows signs of a partial discharge, all the three phases seem to be labeled as a PD class, ie, class 1.", "votes": null }, { "id": "462351", "postDate": "01/28/2019 06:50:42", "content": "That’s what it looks like to me as well.
", "rawMarkdown": "That’s what it looks like to me as well.", "votes": null }, { "id": "462489", "postDate": "01/28/2019 11:11:39", "content": "That's interesting to know.\nBut strangely, \nIn the train set most positive measurements have all 3 phases labeled as faulty,\nplus there are indeed cases where only 1 out of the 3 phases are labeled faulty.
", "rawMarkdown": "That's interesting to know.\nBut strangely, \nIn the train set most positive measurements have all 3 phases labeled as faulty,\nplus there are indeed cases where only 1 out of the 3 phases are labeled faulty.", "votes": null }, { "id": "464886", "postDate": "02/01/2019 18:00:21", "content": "Yes, quite frankly I don't think the folks that put up the competition know what they're looking for. Otherwise, I don't understand how the scores seem so weak. I'm happy to use this competition to build skills working with these very large (to me) datasets, but I'm not really expecting to find anything terribly enlightening,
", "rawMarkdown": "Yes, quite frankly I don't think the folks that put up the competition know what they're looking for. Otherwise, I don't understand how the scores seem so weak. I'm happy to use this competition to build skills working with these very large (to me) datasets, but I'm not really expecting to find anything terribly enlightening,", "votes": null } ], "comments": [ { "id": 461838, "author_name": "ylchan87", "author_url": "", "post_date": "01/27/2019 05:30:58", "content": "Welcome to the party (of confused coding monkeys :P)\nThere's a similar post, but without conclusion.\nNevertheless should give you some info.\nhttps://www.kaggle.com/c/vsb-power-line-fault-detection/discussion/77600
\n\nAlso concerning the domain knowledge on PD you can look at the pinned top post at the forum
", "votes": null, "replies": [] }, { "id": 462338, "author_name": "bikiranguha", "author_url": "", "post_date": "01/28/2019 06:31:21", "content": "Coming from a power systems background, my guess would be that case is a single phase fault, ie, only one phase (which would be the phase representing Sig230) suffers from partial discharge. The remaining two are fine. Single phase faults are very common in power systems, rather all the three phases (or multiple phases) being affected simultaneously is much rarer. I am guessing that out of the three phases, if any phase shows signs of a partial discharge, all the three phases seem to be labeled as a PD class, ie, class 1.
", "votes": null, "replies": [ { "id": 462351, "author_name": "bhendr2", "author_url": "", "post_date": "01/28/2019 06:50:42", "content": "That’s what it looks like to me as well.
", "votes": null, "replies": [] }, { "id": 462489, "author_name": "ylchan87", "author_url": "", "post_date": "01/28/2019 11:11:39", "content": "That's interesting to know.\nBut strangely, \nIn the train set most positive measurements have all 3 phases labeled as faulty,\nplus there are indeed cases where only 1 out of the 3 phases are labeled faulty.
", "votes": null, "replies": [] }, { "id": 464886, "author_name": "bhendr2", "author_url": "", "post_date": "02/01/2019 18:00:21", "content": "Yes, quite frankly I don't think the folks that put up the competition know what they're looking for. Otherwise, I don't understand how the scores seem so weak. I'm happy to use this competition to build skills working with these very large (to me) datasets, but I'm not really expecting to find anything terribly enlightening,
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "461712": "Hey everyone. Late to the party here, so I'm trying to get up to speed. To better understand what we're looking for I plotted out a few signals from the training set that are marked as having partial discharge signatures. They are signals 228, 229, and 230. I've noticed that many of the PD detections come in triplets of corresponding phase differences, but not all. Knowing that, I'm confused as to why signals 228, and 229 are marked as showing PD. Sig230 seems to show the pattern pointed out in the posted example, while the other two seem to only show transient effects. Am I missing something here?\n\nThanks for the clarification,\n\nBen", "461838": "Welcome to the party (of confused coding monkeys :P)\nThere's a similar post, but without conclusion.\nNevertheless should give you some info.\nhttps://www.kaggle.com/c/vsb-power-line-fault-detection/discussion/77600\n\nAlso concerning the domain knowledge on PD you can look at the pinned top post at the forum", "462338": "Coming from a power systems background, my guess would be that case is a single phase fault, ie, only one phase (which would be the phase representing Sig230) suffers from partial discharge. The remaining two are fine. Single phase faults are very common in power systems, rather all the three phases (or multiple phases) being affected simultaneously is much rarer. I am guessing that out of the three phases, if any phase shows signs of a partial discharge, all the three phases seem to be labeled as a PD class, ie, class 1.", "462351": "That’s what it looks like to me as well.", "462489": "That's interesting to know.\nBut strangely, \nIn the train set most positive measurements have all 3 phases labeled as faulty,\nplus there are indeed cases where only 1 out of the 3 phases are labeled faulty.", "464886": "Yes, quite frankly I don't think the folks that put up the competition know what they're looking for. Otherwise, I don't understand how the scores seem so weak. I'm happy to use this competition to build skills working with these very large (to me) datasets, but I'm not really expecting to find anything terribly enlightening," }, "source": "meta" }