Since I know far more about campus utility production and distribution systems of chilled water, steam, electricity, and hot water than ML, I want discuss the likelihood of developing a useful solution to this competition. It is clear from the leaked data that many of the sites are universities with district energy systems, i.e. centralized plants that deliver energy products (chilled water, steam, etc.) to the campus buildings. In those cases, there is no production/generating equipment in the buildings, and the energy consumption is metered at the building entrance. Stand-alone buildings contain internal production/generating equipment. So why does that matter?
\n\nFor district energy supplied buildings, the various products are fairly independent, although there are interactions if heating and cooling are supplied simultaneously. For stand-alone buildings, unless the meter readings are properly prepared, they will double count consumption. For example, a building chiller that produces chilled water will use electricity that will be recorded at the electrical meter, but the energy output of the chilled water delivered will also be metered. The relationship between the output of the chiller and its input is an efficiency measure e.g., KW/Ton, and must be subtracted from the electrical meter reading to give a true reading of building energy consumption. Suppose further that the building is supplied by rooftop units which contain independent fans, refrigeration units, and heating, possibly electric, all in a single enclosure. Now you're seeing everything as electric. A similar building will have an entirely different energy profile. Site information won't help you here. You need an independent solution for each building.
\n\nThis suggests to me that categorization of the building metadata needs to be improved to include such information, and that, probably, this entire competition is a waste of time for such a large amorphous set of buildings.
", "messages": [ { "id": 678024, "postDate": "2019-11-20T22:46:29.427Z", "content": "Since I know far more about campus utility production and distribution systems of chilled water, steam, electricity, and hot water than ML, I want discuss the likelihood of developing a useful solution to this competition. It is clear from the leaked data that many of the sites are universities with district energy systems, i.e. centralized plants that deliver energy products (chilled water, steam, etc.) to the campus buildings. In those cases, there is no production/generating equipment in the buildings, and the energy consumption is metered at the building entrance. Stand-alone buildings contain internal production/generating equipment. So why does that matter?
\n\nFor district energy supplied buildings, the various products are fairly independent, although there are interactions if heating and cooling are supplied simultaneously. For stand-alone buildings, unless the meter readings are properly prepared, they will double count consumption. For example, a building chiller that produces chilled water will use electricity that will be recorded at the electrical meter, but the energy output of the chilled water delivered will also be metered. The relationship between the output of the chiller and its input is an efficiency measure e.g., KW/Ton, and must be subtracted from the electrical meter reading to give a true reading of building energy consumption. Suppose further that the building is supplied by rooftop units which contain independent fans, refrigeration units, and heating, possibly electric, all in a single enclosure. Now you're seeing everything as electric. A similar building will have an entirely different energy profile. Site information won't help you here. You need an independent solution for each building.
\n\nThis suggests to me that categorization of the building metadata needs to be improved to include such information, and that, probably, this entire competition is a waste of time for such a large amorphous set of buildings.
", "rawMarkdown": "Since I know far more about campus utility production and distribution systems of chilled water, steam, electricity, and hot water than ML, I want discuss the likelihood of developing a useful solution to this competition. It is clear from the leaked data that many of the sites are universities with district energy systems, i.e. centralized plants that deliver energy products (chilled water, steam, etc.) to the campus buildings. In those cases, there is no production/generating equipment in the buildings, and the energy consumption is metered at the building entrance. Stand-alone buildings contain internal production/generating equipment. So why does that matter? \n\nFor district energy supplied buildings, the various products are fairly independent, although there are interactions if heating and cooling are supplied simultaneously. For stand-alone buildings, unless the meter readings are properly prepared, they will double count consumption. For example, a building chiller that produces chilled water will use electricity that will be recorded at the electrical meter, but the energy output of the chilled water delivered will also be metered. The relationship between the output of the chiller and its input is an efficiency measure e.g., KW/Ton, and must be subtracted from the electrical meter reading to give a true reading of building energy consumption. Suppose further that the building is supplied by rooftop units which contain independent fans, refrigeration units, and heating, possibly electric, all in a single enclosure. Now you're seeing everything as electric. A similar building will have an entirely different energy profile. Site information won't help you here. You need an independent solution for each building.\n\nThis suggests to me that categorization of the building metadata needs to be improved to include such information, and that, probably, this entire competition is a waste of time for such a large amorphous set of buildings. ", "votes": 1 } ], "comments": [], "raw_markdown_by_id": { "678024": "Since I know far more about campus utility production and distribution systems of chilled water, steam, electricity, and hot water than ML, I want discuss the likelihood of developing a useful solution to this competition. It is clear from the leaked data that many of the sites are universities with district energy systems, i.e. centralized plants that deliver energy products (chilled water, steam, etc.) to the campus buildings. In those cases, there is no production/generating equipment in the buildings, and the energy consumption is metered at the building entrance. Stand-alone buildings contain internal production/generating equipment. So why does that matter? \n\nFor district energy supplied buildings, the various products are fairly independent, although there are interactions if heating and cooling are supplied simultaneously. For stand-alone buildings, unless the meter readings are properly prepared, they will double count consumption. For example, a building chiller that produces chilled water will use electricity that will be recorded at the electrical meter, but the energy output of the chilled water delivered will also be metered. The relationship between the output of the chiller and its input is an efficiency measure e.g., KW/Ton, and must be subtracted from the electrical meter reading to give a true reading of building energy consumption. Suppose further that the building is supplied by rooftop units which contain independent fans, refrigeration units, and heating, possibly electric, all in a single enclosure. Now you're seeing everything as electric. A similar building will have an entirely different energy profile. Site information won't help you here. You need an independent solution for each building.\n\nThis suggests to me that categorization of the building metadata needs to be improved to include such information, and that, probably, this entire competition is a waste of time for such a large amorphous set of buildings. " } }