System and method for detecting the degree of filter fouling

Abstract

A system and method for identifying the degree of dirtiness of a current filter. The system includes sensors (308, 310, 322, 324, 1808, 1810, 1822, 1824) that are coupled to a controller (104) to collect pressure differential sensor data and flow rate data associated with a material flow passing through filters (306, 320, 1806, 1820). The controller (104) applies a first filtered data set to a portion of the pressure differential sensor data and the flow rate data and applies a second filtered data set to the first filtered data set to further smooth the first filtered data set. The controller (104) further applies an edge detection filter to the second filtered data set, thereby obtaining an edge detection filtered data set and using the edge detection filtered data set and the flow rate data to determine a filter replacement threshold and an optimal filter replacement date.

Description

Technical Field

[0001] The present invention generally relates to filters, and more particularly to the efficiency and life of filters. Background Art

[0002] In commercial buildings, air handling unit (AHU) or rooftop unit (RTU) air filters should be replaced correctly. Premature replacement will result in additional hardware and labor costs. Insufficient replacement will lead to increased energy costs and can endanger the health of occupants.

[0003] In standard practice, engineers blindly replace filters according to a fixed schedule, such as replacing filters every three months, without checking the actual degree of dirtiness. This method is not accurate. Depending on outdoor air quality (OAQ) and indoor air quality (IAQ) conditions and the intake fan status, the degree of dirtiness can increase faster or slower. Therefore, it brings additional equipment / labor costs or higher energy costs again.

[0004] Another known practice is to replace the filter when the pressure difference (DP) reaches a threshold. However, due to sensor data interference being inherent to sensors, this method is unreliable. When the current filter DP value varies significantly due to abnormal and / or interfering sensor readings, a signal indicating that the filter is dirty is issued and the readings become untrustworthy. Similar problems exist in the filtration of other media or materials (such as water, oil, etc.).

[0005] There is a need for a method to determine that a filter is dirty and needs to be replaced, while avoiding premature / frequent replacement of filters or operating a system with a filter that needs to be replaced. Summary of the Invention

[0006] The present invention describes a method and system for using multiple sensors to monitor the efficiency of the dirty side or input side of a filter and the output side of the filter. Then the sensor data is collected, averaged, or otherwise standardized to create a time-related filter curve. Then the filter curve is further processed to predict when the threshold for filter replacement will be reached. Determine the date for replacing the filter. The date for replacing the filter also depends on the filter cost and labor cost required for replacing the filter. The filter can be an air filter, a water filter, or other types of material filters, where there is a pressure difference between different sides of the filter.

[0007] When studying the following drawings and the detailed description, other devices, apparatuses, systems, methods, features, and advantages of the present invention will be obvious or will become obvious to those skilled in the art. The intention is that all such additional systems, methods, features, and advantages are included in this specification, within the scope of the present invention, and are protected by the appended claims. Brief Description of the Drawings

[0008] The present invention can be better understood by reference to the following drawings. The components in the drawings are not necessarily drawn to scale, but rather the emphasis is on explaining the principles of the present invention. In the drawings, like reference numerals indicate corresponding parts in different views.

[0009] Figure 1 A schematic diagram showing a first building having a main building controller coupled to a network and a second building having a building automation system according to an exemplary embodiment of the present invention.

[0010] Figure 2 Showing according to an exemplary embodiment of the present invention Figure 1 a schematic diagram of a building controller.

[0011] Figure 3 Showing according to an exemplary embodiment of the present invention as Figure 1 a schematic diagram of an air flow system that is part of a building automation system of a building.

[0012] Figure 4 Showing according to an exemplary embodiment of the present invention data from input sensors of an air flow system from monitoring Figure 3 a chart.

[0013] Figure 5 Showing according to an exemplary embodiment data from Figure 3 a filter curve of processed data of a system.

[0014] Figure 6 Showing according to an exemplary embodiment of the present invention different types of sensor data instances that can be collected and processed from elements of a system from Figure 3 a schematic diagram.

[0015] Figure 7 Showing according to an exemplary embodiment cleaned sensor data from a DP and data from the flow rate from Figure 3 a chart.

[0016] Figure 8 Showing according to an exemplary embodiment of the present invention data from Figure 7 a day a filter curve of a chart.

[0017] Figure 9 Showing according to an exemplary embodiment of the present invention a chart of multi-day reference DP values after applying a regression operation (such as a least squares operation).

[0018] Figure 10A graph showing multi - day reference DP values after applying least - squares operations twice according to an exemplary embodiment of the present invention.

[0019] Figure 11 A graph showing the degree of dirtiness over multiple days after considering fan status data according to an exemplary embodiment of the present invention.

[0020] Figure 12 A graph showing the degree of dirtiness after applying edge detection according to an exemplary embodiment of the present invention.

[0021] Figure 13 Depicts according to an exemplary embodiment of the present invention Figure 11 A graph of the dynamic segmentation of data.

[0022] Figure 14 A schematic diagram showing a learning method for identifying a filter model according to an exemplary embodiment of the present invention.

[0023] Figure 15 A graph showing an example of daily air quality at several sampling points in a predetermined geographical area according to an exemplary embodiment of the present invention.

[0024] Figure 16 A graph showing the estimated degree of dirtiness based on data according to an exemplary embodiment of the present invention.

[0025] Figure 17 A graph showing the optimal filter replacement date derived from processed data according to an exemplary embodiment of the present invention.

[0026] Figure 18 Shows according to an exemplary embodiment of the present invention as Figure 1 A schematic diagram of a water flow system that is part of a building automation system for a building.

[0027] Figure 19 A flowchart showing a method for determining when a filter needs to be replaced according to an exemplary embodiment of the present invention. Detailed Description

[0028] The present invention describes a method and system for using multiple sensors to monitor the efficiency of the dirty side or input side of a filter and the output side of the filter.

[0029] In Figure 1In FIG. 100, a schematic diagram of a network - centric building 102 with a main building controller 104 coupled to a network 106 and a building 108 with a building automation system 110 is depicted according to an exemplary embodiment of the present invention. The building automation system 110 has a controller 112 connected to the network 106. The building automation system 110 is provided, for example, by Siemens. The controller 112 can be a dedicated computer running an operating system such as LINUX or WINDOWS. The network connection can be a wired Ethernet, a wireless Ethernet, a cellular, a packet, an ISDN, or other types of network connections that can provide data for transmission via the network. The main building controller 104 is shown within the building 102, but in fact, the main building controller can be positioned as software executed on a server located in a server farm accessible via the Internet in a manner commonly referred to as cloud computing.

[0030] Figure 2 A schematic diagram 200 of a main building controller 104 according to an exemplary embodiment of the present invention is shown. Figure 1 The main building controller 104 has a controller / processor 202 that is coupled to a memory 206, a data storage 212, a network interface 214, an input / output interface 216, a display interface 218, and a power supply 220, all of which are connected via a bus 204. The bus 204 is shown as a power / data bus, but in fact, the main building controller 104 can have multiple separate buses that include data, address, and power. The memory 206 is divided into an operating system memory 208 and an application memory 210. The application memory 210 contains instructions that, when executed, are used to store filter data in the data storage 212 and process the data to determine whether a filter needs to be replaced and the date of filter replacement. The data storage 212 is depicted within the main building controller 104, but in other embodiments, the data storage 212 can reside externally or even on the network 106 or in the cloud. Also, in some embodiments, the main building controller 104 can be implemented in a building with a building automation system 110.

[0031] In Figure 3 FIG. 200, a schematic diagram of a main building controller 104 is depicted as Figure 1Schematic of an air flow system 300 that is part of a building automation system 110 for a building 108. Outdoor / external air 302 enters the building 108 and passes through a damper 304. A pressure sensor 308 measures the flow or pressure of air entering a physical filter, such as air filter 306. The flow or pressure of air leaving the filter is measured by a pressure sensor 310. It should be understood that pressure sensors and flow sensors are used interchangeably. The pressure drop determined at the DP1 sub-controller 312 is determined and the pressure drop is sampled periodically (pressure differential sensor data). Note that other embodiments can have any number of pressure differential sensors. The resulting data is sent to a data store for further processing. DP1 is a sub-controller connected to a building controller. In the current embodiment, the air filter 306 can be a pleated fiberglass filter. In other embodiments, other types of filters and filter materials can be used.

[0032] The air flow from the air filter 306 passes through air coolers 314 and 316. A centrifugal fan 318 increases the air pressure or flow through a pressure sensor 324 and then through an air filter 320 and a pressure sensor 322. A pressure differential DP2 sub-controller 326 determines the pressure drop across the air filter and periodically provides data to a data store for further processing. Additional data, such as current consumption and fan speed, is generated by sensors associated with the centrifugal fan and is also stored in the data store. Another centrifugal fan 328 increases the flow or pressure of the return air 330. A portion 332 of the return air flows through a damper 334. Another portion 338 of the return air is exhausted from the building 108 via an air damper 336.

[0033] Turning Figure 4 , according to an exemplary embodiment of the present invention depicts from monitoring Figure 3Graphs 402, 404, 406, 408, 410, and 412 of data (a predefined set or period of data) from the input sensors 306, 308, 322, and 324 of the air flow system 300. As shown in graph 408, the pre-filter differential pressure (DP) values measured at the DP1 sub-controller 312 vary significantly during the day. In the graph, the x-axis is the sample number. In the current implementation, the sampling rate of the data at the DP1 sub-controller 312 is 15 minutes, but in other implementations, different durations can be used. In the current implementation, the duration is one day, but in other implementations, other durations can be used. Additionally, there are cases of sensor readings with anomalies and / or disturbances, as shown in graph 408, where the pre-filter DP has different values when the flow 412 is constant. An increase in the DP of the flow results in an increase in the filter curve. However, graphs 402 and 404 are the measured filter curves, which are vertical lines during the anomaly and / or disturbance periods. These anomaly and / or disturbance periods can be caused by faults, maintenance, abnormal and disturbing situations, and need to be identified and considered during the dirtiness detection. Otherwise, the results will not be as reliable as often occurred in the previously known methods. The theoretical filter curve is shown in graph 412, where the filter curve is plotted on a DP-flow graph. Due to the basic physical principles, it is inaccurate to estimate the dirtiness level using only the DP: we need at least the DP and a flow meter to estimate the dirtiness of the filter. Additionally, data analysis indicates that we need to aggregate data over multiple days to remove the disturbances in the sensor data when processing the data. It should be noted that in other exemplary implementations, a post-filter or other filters can be used.

[0034] In Figure 5 , a graph 502 of filter curves 504 to 510 of processed data from the Figure 3 air flow system 300 is depicted according to an exemplary implementation. Sensor data is collected from the centrifugal fan 318 and the DP2 sub-controller 326. It should be noted that in other implementations, more or fewer differential pressure sensors can be used for other HVAC systems. When processing at the Figure 1 main building controller 104 of the Figure 5 for more than a predefined time, the data for the graph is obtained. Curve 504 shows that as the air flow rate increases, the fan pressure from the centrifugal fan 318 increases (the normalized value of the differential pressure), which is commonly referred to as the system curve. As the filter gets dirty, the system curve 506 has moved upward, indicating that the air flow rate decreases as the fan pressure increases. With a clean filter, the fan speed curve 508 of the centrifugal fan 318 is lower compared to that of the dirty filter.

[0035] Among them, the system curve 504 and the fan curve of the centrifugal fan 318 intersect or converge at the first rotational speed 508, and the operating state of the clean air filter can be recognized (point "A" 512). Similarly, the system curve 506 and the fan curve of the dirty air filter rotational speed 510 intersect or converge in the operating state of the dirty air filter, which can be recognized (point "B" 514). The initial point can be recognized in the building automation system 110 considering filter materials and environmental conditions such as smoke. During the operation of the building automation system, over time, the operating state of the dirty air filter can be modified to more accurately signal when the filter must be replaced. Such a signal is typically an alarm message or alert generated by the building automation system 110. More sensor data from other sensors can be used in other embodiments and combined with or used in place of the fan rotational speed and air flow rate.

[0036] In Figure 6 , an exemplary embodiment according to the present invention shows a schematic diagram 600 representing examples of different types of input sensor data 602 that can be collected and processed from the components of the air flow system 300 of Figure 3 . Some examples of the input sensor data 602 include pressure difference, intake fan rotational speed, air flow rate, fan status, and fan energy (energy sensor data). In other embodiments, additional or different combinations of sensor data associated with air pressure / air flow can be employed. In still other embodiments, the sensors can be associated with different types of media being filtered, such as liquids (water, oil, chemicals) filtered by a liquid filter (a filter manufactured for filtering liquids). The input sensor data is sourced from Figure 1 's building automation system 110 and sent to the main building controller 104 via the network 106. The input sensor data 602 is then processed using the algorithm 608 to generate outputs 604 such as a dirtiness index, an estimated filter replacement date, and an estimate of energy waste.

[0037] Turning to Figure 7 , according to an exemplary embodiment, a chart of clean sensor data from the DP2 sub - controller 326 and the flow rate from Figure 3 is depicted. The process starts with the raw input data of the intake fan variable speed drive (svd) / flow rate 702 (svd provided by a sensor located in the fan 318), the ratio of the pre - filter DP 704 and the flow rate 706 (data for one day in the current example). Since the DP2 sub - controller 326 sensor values vary significantly over time, as explained previously, detecting the dirtiness level based solely on the DP readings is unreliable.

[0038] In Figure 8in which, according to an exemplary embodiment of the present invention, a graph of filter curves 802 and 804 in the measurement of the air flow system 300 over one day from Figure 7 is shown. Due to the data being highly interfering, it is not practical to use only one day's data to detect the "degree of dirtiness" of the filter. The following data is associated with filters having approximately the same degree of dirtiness, but their filter curves are significantly different. Given a flow rate f and a DP value "d". If "d" is defined as:

[0039]

[0040] and

[0041] A = [1, f, f 2 , f 3

[0042] then f ∈ R n is a flow rate vector, which is defined as

[0043]

[0044] In other embodiments, the order of "A" can be higher or lower, such as

[0045] A = [1, f]

[0046] A = [1, f, f 2

[0047] or even

[0048] A = [1, f, f 2 , f 3 , f 4 .

[0049] Generally, the more accurate the flow meter, the better the performance provided by the higher-order model. Therefore,

[0050] d = Ab,

[0051] where "b" is the unknown coefficient of the filter curve. However, "b" can be found using the least squares regression method

[0052] b = A + b,

[0053] and "A + " is the pseudo-inverse matrix of "A". The resulting filter curves 802 and 804 are obtained.

[0054] Turning to Figure 9 ​​, An exemplary embodiment according to the present invention depicts a graph 902 of multi - day reference DP values after applying a regression operation (such as a least - squares operation). The daily filter parameters are plotted for all times (in this example, one year). The x - axis is the date. The y - axis is the expected DP at a reference flow rate (in this example, 18,000 cubic feet per minute). As shown in graph 902, the data is very noisy, making it unclear how many and which are the filter replacement days in the dataset. For simplicity, a first - order filter curve model is used to illustrate the method. Given the DP level d[i] at time instance i, we have a vector "d" with "M" elements:

[0055]

[0056] The A matrix is re - defined as:

[0057]

[0058] And it is an M - by - (N + 1) matrix, where N is the number of days. The degree of fouling on the j - th day is denoted as b[j], and the common slope of the filter curve is b0, resulting in the following equation:

[0059] d = Ab

[0060]

[0061] The vector "b" can also be solved by the least - squares regression method:

[0062] b = A + d

[0063] The multi - day reference DP values obtained after one least - squares regression are depicted in graph 902.

[0064] In Figure 10 , an exemplary embodiment according to the present invention depicts a graph 1002 of multi - day reference DP values after applying two least - squares operations. The data smoothed using the least - squares method is shown as a dark solid line in graph 1002. The lighter lines are the unsmoothed input (data).

[0065] Turning to Figure 11 , an exemplary embodiment according to the present invention shows graphs 1102 and 1104 of the multi - day fouling degree after considering the fan status data. The fouling degree (Df) is calculated and the fan status is applied to the calculation. A descending ramp - up and ramp - down time is shown in graph 1102, depicting two filter replacement situations over a year. It should be noted that the second Df drop around day 250 is of approximately the same importance as the drop on day 50. If a window of two adjacent days is used, false alarms will occur and more than two filter replacements will be detected.

[0066] In Figure 12 , Chart 1202 of the degree of fouling after applying edge detection is shown according to an exemplary embodiment. The Sobel filter (also known as the Sobel-Feldman operator) is used for the descending edge detection of the Df curve. In the current embodiment, the Sobel filter is used. In other embodiments, different edge detection methods can be employed, such as the Canny, Prewitt, or Laplacian edge detection methods. The resulting peak identifies the filter replacement date.

[0067] Turning to Figure 13 , Chart 1302 and 1304 of the dynamic segmentation of the Figure 11 data are shown according to an exemplary embodiment. Dynamic segmentation allows multiple attribute sets to be associated with any part of the linear feature. Using the current data, the peaks are clustered into segments. Using the standard deviation, the filter replacement threshold is calculated. Points above the dynamic threshold are selected. The selected points are identified as the small dots in Chart 1302. These points are then segmented into two clusters that identify two filter replacement dates.

[0068] In Figure 14 , a schematic diagram 1400 of a learning method for identifying a filter model is shown according to an exemplary embodiment. When there are particles in the air that can be captured by the filter, the filter gets dirty. If the degree of fouling of the filter is Df, the accumulation of the degree of fouling is calculated by the following filter model.

[0069]

[0070] where d A (t) is the degree of air fouling at time t, and the degree of air fouling is associated with the air quality (AQ). The integral of the air flow f F (t) multiplied by d A (t) is equal to the total number of particles passing through the filter from time t0 to t N . The constant k represents the ability of the filter to capture particles. As shown in Chart 1502, the parameter k is in the "filter model" box and will be learned based on the AQ database stored in the data storage 212 or accessed via the network 106. Additionally, the relationship between each air quality sensor data (such as particulate matter 10um (PM10), PM2.5, pollen, etc.) and the combined air fouling degree (d A (t)) can be identified via a machine learning process. Therefore, the data in the air quality database is combined with the historical fouling degree and flow sensor data within a predetermined historical period to generate a filter model.

[0071] Steering Figure 15 , Chart 1502 shows an example of the daily air quality from several sampling points in a predetermined geographical area according to an exemplary embodiment. Given an air quality metric, such as Figure 15 the daily PM10 example in, the rate of accumulation of dirtiness can be estimated. In the current example, the daily PM10 data from several locations and the historical sensor data (DP and flow) can be used to build a filter model. In this case, given the filter type and typical HVAC usage, in Chart 1602, the algorithm estimates that the filter dirtiness increases from 0% to 100% within 4 months ( Figure 16 Chart showing the estimated dirtiness based on data according to an exemplary embodiment).

[0072] In Figure 17 , Chart 1702 showing the optimal filter replacement date derived from the processed data is shown according to an exemplary embodiment. As the filter gets dirtier, the intake fan (i.e., centrifugal fan 318) will waste more energy to provide the same airflow. Changing the filter more frequently will reduce fan energy waste, but requires more labor and hardware costs 1706. There is an optimal filter replacement frequency to minimize the total cost 1708. As shown in Chart 1702, the vertical line 1704 at time T f is the optimal filter replacement time.

[0073] To estimate the energy saving potential by changing the filter, the fan speed is recorded just after installing a new filter. The drive efficiency ratio R is defined as:

[0074] R = fan speed / flow

[0075] For example, at the current dirtiness level, the R of the filter is R1 (i.e., Figure 3 the air filter 320) is 4.2. Just after installing the filter, its ratio R2 is 3.39. Then the estimated energy saving ratio K is

[0076] K = (R2 / R1) 3

[0077] In the current example, K = 1.22, i.e., there is a 22% energy waste due to the dirty filter. Given the energy price and the estimated intake fan energy consumption, we can estimate the energy saving by changing the filter. The curve is shown as the energy cost curve 1710.

[0078] Steering Figure 18 , according to an exemplary embodiment of the present invention shows as Figure 1Schematic diagram 1800 of a water (liquid) flow system that is part of the building automation system 110 of building 108. Water 1802 from the water tower enters building 108 and passes through valve 1804. Pressure sensor 1808 measures the flow or pressure of the water entering water filter 1806. The flow or pressure of the water leaving water filter 1806 is measured by pressure sensor 1810. It should be understood that the terms pressure sensor and flow sensor are used interchangeably. Pressure drop DP1 1812 is determined and sampled regularly. Since pressure difference DP1 1812 is a sub - controller connected to building controller 112, the resulting data is sent to the data storage area for further processing. In the current embodiment, water filter 1806 can be a pleated paper filter. In other embodiments, other types of filters and filter materials, such as fiberglass or sand, can be used.

[0079] Water from water filter 1806 passes through water coolers 1814 and 1816. Pump 1818 increases the pressure or flow of the water that passes through pressure sensor 1824 and then through water filter 1820 and pressure sensor 1822. Pressure difference DP2 sub - controller 1826 determines the pressure drop across the air filter and regularly provides data to the data storage area for further processing. Additional data, such as current consumption and fan speed, is generated by sensors associated with the centrifugal fan and is also stored in the data storage area. Return water 1830 passes through pump 1828 through valve 1836, and return water 1838 leaves the building for the cooling tower. By processing the sensor data from water flow system 1800, a method similar to that used to determine the optimal filter replacement using air flow system 300 can be used.

[0080] In Figure 19In FIG. 1900, a flowchart of a method for determining when a filter needs to be replaced is shown according to an exemplary embodiment of the present invention. In a building automation system 110, sensor data is collected in step 1902 from sensors such as 324 and 322 at a sub - controller DP2326 associated with flow or pressure drop across the filter. In step 1904, the collected sensor data is processed by a main building controller 104, and in the current embodiment, a numerical regression method (such as least - squares method, robust least - squares method, decision tree, support vector machine regression) is employed to reduce the interference in the data. In step 1906, further filtering of the data results in a second filtered data set. In step 1908, a dirt level curve is generated. In step 1910, edge - detection filtering such as using a Sobel filter is applied to the dirt level curve to identify the peak filter replacement day and obtain edge - detection - filtered data. In step 1912, the peaks are segmented into clusters and a threshold for filter replacement is determined. Then in step 1914, historical dirt level data is used together with the processed data, labor and material costs, and energy costs to determine the optimal filter replacement date.

[0081] Those skilled in the art will understand and appreciate that one or more of the processes, sub - processes, or process steps described in conjunction with Figure 19 can be performed by hardware and / or software (machine - readable instructions). If the method is performed by software, the software can reside in a software memory in a suitable electronic processing component or system (such as one or more of the functional components or modules schematically depicted in the figures).

[0082] The software in the software memory can include an ordered list of executable instructions for implementing logical functions (i.e., "logic" that can be implemented in a digital form such as digital circuits or source code or in an analog form such as analog circuits or analog energy such as analog electricity, sound, or video signals), and the software can be selectively embodied in any computer-readable medium for use by or in conjunction with an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems capable of selectively extracting and executing instructions from the instruction execution system, apparatus, or device). In the context of the present disclosure, a "computer-readable medium" is any tangible means that can contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A tangible computer-readable medium can selectively be, for example but not limited to, an electrical, magnetic, optical, electromagnetic, or semiconductor system, apparatus, or device. A more detailed but still non-exhaustive list of tangible computer-readable media will include the following: portable computer disks (magnetic), RAM (electronic), read-only memory "ROM" (electronic), erasable programmable read-only memory (EPROM or flash memory) (electronic), and portable compact disc read-only memory "CDROM" (optical). It should be noted that a tangible computer-readable medium can even be paper (punched cards or punched tapes) or another suitable medium on which instructions can be captured in electronic form and subsequently (if necessary) compiled, interpreted, or otherwise processed and stored in a computer memory.

[0083] The detailed description of one or more embodiments of the method of middleware services for an integrated building server for direct communication with equipment, panels, and points is presented herein only by way of example and not in a limiting manner. It will be recognized that the advantages of certain individual features and functions described herein can be obtained without combining with other features and functions described herein. In addition, it will be recognized that various alternatives, modifications, variations, or improvements of the embodiments disclosed above and other features and functions, or alternatives thereof, can be expected to be combined into many other different embodiments, systems, or applications. Those skilled in the art can then make alternatives, modifications, variations, or improvements that are not currently foreseen or anticipated, and these alternatives, modifications, variations, or improvements should also be included in the appended claims. Therefore, the nature and scope of any appended claims should not be limited to the description of the embodiments contained herein.

Claims

1. A method for identifying the degree of fouling of a current filter, comprising: Collecting differential pressure sensor data and flow rate data associated with a material flow passing through a physical filter; Storing the differential pressure sensor data and the flow rate data in a data storage area; Using a numerical regression method to filter a predetermined set of the differential pressure sensor data and the flow rate data to smooth the predetermined set of the differential pressure sensor data, thereby obtaining a first filtered data set; Using a numerical regression method to filter the first filtered data set to further smooth the first filtered data, thereby obtaining a second filtered data set; Applying an edge detection filter to the second filtered data set, thereby obtaining an edge detection filtered data set; And Using the edge detection filtered data set and the flow rate data to determine a replacement threshold and an optimal filter replacement date for the physical filter, wherein the physical filter is an air filter or a liquid filter.

2. The method for identifying the degree of fouling of a current filter according to claim 1, further comprising: Collecting energy sensor data associated with the material flow passing through the physical filter; Storing the energy sensor data in the data storage area; And Using the edge detection filtered data set and the energy sensor data to determine a replacement threshold and an optimal filter replacement date for the physical filter.

3. The method for identifying the degree of fouling of the current filter according to claim 1, wherein, At least one of historical outdoor air quality data or historical filter fouling degree is also used to determine the optimal filter replacement date.

4. The method for identifying the degree of fouling of the current filter according to claim 1, wherein, Applying the edge detection filter is to apply an edge detection filter selected from the group consisting of Sobel, Canny, Prewitt or Laplacian edge detection methods.

5. The method for identifying the degree of dirtiness of the current filter according to claim 1, wherein, Determining the optimal filter replacement date further includes using energy sensor data.

6. A system for identifying the degree of fouling of a current filter, comprising: A plurality of sensors coupled to a controller, the sensors collecting differential pressure sensor data and flow rate data associated with a material flow passing through a physical filter; A data storage area accessed by a building controller, the data storage area storing the differential pressure sensor data and the flow rate data; A first filtering method generates a first filtered data set by the least squares regression method, and the first filtering method is applied by the controller to a part of the differential pressure sensor data and the flow rate data included in the data storage area; A second filtering method generates a second filtered data set by the least squares regression method, and the second filtering method is applied by the controller to the first filtered data set to further smooth the first filtered data set; An edge detection filter, the edge detection filter is applied by the controller to the second filtered data set, thereby obtaining an edge detection filtered data set; and A filter replacement threshold and an optimal filter replacement date, the filter replacement threshold and the optimal filter replacement date are determined by the controller using the edge detection filtered data set and the flow rate data, wherein the physical filter is an air filter or a liquid filter.

7. The system according to claim 6, further comprising: Energy sensor data collected by an energy sensor, the energy sensor data being associated with a material flow passing through the physical filter, wherein the energy sensor data is stored in the data storage area; and Using the edge detection filtered data set and the energy sensor data to determine a replacement threshold and an optimal filter replacement date for the physical filter.

8. The system according to claim 6, wherein, The optimal filter replacement date determined by the controller also uses at least one of historical outdoor air quality data or historical filter dirtiness.

9. The system according to claim 6, wherein, The edge detection filter is selected from the group consisting of Sobel, Canny, Difference of Gaussians, Prewitt, Scharr, or Laplacian filters.

10. The system according to claim 6, wherein, Determining the optimal filter replacement date includes using a predetermined historical period of data.

Images