Indoor air state monitoring point optimal arrangement method suitable for enkf data assimilation

Abstract

The application discloses a kind of indoor air state monitoring point optimization arrangement method suitable for EnKF data assimilation, belong to building environment test field.First, build working condition sample set, in combination with engineering case demand and feasibility set candidate air state measuring point position, calculate the Spearman correlation coefficient between candidate measuring point state and boundary condition parameter variable, according to the air state simulation value of each candidate measuring point under corresponding working condition, calculate the standard deviation of the retained candidate measuring point state in each boundary condition parameter variable interval, the number m of measuring point is not less than the total number n of condition parameter variable, root calculates the change amount of measuring point state caused by unit condition parameter variable change in each boundary condition parameter variable interval, then calculate the extreme value of coefficient matrix corresponding to each candidate measuring point combination for data assimilation, the minimum number of monitoring equipment required in measurement can be determined by the application, and investment cost is reasonably controlled.

Description

Technical Field

[0001] The present invention belongs to the field of building environment testing, and in particular relates to an indoor air state monitoring point optimization arrangement method suitable for EnKF data assimilation. Background Art

[0002] Indoor air quality assessment is fundamental to optimizing the operation of air conditioning and ventilation systems. Because air quality is subject to disturbances from multiple indoor and outdoor factors and exhibits spatial and temporal heterogeneity, accurately determining the distribution of air quality parameters requires deploying numerous measuring instruments across a wide spatial area. This leads to high initial investment and maintenance costs, inconvenience for residents, and is impractical in practical projects. Ensemble Kalman Filter (EnKF) data assimilation technology effectively combines limited measurement data with multi-physics computational fluid dynamics (CFD) simulations. This allows the boundary condition parameters of the current operating condition of the numerical model to be inferred from a small amount of air quality parameter measurements, and global distribution information of the multi-physics field can then be obtained through CFD. While this method has proven effective in building environmental simulation and evaluation, its performance is significantly impacted by the properties of the measurement data. Environmental condition monitoring points in practical projects are often arranged based on area indicators or empirical experience, failing to provide targeted and useful measurement data for EnKF data assimilation. When using EnKF data assimilation, finding suitable condition monitoring points is often a trial-and-error process, which wastes significant time and computational resources. Summary of the Invention

[0003] Based on the above shortcomings, the present invention provides an indoor air condition monitoring point optimization layout method suitable for EnKF data assimilation, which solves the problem of poor data assimilation effect caused by improper measurement point layout.

[0004] The technology adopted by the present invention is as follows: a method for optimizing the layout of indoor air condition monitoring points suitable for EnKF data assimilation, the steps are as follows:

[0005] Step 1: Adaptively construct a set of working condition samples. First, construct an original basic working condition sample set through the extreme value combination of each system boundary condition variable. Perform CFD simulation on the working condition samples in the basic working condition sample set to obtain the corresponding physical field simulation result set. Then, linearly interpolate the adjacent working condition samples in the physical field simulation result set to create a median parameter variable working condition, and at the same time, linearly interpolate to estimate the indoor physical field under the median parameter variable. Compare the interpolated estimated physical field with the physical field obtained by CFD simulation under the same parameter variable, and calculate the error of the linear interpolation of the physical field. If the error is greater than the set limit value, the median parameter variable working condition is added to the physical field simulation result set, and the physical field CFD simulation result corresponding to the working condition is added to the physical field simulation result set. If the error is less than the limit value, no supplement is required, and so on, until the error between the linear interpolation result and the CFD simulation result of the physical field of the median parameter variable working condition of the adjacent samples is less than the limit value. At this time, the basic working condition sample set and its corresponding physical field simulation result set are completed.

[0006] Step 2: Based on the requirements and feasibility of the project case, the candidate air state measurement point locations are set. A total of m candidate measurement points are set at equal intervals in the building space. The air state simulation value of each candidate measurement point under the corresponding working condition is derived from the physical field CFD simulation results of each working condition sample.

[0007] Step 3: Based on the basic working condition sample set and the air state simulation value of each candidate measuring point under the corresponding working condition, calculate the Spearman correlation coefficient between the candidate measuring point state and the boundary condition parameter variable, so as to analyze the monotonicity of the candidate measuring point state within the boundary condition parameter variable interval. Represents the air state at the measuring point j, j = 1, 2, 3, ..., m; for any s i If the condition parameter variable with uncertainty in the system boundary condition is s, then the j measurement point is retained, otherwise the j measurement point is eliminated from the candidate measurement points. i , i=1,2,3,…,n, the variable vector is s=[s1 s2 … s i … s n ] T , where n is the total number of conditional parameter variables; in the EnKF algorithm, the key EnKF filter equation is shown in formula (1),

[0008]

[0009] Where s' represents the conditional parameter variable vector after data assimilation, is the state vector of the monitoring point, is the covariance between the conditional parameter variable and the state prediction value of the monitoring point, is the covariance between the state prediction values ​​of each monitoring point, Y is the measurement value vector, M represents the projection matrix of the measurement value vector to the prediction value vector, C ∈∈ is the measurement error covariance matrix;

[0010] To make the solution of s' unique, when the boundary condition parameter variables take different values, the measurement point states used for EnKF data assimilation are not equal. Therefore, within the range of variation of the boundary condition parameter variables, the measurement point states should be monotonic, that is:

[0011]

[0012] Step 4: Based on the simulated values ​​of the air state at each candidate measuring point under the corresponding working conditions, calculate the standard deviation of the state of the candidate measuring point retained within the variable interval of each boundary condition parameter, and compare it with the standard deviation of the instrument measurement error. If If it holds, then the jth measuring point is retained, otherwise it is eliminated from the candidate measuring points, σ j is the standard deviation of the air state error at the measuring point j; when there is an error in the measurement, C ∈∈ ≠0, that is, the true value of the air state exists within the error range of the measured value. To ensure the stability of the calculation, when the difference between the simulated value and the measured value is less than the measurement error, the simulation result is considered to be close to the actual working condition. Therefore, within the range of the boundary condition parameter variable, if the change in the state of the measuring point is always less than the measurement error, it cannot be used for EnKF data assimilation, that is:

[0013]

[0014] Step 5: The number of measurement points m required for EnKF data assimilation is not less than the total number of conditional parameter variables n. Take m = n, and group the candidate measurement points retained in step 4 into groups of m. Let the number of retained candidate measurement points be d, then the total number of candidate measurement points is d. According to the simulated values ​​of the air state of each candidate measuring point under the corresponding working conditions, the Pearson correlation coefficient between the states of each candidate measuring point in each combination under the working condition change is calculated to analyze the linear correlation between the states of the candidate measuring points. If any two measurement points in a combination are established, the candidate measurement point combination is retained, otherwise the candidate measurement point combination is eliminated; when the measurement is absolutely accurate, C ∈∈ =0, in order to make formula (1) meaningful, then:

[0015]

[0016]

[0017]

[0018] Where, represents the covariance between the state prediction values ​​of measurement point a and measurement point b, where a, b = 1, 2, 3, …, m, where m is the total number of measurement points; represents the Pearson correlation coefficient between the state prediction values ​​of measuring point a and measuring point b. According to formula (6), the states of any two measuring points used for EnKF data assimilation need to be nonlinearly correlated.

[0019] Step 6: Calculate the state change β of the measuring point caused by the change of the unit condition parameter variable within each boundary condition parameter variable interval based on the basic working condition sample set and the air state simulation value of each candidate measuring point under the corresponding working condition. i,j , and then calculate the extreme values ​​of the coefficient matrix A corresponding to each candidate measurement point combination for data assimilation, that is, the minimum value min(|A|) and the maximum value max(|A|). If min(|A|)·max(|A|)>0, then the measurement point combination is retained, otherwise the measurement point combination is eliminated. Among them, the retained measurement point combinations are all determined to be measurement point combinations suitable for EnKF data assimilation. The number of measurement points, measurement point locations, and measured state parameter objects contained in each combination are output in sequence to obtain a monitoring point layout plan suitable for EnKF data assimilation, including the measured state parameters, measurement point number, and measurement point locations; Among them, for any boundary condition parameter variable:

[0020]

[0021]

[0022]

[0023]

[0024]

[0025] β i =[β i,1 β i,2 … β i,j … β i,m ] (12)

[0026]

[0027]

[0028] Let A be the coefficient matrix of the above equations, that is:

[0029]

[0030] To make Δs i To have a unique solution, the rank of the coefficient matrix of the system of equations must be equal to the total number of conditional parameter variables, that is:

[0031] rank(A)=n (16)

[0032] Then m≥n. When m=n, to make rank(A)=n hold, then:

[0033]

[0034] Where, α i is the conditional parameter variable s i The ratio vector of the state prediction error caused by relative to the total state prediction error, Indicates that the conditional parameter variable s i In the case of a single change i The covariance between the predicted value of the monitoring point state and For the conditional parameter variable s i The covariance between the state prediction values ​​of each monitoring point under the condition of individual changes, Δs i is the conditional parameter variable s i The correction amount, β i For the variable s i The change vector of the air state at each monitoring point under each unit change of , β i,j The air state at the measurement point j is represented by the variable s i The amount of change per unit change in Y j is the measured value at measuring point j.

[0035] Furthermore, the present invention provides a system for optimizing the layout of indoor air status monitoring points suitable for EnKF data assimilation, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The processor processes and executes the computer program to implement the above-mentioned method for optimizing the layout of indoor air status monitoring points suitable for EnKF data assimilation, and outputs a monitoring point layout plan suitable for EnKF data assimilation.

[0036] Advantages and benefits of the present invention: This invention establishes an effective measurement scheme for indoor multi-physics field EnKF data assimilation, ensuring a unique solution for the boundary condition parameters derived from the assimilation of measured data with numerical simulations. The corresponding multi-physics field simulation results are highly accurate. This invention can determine the minimum number of monitoring devices required for measurement, effectively controlling investment costs. It also avoids arduous manual trial-and-error work and conserves significant computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1Flowchart of the technology for optimizing the placement of indoor air condition monitoring points for EnKF data assimilation

[0038] Figure 2 Adaptive construction working condition sample diagram,

[0039] (a) Schematic diagram of the sample working condition for adaptive construction, (b) flow chart of the sample working condition for adaptive construction;

[0040] Figure 3 This is a scale model diagram of the architectural space;

[0041] Figure 4 is the distribution map of candidate measuring points,

[0042] Among them, (a) the number map of candidate temperature measurement points, (b) the number map of wind speed measurement points;

[0043] Figure 5 Screening result graphs for measurement plan optimization;

[0044] Figure 6 This is a comparison chart of wind speed field simulation errors;

[0045] Figure 7 This is a comparison chart of temperature field simulation errors; DETAILED DESCRIPTION

[0046] The system boundary conditions act directly on the spatial multi-physics field distribution. The various factors in the boundary conditions have different degrees of influence on the air state at different spatial points. The air environment formed under different working conditions may have the same state at some locations. If these locations are used as air state monitoring points for EnKF data assimilation, it is obviously impossible to obtain unique and accurate boundary condition parameters. Therefore, the key to optimizing the monitoring scheme lies in analyzing the correlation between the air state at each location and the various factors in the system boundary conditions. Starting from the basic principles of the EnKF algorithm, the present invention derives the basic principles that the measurement scheme should meet, and creates a method for optimizing the layout of indoor air state monitoring points suitable for EnKF data assimilation. The present invention is further explained in conjunction with the accompanying drawings and specific embodiments.

[0047] Example 1

[0048] In order to achieve the optimal arrangement of indoor air status monitoring points, such as Figure 1 As shown in the figure, a method for optimizing the layout of indoor air condition monitoring points suitable for EnKF data assimilation is provided. The steps are as follows:

[0049] Step 1: Figure 2As shown, the working condition sample set is adaptively constructed. First, the original basic working condition sample set is constructed by combining the extreme values ​​of the boundary condition variables of each system. The working condition samples in the basic working condition sample set are subjected to CFD simulation to obtain the corresponding physical field simulation result set. Then, the median parameter variable working condition is created by linear interpolation of adjacent working condition samples in the physical field simulation result set, and the indoor physical field under the median parameter variable is estimated by linear interpolation at the same time. The physical field estimated by this interpolation is compared with the physical field obtained by CFD simulation under the same parameter variable, and the error of the linear interpolation of the physical field is calculated. If the error is greater than the set limit value, the median parameter variable working condition is added to the physical field simulation result set, and the physical field CFD simulation result corresponding to the working condition is added to the physical field simulation result set. If the error is less than the limit value, no supplement is required, and so on, until the error between the linear interpolation result and the CFD simulation result of the physical field of the median parameter variable working condition of the adjacent samples is less than the limit value. At this time, the basic working condition sample set and its corresponding physical field simulation result set are completed.

[0050] Step 2: Based on the requirements and feasibility of the project case, the candidate air state measurement point locations are set. A total of m candidate measurement points are set at equal intervals in the building space. The air state simulation value of each candidate measurement point under the corresponding working condition is derived from the physical field CFD simulation results of each working condition sample.

[0051] Step 3: Based on the basic working condition sample set and the air state simulation value of each candidate measuring point under the corresponding working condition, calculate the Spearman correlation coefficient between the candidate measuring point state and the boundary condition parameter variable, so as to analyze the monotonicity of the candidate measuring point state within the boundary condition parameter variable interval. Represents the air state at the measuring point j, j = 1, 2, 3, ..., m, for any s i If the condition parameter variable with uncertainty in the system boundary condition is s, then the j measurement point is retained, otherwise the j measurement point is eliminated from the candidate measurement points. i , i=1,2,3,…,n, the variable vector is s=[s1 s2 … s i … s n ] T , where n is the total number of conditional parameter variables;

[0052] In the EnKF algorithm, the key EnKF filter equation is shown in formula (1):

[0053]

[0054] Where s' represents the conditional parameter variable vector after data assimilation, is the state vector of the monitoring point, is the covariance between the conditional parameter variable and the state prediction value of the monitoring point, is the covariance between the state prediction values ​​of each monitoring point, Y is the measurement value vector, M represents the projection matrix of the measurement value vector to the prediction value vector, C ∈∈ is the measurement error covariance matrix;

[0055] To make the solution of s' unique, when the boundary condition parameter variables take different values, the measurement point states used for EnKF data assimilation are not equal. Therefore, within the range of variation of the boundary condition parameter variables, the measurement point states should be monotonic, that is:

[0056]

[0057] Step 4: Based on the simulated values ​​of the air state at each candidate measuring point under the corresponding working conditions, calculate the standard deviation of the state of the candidate measuring point retained within the variable interval of each boundary condition parameter, and compare it with the standard deviation of the instrument measurement error. If If it holds, then the jth measuring point is retained, otherwise it is eliminated from the candidate measuring points, σ j is the standard deviation of the air state error at the measuring point j; when there is an error in the measurement, C ∈∈ ≠0, that is, the true value of the air state exists within the error range of the measured value. To ensure the stability of the calculation, when the difference between the simulated value and the measured value is less than the measurement error, the simulation result is considered to be close to the actual working condition. Therefore, within the range of the boundary condition parameter variable, if the change in the state of the measuring point is always less than the measurement error, it cannot be used for EnKF data assimilation, that is:

[0058]

[0059] Step 5: Since the number of measurement points m required for EnKF data assimilation is not less than the total number of conditional parameter variables n, m=n is taken. The candidate measurement points retained in step 4 are grouped into groups of m. If the number of candidate measurement points retained is d, the total number of candidate measurement points is d. According to the simulated values ​​of the air state of each candidate measuring point under the corresponding working conditions, the Pearson correlation coefficient between the states of each candidate measuring point in each combination under the working condition change is calculated to analyze the linear correlation between the states of the candidate measuring points. If any two measurement points in a combination are established, the candidate measurement point combination is retained, otherwise the candidate measurement point combination is eliminated; when the measurement is absolutely accurate, C ∈∈ =0, in order to make formula (1) meaningful, then:

[0060]

[0061]

[0062]

[0063] Where, represents the covariance between the state prediction values ​​of measurement point a and measurement point b, where a, b = 1, 2, 3, …, m, where m is the total number of measurement points; represents the Pearson correlation coefficient between the state prediction values ​​of measuring point a and measuring point b. According to formula (6), the states of any two measuring points used for EnKF data assimilation need to be nonlinearly correlated.

[0064] Step 6: Calculate the state change β of the measuring point caused by the change of the unit condition parameter variable within each boundary condition parameter variable interval based on the basic working condition sample set and the air state simulation value of each candidate measuring point under the corresponding working condition. i,j , and then calculate the extreme values ​​of the coefficient matrix A corresponding to each candidate measurement point combination for data assimilation, that is, the minimum value min(|A|) and the maximum value max(|A|). If min(|A|)·max(|A|)>0, then the measurement point combination is retained, otherwise the measurement point combination is eliminated. Among them, the retained measurement point combinations are all determined to be measurement point combinations suitable for EnKF data assimilation. The number of measurement points, measurement point locations, and measured state parameter objects contained in each combination are output in sequence to obtain a monitoring point layout plan suitable for EnKF data assimilation, including the measured state parameters, measurement point number, and measurement point locations; Among them, for any boundary condition parameter variable:

[0065]

[0066]

[0067]

[0068]

[0069]

[0070] β i =[β i,1 β i,2 … β i,j … β i,m ] (12)

[0071]

[0072]

[0073] Let A be the coefficient matrix of the above equations, that is:

[0074]

[0075] To make Δs i To have a unique solution, the rank of the coefficient matrix of the system of equations must be equal to the total number of conditional parameter variables, that is:

[0076] rank(A)=n (16)

[0077] Then m≥n. When m=n, to make rank(A)=n hold, then:

[0078]

[0079] Where, α i is the conditional parameter variable s i The ratio vector of the state prediction error caused by relative to the total state prediction error, Indicates that the conditional parameter variable s i In the case of a single change i The covariance between the predicted value of the monitoring point state and For the conditional parameter variable s i The covariance between the state prediction values ​​of each monitoring point under the condition of individual changes, Δs i is the conditional parameter variable s i The correction amount, β i For the variable s i The change vector of the air state at each monitoring point under each unit change of , β i,j The air state at the measurement point j is represented by variable s i The amount of change per unit change in Y j is the measured value at measuring point j.

[0080] Therefore, the measurement scheme used for EnKF data assimilation must meet the following five principles:

[0081] (1) The states of any two measurement points used for EnKF data assimilation are nonlinearly correlated.

[0082] (2) Within the range of variation of boundary condition parameter variables, the standard deviation of the variation of the measuring point state is greater than the standard deviation of the measurement error.

[0083] (3) Within the range of variation of each boundary condition parameter variable, the state of the measuring point should be monotonic, that is:

[0084] (4) The number of measuring points is not less than the total number of conditional parameter variables, m ≥ n;

[0085] (5) When m=n, the determinant of the coefficient matrix consisting of the measurement state changes caused by the changes in the unit boundary condition parameter variables is not 0,

[0086] Example 2

[0087] This embodiment discloses a system for optimizing the layout of indoor air status monitoring points suitable for EnKF data assimilation, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The processor processes and executes the computer program to implement the method for optimizing the layout of indoor air status monitoring points suitable for EnKF data assimilation as described in Example 1, and outputs a monitoring point layout plan suitable for EnKF data assimilation.

[0088] Example 3

[0089] like Figure 3 The figure shows a scaled model of an ideal building space. The dimensions of this small room are length × width × height = 1m × 0.75m × 0.5m. Two air-conditioning supply vents 1 are located on the ceiling, coordinating the air supply to the room. Return air vents 2 are symmetrically located below the walls on either side. The indoor heat source is evenly distributed on the floor, with no heat transfer to other building walls. The system components of this building's indoor thermal environment contain three uncertain boundary condition parameter variables: air-conditioning supply air velocity, supply air temperature, and floor heat dissipation. The corresponding operating ranges are shown in Table 1.

[0090] Table 1. Numerical intervals of boundary condition parameter variables

[0091]

[0092]

[0093] The goal of this embodiment is to analyze the indoor air temperature and flow rate distribution. It is known that the candidate measurement points of these two air states are evenly distributed in the length, width and height directions of the space. Figure 4 Shown are the coordinates and numbers of candidate measurement points for temperature and wind speed.

[0094] According to the extreme values ​​of the boundary condition parameter variables in Table 1, 8 basic working condition samples are established to form the original sample set. The boundary condition parameter variable values ​​corresponding to each sample are shown in Table 2. The working condition samples in the set are subjected to CFD simulation to obtain a set of corresponding physical field (airflow field and temperature field) simulation results. Then, the working condition samples in the above set are linearly interpolated to create a median parameter variable working condition, and at the same time, the indoor physical field under the median parameter variable is linearly interpolated to estimate. For example: the linear interpolation of working condition 1 and working condition 2 in Table 2 gives a supply air speed of 0.4m / s, a supply air temperature of 14°C, and a ground heat flux density of 30W / m 2The median parameter variable working condition is used, and the airflow field and temperature field obtained from the CFD simulation of working condition 1 and working condition 2 are linearly interpolated to estimate the airflow field and temperature field under the median parameter variable working condition. The physical field estimated by this interpolation is compared with the physical field obtained by CFD simulation under the corresponding median parameter variable working condition, and the error of the linear interpolation is calculated. If the error between the interpolation result of the airflow field or temperature field and the CFD simulation result is greater than the set limit value (the relative error limit value set in this case is 3%), then the median parameter variable working condition is added to the above working condition sample set, and the physical field CFD simulation result corresponding to the working condition is added to the physical field set. If the error is less than the limit value, no supplement is required. This process is repeated until the error between the linear interpolation result and the CFD simulation result of the physical field of the median parameter variable working condition of adjacent samples is less than the limit value. At this time, the working condition sample set and its corresponding physical field CFD simulation result set are completed. After calculation, the final working condition sample set of this embodiment has a total of 25 samples, as shown in Table 3.

[0095] Table 2 Original basic working condition sample set

[0096]

[0097] Table 3 Adaptively generated working condition sample set

[0098]

[0099] Based on the locations of the aforementioned candidate measuring points, the simulated values ​​of the air temperature and flow rate at each candidate measuring point under various working conditions are derived from the physical field CFD simulation results of the above working condition samples. On this basis, the Spearman correlation coefficient between the candidate measuring point state (temperature, flow rate) and the boundary condition parameter variables is calculated and a conditional judgment is performed. If the Spearman correlation coefficient between the candidate measuring point temperature or flow rate and the boundary condition parameter variable is 1, the temperature or wind speed candidate measuring point is retained; otherwise, the temperature or wind speed measuring point is eliminated from the candidate measuring points.

[0100] Next, based on the simulation results of the above working condition samples, the standard deviation of the candidate measurement point states retained in the previous step is calculated within the variable intervals of each boundary condition parameter, and compared with the standard deviation of the instrument measurement error. Within the variable intervals of each boundary condition parameter, if the standard deviation of the air temperature change of the candidate temperature measurement point is greater than the standard deviation of the temperature sensor measurement error (0.1°C in this case), then the candidate temperature measurement point is retained; otherwise, the temperature measurement point is eliminated. Similarly, if the standard deviation of the air flow velocity change of the candidate wind speed measurement point is greater than the standard deviation of the wind speed sensor measurement error (0.01m / s in this case), then the candidate wind speed measurement point is retained; otherwise, the wind speed measurement point is eliminated.

[0101] Since there are three boundary condition parameter variables in this embodiment, EnKF data assimilation requires at least three measurement points. The candidate temperature or wind speed measurement points retained in the previous step are grouped into groups of three. Based on the simulation results of the operating condition samples, the Pearson correlation coefficient between the states of the candidate measurement points in each combination is calculated when the operating condition changes. If the Pearson correlation coefficient between the air states of any two measurement points in a candidate measurement point combination (the air temperature at the temperature measurement point or the air velocity at the wind speed measurement point) is not 1, meaning that the air states of the two measurement points are nonlinearly correlated, then the candidate measurement point combination is retained; otherwise, it is discarded.

[0102] Based on the simulation results of the operating condition samples, the change in the measurement point state caused by the change in the unit condition parameter variable within each boundary condition parameter variable interval is calculated. Then, the extreme values ​​of the coefficient matrix A corresponding to each candidate measurement point combination for data assimilation are calculated, namely the minimum value min(A) and the maximum value max(A). If min(A)·max(A)>0, the measurement point combination is determined to be suitable for EnKF data assimilation. The information contained in this combination, such as the number of measurement points, measurement point locations, and measured state parameters, is sequentially output as the optimized measurement plan.

[0103] Figure 5 The following table shows the measurement point combination suitable for EnKF data assimilation after optimization screening. The comparison of the simulation errors after the non-optimized measurement scheme and the optimized measurement scheme were used for EnKF data assimilation of the wind speed and temperature field simulation in the small room is shown in Figure 2. Figure 6 and Figure 7 Therefore, the optimization arrangement method proposed in Example 1 can significantly improve the EnKF data assimilation accuracy.

Claims

1. A method for optimizing the placement of indoor air condition monitoring points suitable for EnKF data assimilation, characterized in that: Here are the steps: Step 1: Adaptively construct a working condition sample set. First, construct an original basic working condition sample set through the extreme value combination of each system boundary condition parameter variable, perform CFD simulation on the working condition samples in the basic working condition sample set to obtain the corresponding physical field CFD simulation result set, and then linearly interpolate the adjacent working condition samples in the basic working condition sample set to create a median parameter variable working condition, and at the same time linearly interpolate to estimate the indoor physical field under the median parameter variable working condition, compare the interpolated estimated physical field with the physical field obtained by CFD simulation under the same parameter variable, and calculate the error of the linear interpolation of the physical field. If the error is greater than the set limit value, the median parameter variable working condition is added to the basic working condition sample set, and the physical field CFD simulation result corresponding to the median parameter variable working condition is added to the physical field CFD simulation result set. If the error is less than the limit value, no supplement is required, and so on, until the error between the linear interpolation result of the physical field under the median parameter variable working condition of the adjacent samples and the physical field CFD simulation result is less than the limit value. At this time, the basic working condition sample set and its corresponding physical field CFD simulation result set are completed. Step 2: Based on the requirements and feasibility of the project case, the candidate air state simulation value measurement point locations are set. A total of m candidate monitoring points are set at equal intervals in the building space. The air state simulation value of each candidate monitoring point under the corresponding working condition is derived from the physical field CFD simulation results of each working condition sample. Step 3: Based on the basic working condition sample set and the air state simulation value of each candidate monitoring point under the corresponding working condition, calculate the Spearman correlation coefficient between the candidate monitoring point state and the boundary condition parameter variable, so as to analyze the monotonicity of the candidate monitoring point state within the boundary condition parameter variable interval. , Represents the simulated value of the air state at monitoring point j, j = 1, 2, 3, ..., m; for any If the condition is established, then the j monitoring point is retained, otherwise the j monitoring point is eliminated from the candidate monitoring points; where, the boundary condition parameter variable with uncertainty in the system boundary condition is s i , i=1,2,3,…,n, the variable vector is , where n is the total number of boundary condition parameter variables; in the EnKF algorithm, the key EnKF filter equation is shown in formula (1), （1） Where s' represents the boundary condition parameter variable vector after data assimilation, is the state simulation value vector at the monitoring point, is the covariance matrix between the boundary condition parameter variables and the state simulation values ​​at the monitoring points, is the covariance matrix between the state simulation values ​​at each monitoring point, Y is the measurement value vector, M represents the projection matrix of the measurement value vector to the state simulation value vector, is the measurement error covariance matrix; To make The solution of is unique. When the boundary condition parameter variables take different values, the monitoring point states used for EnKF data assimilation do not have equal conditions. Therefore, within the range of variation of the boundary condition parameter variables, the monitoring point states should be monotonic, that is: （2） Step 4: Based on the simulated values ​​of the air state at each candidate monitoring point under the corresponding working conditions, calculate the standard deviation of the state of the candidate monitoring point retained within the variable interval of each boundary condition parameter, and compare it with the standard deviation of the instrument measurement error. If it is established, then the j monitoring point is retained, otherwise the j monitoring point is eliminated from the candidate monitoring points. is the standard deviation of the air state simulation value error at monitoring point j; when there is an error in the measurement, , that is, the true value of the air state exists within the error range of the measured value; in order to ensure the stability of the calculation, when the difference between the simulated value and the measured value is less than the measurement error, the simulation result is considered to be close to the actual working condition; therefore, within the range of the boundary condition parameter variable, if the change in the state of the monitoring point is always less than the measurement error, it cannot be used for EnKF data assimilation, that is: （3） Step 5: The number of monitoring points m required for EnKF data assimilation is not less than the total number of boundary condition parameter variables n. Take m=n, and combine the candidate monitoring points retained in step 4 into groups of m. Let the number of candidate monitoring points retained be d, then the total number of monitoring points is d. According to the simulated values ​​of the air state of each candidate monitoring point under the corresponding working conditions, the Pearson correlation coefficient between the states of each candidate monitoring point in each combination under the working condition change is calculated to analyze the linear correlation between the states of the candidate monitoring points. If any two monitoring points in a combination are true, the candidate monitoring point combination is retained, otherwise the candidate monitoring point combination is eliminated; when the measurement is absolutely accurate, , in order to make formula (1) meaningful, then: （4） （5） （6） Where, represents the covariance matrix between the state simulation values ​​of monitoring point a and monitoring point b, where a, b = 1, 2, 3, …, m, and m is the total number of monitoring points; represents the Pearson correlation coefficient between the state simulation values ​​of monitoring point a and monitoring point b. According to formula (6), the states of any two monitoring points used for EnKF data assimilation need to be nonlinearly correlated. Step 6: Calculate the state change of the monitoring point caused by the change of the unit boundary condition parameter variable within each boundary condition parameter variable interval based on the basic working condition sample set and the air state simulation value of each candidate monitoring point under the corresponding working condition. , and then calculate the extreme value of the coefficient matrix A corresponding to each candidate monitoring point combination for data assimilation, that is, the minimum value and maximum value ,like , then the monitoring point combination is retained, otherwise the monitoring point combination is eliminated. The retained monitoring point combinations are all determined to be monitoring point combinations suitable for EnKF data assimilation. The number of monitoring points, monitoring point locations and measured state parameter objects contained in each combination are output in sequence to obtain a monitoring point layout scheme suitable for EnKF data assimilation, including measured state parameters, number of monitoring points and monitoring point locations. For any boundary condition parameter variable: （7） （8） （9） （10） （11） （12） （13） （14） Let A be the coefficient matrix of the above equations, that is: （15） To make To have a unique solution, the rank of the coefficient matrix of the system of equations must be equal to the total number of boundary condition parameter variables, that is: （16） Then m≥n, when m=n, we need to make If established, then: （17） Where, is the boundary condition parameter variable s i The ratio vector of the state simulation value error caused by to the total state simulation value error, Indicates the boundary condition parameter variable s i In the case of a single change i The covariance matrix between the simulated value of the monitoring point state, is the parameter variable s under the boundary condition i The covariance matrix between the state simulation values ​​of each monitoring point under the condition of individual changes, is the boundary condition parameter variable s i The correction amount, For the variable s i The change vector of the air state at each monitoring point under each unit change of The air state at the measurement point j is represented by variable s i The amount of change per unit change in Y j is the measured value at monitoring point j.

2. A system for optimizing the placement of indoor air condition monitoring points suitable for EnKF data assimilation, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The processor executes the computer program to implement the indoor air condition monitoring point optimization layout method suitable for EnKF data assimilation according to claim 1, and outputs a monitoring point layout plan suitable for EnKF data assimilation.

Images