CN109583008B - General modeling method for water chiller energy efficiency model based on stepwise regression - Google Patents
General modeling method for water chiller energy efficiency model based on stepwise regression Download PDFInfo
- Publication number
- CN109583008B CN109583008B CN201811212546.8A CN201811212546A CN109583008B CN 109583008 B CN109583008 B CN 109583008B CN 201811212546 A CN201811212546 A CN 201811212546A CN 109583008 B CN109583008 B CN 109583008B
- Authority
- CN
- China
- Prior art keywords
- model
- data
- energy efficiency
- parameters
- measurable
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Geometry (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Air Conditioning Control Device (AREA)
Abstract
The invention discloses a general modeling method for a water chilling unit energy efficiency model based on stepwise regression, and because measurable parameters affecting the water chilling unit energy efficiency can be different in different actual projects, the invention provides an optimal energy efficiency model (SR) meeting different operation conditions of the current unit by adopting a stepwise regression method to perform optimization screening on model factor items on the basis of programming a model factor item set representing the unit energy efficiency. And because the improvement of the data quality is the basis for ensuring the precision of the simulation model, the invention provides a multi-index fusion wavelet denoising method for effectively removing noise components concentrated on a high-frequency part of a signal (namely the monitoring data of measurable parameters). The method creatively solves the defects of low simulation precision and limited application range of the traditional black/gray box model, and lays a foundation for optimizing the unit operation strategy and diagnosing the energy conservation of the water chilling unit.
Description
Technical Field
The invention relates to a method for constructing an optimal energy efficiency model of a water chilling unit by using on-site measurable influence parameters, in particular to a general modeling method for the energy efficiency model of the water chilling unit based on stepwise regression.
Background
For different functional types of buildings, the energy consumption of the water chilling unit accounts for more than 40% of the energy consumption of the air conditioning system, and the energy consumption of the water chilling unit accounts for about 20-37% of the total energy consumption of the building. As the most important energy consumption part of the air conditioning system, it is important to realize energy conservation diagnosis of the water chilling unit. At present, the performance of the water chilling unit is evaluated by adopting simulation, and the actual operation characteristics of the unit are reflected through a reasonable and accurate energy efficiency model, so that a foundation is laid for optimizing the operation strategy of the unit. In general, the water chiller can simulate model development mainly comprises three parts of parameter analysis influence, modeling method selection and experimental data processing.
Although the performance influence parameters of the water chilling unit are discussed more or less in the prior researches, the measurable influence parameters are different for different actual projects. Therefore, how to reasonably and accurately classify the influence parameters and select the influence parameters suitable for the current actual engineering to construct a water chilling unit performance simulation model is still a problem to be solved. At present, simulation models constructed by means of performance influence parameters of a water chiller and different modeling methods are mainly divided into a mechanism model, an ash box model, a black box model and an artificial intelligent model. The gray box model and the black box model can lead the research focus to be transferred from the internal characteristics of the concerned model and the complex thermodynamic equation to directly consider the output performance of the system, and the performance simulation of various working conditions of the water chilling unit is very simple and quick and is very suitable for site engineers, so that the construction of the black/gray box model of the water chilling unit by using the data driving method is the best choice for researching the running performance of the water chilling unit. However, for a certain gray box/black box model, if input parameters required by the model cannot be completely measured in actual engineering or a unit runs under a working condition that the model is not applicable, the current model is not available or the accuracy is low. Therefore, how to construct performance simulation models suitable for different water chilling units and operation conditions is particularly important. In the practical application process, the black box or ash box model needs a large amount of samples or measured data reflecting the performance characteristics of the chiller to solve model coefficients, the applicability of the model to different types of units is verified, and at the moment, the improvement of the data quality can effectively ensure the model precision. The traditional data preprocessing mainly comprises format errors, unreasonable data and transient data rejection, but because of instrument and meter precision or mechanical tool deviation, electric/magnetic field interference and circuit noise in the monitoring process, the sampling signals (i.e. the monitoring data) are inevitably polluted by noise, and the size and change rule of the noise signals cannot be predicted. Therefore, means for effectively eliminating noise signals are required to be studied on the basis of the traditional data preprocessing method.
In summary, in the conventional chiller model, there is a problem that the simulation model has low accuracy and the application range is limited due to the fixed model structure and the fixed model input parameters.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a general modeling method for a water chiller energy efficiency model based on stepwise regression, which can ensure the fitting effect and the model precision of a regression model and construct a current optimal energy efficiency model of a machine set.
A general modeling method for a water chiller energy efficiency model based on stepwise regression comprises the following steps:
determining a model factor item set of an energy efficiency model, wherein the model factor item set comprises three types of model factor items, namely a first type of model factor item is composed of secondary refrigerant side parameters, a second type of model factor item is composed of refrigerant side parameters, and a third type of model factor item is composed of compressor regulating parameters;
the first model factor item is obtained through collecting and sorting model factor items in an existing water chiller energy consumption regression model, and the second model factor item and the third model factor item are obtained through analysis of influence parameter action mechanisms;
step two, aiming at a certain actual project, collecting historical operation data of field measurable parameters of the water chilling unit;
judging whether the field measurable parameter in the step two is a composition parameter of a model factor item in the model factor item set, if so, carrying out data preprocessing on the field measurable parameter, wherein the method comprises the following specific steps of:
(1) Reading the data of each measurable parameter in a set test period by a monitoring platform connected with the water chilling unit to form a data sequence;
(2) Screening the data in the data sequence of each measurable parameter to remove non-ideal data in the data sequence to form a data sequence after screening of each measurable parameter;
(3) Performing multi-index fusion wavelet denoising on the primary screening data of each type of measurable parameter processed by the steps, determining the optimal decomposition reconstruction scale corresponding to each type of measurable parameter data by adopting a multi-index fusion comprehensive evaluation index, determining the optimal wavelet basis function corresponding to each type of measurable parameter data by using a Root Mean Square Error (RMSE) and a signal-to-noise ratio (SNR) index, constructing a reasonable threshold value for each decomposition level on the basis, processing a high-frequency coefficient by using a soft threshold value method, removing noise components concentrated on a high-frequency part, and performing wavelet reconstruction on a low-frequency coefficient and the high-frequency coefficient after the threshold value quantization to obtain a data sequence after noise removal by adopting a multi-index fusion wavelet denoising method;
wherein the steps of determining the optimal decomposition reconstruction scale are as follows:
(a) Respectively solving the root mean square error variation C by adopting the numerical values in the data sequence screened by each measurable parameter vrm Variation C of signal to noise ratio snr Smoothness variation C vr ；
(b) The root mean square error variation C of each type of measurable parameter vrm Variation C of signal to noise ratio snr Smoothness variation C vr The numerical values are weighted and fused by adopting an entropy method to obtain a comprehensive evaluation index sequence of the standard wavelet denoising effect, the comprehensive evaluation index sequence of each type of measurable parameter consists of 10 comprehensive evaluation indexes, and the calculation formula of the comprehensive evaluation indexes is as follows:
CEI(m)＝w vrm (m)·C vrm +w vsnr (m)·C vsnr +w vr (m)C vr
wherein m is a wavelet decomposition reconstruction scale and is a positive integer with a value of 1-10; w (w) vrm (m)、w vsnr (m)、w vr (m) C at m scale vrm 、C snr C vr The weight of the vehicle is occupied;
(c) Subtracting the maximum value in the sequence from each comprehensive evaluation index in the comprehensive evaluation index sequence of each type of measurable parameter respectively for inversion, carrying out 4-order fitting by using a least square method for analyzing the overall change trend of the sequence, removing abnormal values deviating from the whole, and replacing the abnormal values with fitting values corresponding to the 4-order fitting;
(d) Making a curve according to the numerical value in each comprehensive evaluation index sequence, and searching for an obvious inflection point in the curve change, wherein the decomposition reconstruction scale corresponding to the node is the optimal decomposition reconstruction scale of the primary screening data;
the method for determining the optimal wavelet base comprises the following steps: and selecting an optimal wavelet base suitable for actual running data of the water chilling unit by using a Root Mean Square Error (RMSE) and a signal-to-noise ratio (SNR) index, wherein a calculation equation is as follows:
SNR＝10×lg(power signal /power noise )
wherein f (i) is monitoring data after primary screening; m is a wavelet decomposition reconstruction scale, and the value is a positive integer of 1-10;reconstructing data for decomposition at m scale; n is the number of data, and->
The threshold lambda corresponding to each decomposition level is calculated by adopting a length logarithm threshold method:
in the middle ofFor noise estimation, N is the number of wavelet coefficients of each layer;
step four, if the field measurable parameters comprise the chilled water outlet temperature T eo Chilled water inlet temperature T ei Chilled water flow M e Then through the calculation formula Q of the refrigerating capacity e ＝C p M e (T ei -T eo ) Calculate Q e Wherein C p The specific heat is water constant pressure, the specific heat takes the value of 4.2 kJ/(K.kg), the field measurable parameter processed in the third step and the refrigerating capacity calculated by the processed measurable parameter are taken as the known parameters, the model factor items formed by the known parameters are selected from the model factor item set to construct the initial expression of the COP of the energy efficiency model of the water chilling unit, if T eo 、T ei M is as follows e And if the parameters are not field measurable parameters, constructing an initial expression of COP by taking the field measurable parameters processed in the step three as known parameters:
COP＝β 0 X 0 +β 1 X 1 +…+β l X l
wherein l is the number of factor items selected from the model factor item set; beta 0 ,β 1 ,…,β l Fitting coefficients corresponding to the factor items; x is X 0 ,X 1 ,…,X l Respectively the field measurable parameters processed in the third step;
step five, screening each model factor according to the contribution rate of each model factor independent variable in the water chiller energy efficiency model to the COP, and removing model factor with insignificant influence on the COP, so that the influence of each factor in the final regression equation on the COP is significant;
solving fitting coefficients before each model factor item by adopting multiple linear regression, multiplying each model factor after screening by a corresponding coefficient, and summing to obtain an optimal energy efficiency model of the water chiller energy efficiency model COP;
and step seven, collecting measurable parameters related to each model factor item in the COP optimal energy efficiency model under the current running condition of the water chilling unit, preprocessing data of each measurable parameter by adopting the steps (2) - (3), and substituting the measurable parameters preprocessed by the data into the COP optimal energy efficiency model to predict the running performance of the water chilling unit.
The invention has the advantages and positive effects that:
1. the model constructed by the method is essentially a black box model, is very simple and convenient to simulate the performance of each working condition of the unit, and is very suitable for field engineers.
2. For data-driven modeling, the improvement of data quality is the basis of ensuring the precision of a simulation model, and the multi-index fusion wavelet denoising method adopted by the method can effectively remove noise components concentrated on a high-frequency part of a signal (namely monitoring data).
3. Because different actual projects exist in influencing parameters of the water chilling unit energy efficiency model, the method can extract factor items combined by the measurable influencing parameters from the factor item sets aiming at different actual projects on the basis of compiling a model factor item set for representing the unit energy efficiency, and adopts a stepwise regression method to optimize and screen the model factor items so as to construct an optimal energy efficiency model (SR) meeting different operation conditions of the current unit. Therefore, the problem that the use range of the model is limited due to the fact that the traditional energy efficiency model is fixed due to the structural form or input variable is effectively solved.
4. Aiming at the problems of unit operation condition change or unit aging, the traditional energy efficiency model adopts a means of adding correction factors or parameter re-identification to calibrate the simulation model, and once the energy efficiency model exceeds the application range, the phenomenon of poor correction effect can be caused. The method can construct an optimal energy efficiency simulation model (SR) with inconsistent structural forms under different unit operation modes, and greatly expands the application range of the model.
Drawings
FIG. 1 is a flow chart of constructing a chiller optimal energy efficiency model (SR) in accordance with the present invention;
FIG. 2 is a comparison chart of calculation results of accuracy test indexes of energy consumption models of water chilling units;
FIG. 3 (a) is a graph of analysis comparing the actual COP values of the BQ model with the simulated predictions;
FIG. 3 (b) is a graph of a comparison of actual COP measurements with simulated predictions for the DOE-2 model;
FIG. 3 (c) is a graph of analysis comparing the measured value of the GNU model COP with the simulated predicted value;
FIG. 3 (d) is a graph of comparison analysis of MP model COP measured values with simulated predictions;
FIG. 3 (e) is a graph of analysis comparing the measured value of COP of the QHP model with the simulated predicted value;
FIG. 3 (f) is a graph showing the comparison between the measured value of the COP of the SR model and the predicted value of the simulation.
Detailed Description
For a further understanding of the invention, its features and advantages, reference is now made to the following examples, which are illustrated in the accompanying drawings in which:
the invention discloses a general modeling method for a water chiller energy efficiency model based on stepwise regression, which comprises two parts of data preprocessing and model construction. The core idea of the data preprocessing part is that firstly, measurable parameters affecting the energy efficiency of a unit are determined through on-site investigation, and corresponding monitoring data are obtained from a monitoring platform database; secondly, eliminating format errors, unreasonable and transient data in the monitoring data; and finally, carrying out further processing on the primary screening data by utilizing a wavelet denoising method of multi-index fusion. The core idea of the water chilling unit energy efficiency model construction part is that firstly, a model factor item set for representing the energy efficiency of the water chilling unit is compiled according to the measurable influence parameters of the water chilling unit, and the model factor item set is mainly obtained through the arrangement of traditional model factor items and the analysis of influence parameter mechanisms; secondly, selecting factor items formed by parameters from the compiled model factor item set according to the measurable influence parameters of the field unit, and constructing an initial water chiller energy efficiency model; and finally, solving the problem of redundancy of the model factor items by using a stepwise regression method, and constructing an optimal energy efficiency model meeting the current actual engineering through the competition relationship among the factor items.
The general modeling method for the water chilling unit energy efficiency model based on stepwise regression, as shown in fig. 1, comprises the following steps:
step one, determining a model factor item set of an energy efficiency model, wherein the model factor item set comprises three types of model factor items, and the first type of model factor item consists of coolant side parameters, such as a frozen water inlet temperature square item T ei 2 Square term M of cooling water flow c 2 Etc.; the second model factor term being constituted by refrigerant-side parameters, e.g. the condensing pressure logarithmic term ln P c Square term T of condensing temperature c 2 Etc.; the third class of model factor terms consists of compressor tuning parameters such as voltage U, current I, etc.
The first model factor is obtained by collecting and sorting model factor in the existing chiller Energy consumption regression model, the existing chiller Energy consumption regression model is Wang, GNU, DOE-2, SL, BQ, MP, SMP and QHP model, each model can be seen in detail in Empirical model for evaluating power consumption of centrifugal chillers (experience model for evaluating power consumption of centrifugal chiller) published in journal of Energy & Buildings (Energy and building), and in analogy analysis method of building environment system published by the national construction industry publishing society in 2006, deST, models for variable-speed centrifugal chillers (model for variable speed centrifugal chiller) published in journal of Ashrae Transactions (American society of heating, refrigeration and air conditioning engineers), A comparison of empirically based steady-state models for vapor-compression liquid chillers (comparison of steady state experience model for steam compressor chiller) published in journal of Applied Thermal Engineering (year 2003), and model for evaluating the performance of water chiller in online predictor of Applied Energy and water chiller in journal of year 8239-7432 (model for predicting water chiller performance of water chiller in year 4 published in year of Applied Energy and society of year, year-9239-7432);
the second model factor item and the third model factor item are obtained by analyzing the action mechanism of the influencing parameters. The analysis method of the parameter action mechanism is shown in DeST, a simulation analysis method of a building environment system published by the national construction industry Press in 2006, and Computer subroutines for rapid evaluation of refrigerant thermodynamic properties (computer subroutine for rapid calculation of thermodynamic properties of a refrigerant) published in Internationale Journal of Refrigeration (International journal of refrigeration) in 1986.
Step two, aiming at a certain actual project, collecting historical operation data of field measurable parameters of the water chilling unit;
judging whether the field measurable parameter in the step two is a composition parameter of a model factor item in the model factor item set, if so, carrying out data preprocessing on the field measurable parameter, wherein the method comprises the following specific steps of:
(1) Reading the data of each measurable parameter in a set test period by a monitoring platform connected with the water chilling unit to form a data sequence;
(2) And screening the data in the data sequence of each measurable parameter to remove non-ideal data in the data sequence to form a data sequence after screening of each measurable parameter.
As an embodiment of the present invention: the non-ideal data includes format error data, unreasonable data, and transient data.
The format error data is data with a value less than or equal to 0 and non-numerical data, such as: if the temperature, the current, the data with the flow of 0 or less than 0 and the non-numerical data exist in the monitoring data, the data can be identified as the data with the wrong format, and the data should be removed;
the unreasonable data is data which cannot pass through parameter threshold value test and simple energy balance relation test, namely the unreasonable data can be judged, and the unreasonable data can be removed;
the transient data is that the difference between the front and rear monitoring data values of the chilled water supply temperature is more than 0.5 ℃ or the percentage difference between the front and rear monitoring data of the compressor current is more than 10%, so that the front and rear monitoring data can be judged to be transient data and can be eliminated.
(3) And carrying out wavelet denoising with multi-index fusion on the primary screening data of each type of measurable parameters processed by the steps. And determining the optimal decomposition reconstruction scale corresponding to each type of measurable parameter data by adopting the comprehensive evaluation index of multi-index fusion, and determining the optimal wavelet basis function corresponding to each type of measurable parameter data by using the Root Mean Square Error (RMSE) and the signal-to-noise ratio (SNR) index. On the basis, a reasonable threshold is constructed for each decomposition level, a soft threshold method is used for processing the high-frequency coefficient, noise components concentrated on the high-frequency part are removed, and finally wavelet reconstruction is carried out on the low-frequency coefficient and the high-frequency coefficient after threshold quantization, so that a data sequence with noise removed by adopting a multi-index fusion wavelet denoising method can be obtained.
Wherein the steps of determining the optimal decomposition reconstruction scale are as follows:
(e) Respectively solving the root mean square error variation C by adopting the numerical values in the data sequence screened by each measurable parameter vrm Variation C of signal to noise ratio snr Smoothness variation C vr For a specific calculation method, see the thesis of the university of south China's 2012, the research of the wavelet denoising quality evaluation method;
(f) The root mean square error variation C of each type of measurable parameter vrm Variation C of signal to noise ratio snr Smoothness variation C vr The numerical values are weighted and fused by adopting an entropy method to obtain a comprehensive evaluation index sequence of the standard wavelet denoising effect, the comprehensive evaluation index sequence of each type of measurable parameter consists of 10 comprehensive evaluation indexes, and the calculation formula of the comprehensive evaluation indexes is as follows:
CEI(m)＝w vrm (m)·C vrm +w vsnr (m)·C vsnr +w vr (m)C vr
wherein m is a wavelet decomposition reconstruction scale and is a positive integer with a value of 1-10; w (w) vrm (m)、w vsnr (m)、w vr (m) C at m scale vrm 、C snr C vr The calculation method of the occupied weight can be specifically referred to the research on the wavelet denoising quality evaluation method in the university of south China's major school treatises in 2012.
(g) And respectively subtracting the maximum value in the sequence from each comprehensive evaluation index sequence of each type of measurable parameter for inversion, carrying out 4-order fitting by using a least square method for analyzing the overall change trend of the sequence, removing abnormal values deviating from the whole, and replacing the abnormal values with fitting values corresponding to the 4-order fitting.
(h) And (3) making a curve according to the numerical value in each comprehensive evaluation index sequence, and searching for an obvious inflection point in the curve change, wherein the decomposition reconstruction scale corresponding to the node is the optimal decomposition reconstruction scale of the primary screening data.
The method for determining the optimal wavelet base comprises the following steps: and selecting an optimal wavelet base suitable for actual running data of the water chilling unit by using a Root Mean Square Error (RMSE) and a signal-to-noise ratio (SNR) index, wherein a calculation equation is as follows:
SNR＝10×lg(power signal /power noise )
wherein f (i) is monitoring data after primary screening; m is a wavelet decomposition reconstruction scale, and the value is a positive integer of 1-10;reconstructing data for decomposition at m scale; n is a number of data. And->
The threshold lambda corresponding to each decomposition level can be calculated by adopting a length logarithm threshold method:
in the middle ofFor noise estimation, a specific calculation method can be referred to a method for judging noise estimation distortion in wavelet denoising published in journal of noise and vibration control in 2015; n is the number of wavelet coefficients of each layer.
Step four, if the field measurable parameters comprise the chilled water outlet temperature T eo Chilled water inlet temperature T ei Chilled water flow M e Then through the calculation formula Q of the refrigerating capacity e ＝C p M e (T ei -T eo ) Calculate Q e Wherein C p The specific heat is water constant pressure, and the value is 4.2 kJ/(K.kg). And (3) taking the field measurable parameters processed in the step (III) and the refrigerating capacity calculated by the processed measurable parameters as known parameters, and selecting model factor items formed by the known parameters from a model factor item set to construct an initial expression of the water chiller energy efficiency model COP. If T eo 、T ei M is as follows e And if the parameters are not field measurable parameters, constructing an initial expression of COP by taking the field measurable parameters processed in the step three as known parameters:
COP＝β 0 X 0 +β 1 X 1 +…+β l X l
wherein l is the number of factor items selected from the model factor item set; beta 0 ,β 1 ,…,β l Fitting coefficients corresponding to the factor items; x is X 0 ,X 1 ,…,X l And the parameters are field measurable parameters processed in the third step respectively.
Step five, solving the problem of factor term redundancy by adopting a stepwise regression method (the method is described in 2007, published in journal of science and technology information, mathematical evaluation model is established by using stepwise regression analysis), and according to each model factor term X in the water chiller energy efficiency model j The contribution rate of (j=0, 1, …, m) (i.e. independent variable) to COP (i.e. dependent variable) is filtered for each model factor term, and model factor terms with insignificant effect on COP are removed, so that the effect of each factor term on COP in the final regression equation is significant.
Step six, before each model factor itemFitting coefficient beta 0 ,β 1 ,…,β m The optimal energy efficiency model of the water chiller energy efficiency model COP is obtained by multiplying each screened model factor by a corresponding coefficient and summing the model factors by adopting multiple linear regression, and the multiple linear regression solving method can be seen in detail in a mathematical model of multiple linear regression published in the journal of Shenyang engineering college (natural science edition) 2005.
In the stepwise regression method, the contribution coefficient of each model factor term is preferably calculated by adopting the dispersion matrix S of each variable, so that the variable does not need to be normalized, and the calculated amount is reduced.
And step seven, collecting measurable parameters related to each model factor item in the COP optimal energy efficiency model under the current running condition of the water chilling unit, preprocessing data of each measurable parameter by adopting the steps (2) - (3), and substituting the measurable parameters preprocessed by the data into the COP optimal energy efficiency model to predict the running performance (namely COP) of the water chilling unit.
Example 1
(1) The general modeling method provided by the invention is demonstrated by adopting time-by-time operation data of a centrifugal water chilling unit in an energy station of certain Tianjin ecological city in a cold supply season (5 months, 1 day to 9 months, 30 days and 153 days). Through on-site actual investigation, the measurable parameters of the unit mainly comprise the chilled water outlet temperature T eo Chilled water inlet temperature T ei Flow M of chilled water e Inlet temperature T of cooling water ci Outlet temperature T of cooling water co Cooling water flow M c 。
(2) Collecting historical operation data of 6 measurable parameters of the water chiller;
(3) Because the 6 measurable parameters are the constituent parameters of the model factor items in the model factor item set, the time-by-time data of each monitoring parameter is subjected to preliminary screening treatment, format errors, unreasonable and transient data are removed, and 698 steady-state data can be finally obtained.
(4) Cooling water outlet temperature T co Cooling water flow M c The calculation results of the optimal wavelet basis and the decomposition reconstruction scale are shown in Table 1 and are adoptedThe values in the data sequence after each measurable parameter screening are respectively solved for the root mean square error variation Cvrm, the signal to noise ratio variation Crnr and the smoothness variation C vr . And respectively obtaining comprehensive evaluation index sequences of haar, db8 and sym6 wavelet denoising effects by using an entropy method.
(5) And respectively reversing the comprehensive evaluation index sequences of haar, db8 and sym6 wavelet denoising effects of all the measurable parameters, and removing corresponding abnormal values.
(6) And (3) making a curve according to the values in the comprehensive evaluation index sequences corresponding to haar, db8 and sym6 wavelets of each measurable parameter, and determining the optimal reconstruction scale by using an inflection point identification method.
(7) Under the condition of the determined optimal decomposition scale, determining the optimal wavelet base suitable for each measurable parameter in three wavelet bases of haar, db8 and sym6 by using Root Mean Square Error (RMSE) and signal-to-noise ratio (SNR) indexes, wherein each measurable parameter is the chilled water outlet temperature T eo Chilled water inlet temperature T ei Flow M of chilled water e Inlet temperature T of cooling water ci As shown.
Table 1 calculation results of optimal wavelet basis and decomposition reconstruction scale:
(8) According to the calculation results of table 1, determining the threshold value corresponding to each decomposition level of each measurable parameter under the optimal wavelet base and the decomposition reconstruction scale, and carrying out wavelet denoising processing on the data sequence of each parameter by using a soft threshold method.
(9) 6 measurable parameters to be subjected to wavelet denoising and T eo 、T ei M is as follows e Solving the refrigerating capacity Q e As a known parameter. And selecting factor items consisting of 7 known parameters from the model factor item set to construct an initial expression of the water chiller energy efficiency model:
(10) The stepwise regression method provided by the invention is adopted to carry out optimization screening on factor items in an initial expression, and coefficients of the factor items of each model are solved by a multiple linear regression method, so that a COP optimal energy efficiency model (SR) of the water chiller is as follows:
(11) The measurable parameter (T) of the water chilling unit under the current operating condition eo 、T ei 、M e 、T ci 、T co M is as follows c ) Data preprocessing is carried out, and the processed measurable parameters and the calculated refrigerating capacity Q are used for calculating e Substituting the model into a COP optimal energy efficiency model (SR) to perform the prediction simulation of the running performance (namely COP) of the water chilling unit. To verify the accuracy of the optimal energy efficiency model (SR), the processed data are also substituted into DOE-2, MP, QHP and GNU models to calculate COP model values.
(12) The errors between the COP actual measurement value and the simulation value of each model are considered by four indexes of the coefficient of variation CV, the root mean square error RMSE, the average absolute error MAE and the average relative error MRE, and the calculation formula is as follows:
wherein: n is the measured data numberA number; y is i Is the COP actual measurement value;is an analog value. The smaller the values of RMSE, MRE and MAE, the closer the standard measured and simulated values are, the higher the model accuracy. The calculation results of the models CV, RMSE, MAE and the MRE index are shown in fig. 2.
As can be seen from fig. 2, the CV value of the SR model is 4.34%, which satisfies the practical requirements of engineering, but the CV value of the other models is greater than 5%. Compared with the other 5 models, the values of RMSE (0.25), MAE (0.18) and MRE (3.17%) in the SR model are the smallest, which shows that the SR model has the highest precision.
In order to further check the model accuracy, the COP actual measurement value and each model simulation value are subjected to comparative analysis, and the comparative analysis result is shown in the figure.
As can be seen from the comparison result in fig. 3 (a), the degree of dispersion between the actual COP value and the analog value of the BQ model is relatively large, and when the COP of the water chiller is greater than 6, the model precision is reduced; as can be seen from comparison results of FIGS. 3 (b), (d) and (e), COP simulation values in DOE-2, MP and QHP models are all near the measured values, the distribution is uniform, and the model precision is higher than that of a BQ model; as can be seen from the comparison result in fig. 3 (c), the degree of dispersion between the COP measured value and the analog value in the GNU model is the greatest, and the model accuracy is the lowest; as can be seen from comparison results in fig. 3 (f), the measured value and the predicted value of the SR model COP have the lowest degree of dispersion, and the two sets of data are closest, and the comparison results can also prove the feasibility and accuracy of the provided method.
Although the preferred embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the appended claims, which are within the scope of the present invention.
Claims (4)
1. The general modeling method for the water chilling unit energy efficiency model based on stepwise regression is characterized by comprising the following steps of:
determining a model factor item set of an energy efficiency model, wherein the model factor item set comprises three types of model factor items, namely a first type of model factor item is composed of secondary refrigerant side parameters, a second type of model factor item is composed of refrigerant side parameters, and a third type of model factor item is composed of compressor regulating parameters;
the first model factor item is obtained through collecting and sorting model factor items in an existing water chiller energy consumption regression model, and the second model factor item and the third model factor item are obtained through analysis of influence parameter action mechanisms;
step two, aiming at a certain actual project, collecting historical operation data of field measurable parameters of the water chilling unit;
judging whether the field measurable parameter in the step two is a composition parameter of a model factor item in the model factor item set, if so, carrying out data preprocessing on the field measurable parameter, wherein the method comprises the following specific steps of:
(1) Reading the data of each measurable parameter in a set test period by a monitoring platform connected with the water chilling unit to form a data sequence;
(2) Screening the data in the data sequence of each measurable parameter to remove non-ideal data in the data sequence to form a data sequence after screening of each measurable parameter;
(3) Performing multi-index fusion wavelet denoising on the primary screening data of each type of measurable parameter processed by the steps, determining the optimal decomposition reconstruction scale corresponding to each type of measurable parameter data by adopting a multi-index fusion comprehensive evaluation index, determining the optimal wavelet basis function corresponding to each type of measurable parameter data by using a Root Mean Square Error (RMSE) and a signal-to-noise ratio (SNR) index, constructing a reasonable threshold value for each decomposition level on the basis, processing a high-frequency coefficient by using a soft threshold value method, removing noise components concentrated on a high-frequency part, and performing wavelet reconstruction on a low-frequency coefficient and the high-frequency coefficient after the threshold value quantization to obtain a data sequence after noise removal by adopting a multi-index fusion wavelet denoising method;
wherein the steps of determining the optimal decomposition reconstruction scale are as follows:
(a) Respectively solving the root mean square error variation C by adopting the numerical values in the data sequence screened by each measurable parameter vrm Variation C of signal to noise ratio snr Smoothness variation C vr ；
(b) The root mean square error variation C of each type of measurable parameter vrm Variation C of signal to noise ratio snr Smoothness variation C vr The numerical values are weighted and fused by adopting an entropy method to obtain a comprehensive evaluation index sequence of the standard wavelet denoising effect, the comprehensive evaluation index sequence of each type of measurable parameter consists of 10 comprehensive evaluation indexes, and the calculation formula of the comprehensive evaluation indexes is as follows:
CEI(m)＝w vrm (m)·C vrm +w vsnr (m)·C vsnr +w vr (m)C vr
wherein m is a wavelet decomposition reconstruction scale and is a positive integer with a value of 1-10; w (w) vrm (m)、w vsnr (m)、w vr (m) C at m scale vrm 、C snr C vr The weight of the vehicle is occupied;
(c) Subtracting the maximum value in the sequence from each comprehensive evaluation index in the comprehensive evaluation index sequence of each type of measurable parameter respectively for inversion, carrying out 4-order fitting by using a least square method for analyzing the overall change trend of the sequence, removing abnormal values deviating from the whole, and replacing the abnormal values with fitting values corresponding to the 4-order fitting;
(d) Making a curve according to the numerical value in each comprehensive evaluation index sequence, and searching for an obvious inflection point in the curve change, wherein the decomposition reconstruction scale corresponding to the node is the optimal decomposition reconstruction scale of the primary screening data;
the method for determining the optimal wavelet base comprises the following steps: and selecting an optimal wavelet base suitable for actual running data of the water chilling unit by using a Root Mean Square Error (RMSE) and a signal-to-noise ratio (SNR) index, wherein a calculation equation is as follows:
SNR＝10×lg(power signal /power noise )
wherein f (i) is monitoring data after primary screening; m is a wavelet decomposition reconstruction scale, and the value is a positive integer of 1-10;reconstructing data for decomposition at m scale; n is the number of data, and->
The threshold lambda corresponding to each decomposition level is calculated by adopting a length logarithm threshold method:
in the middle ofFor noise estimation, N is the number of wavelet coefficients of each layer;
step four, if the field measurable parameters comprise the chilled water outlet temperature T eo Chilled water inlet temperature T ei Chilled water flow M e Then through the calculation formula Q of the refrigerating capacity e ＝C p M e (T ei -T eo ) Calculate Q e Wherein C p The specific heat is water constant pressure, the specific heat takes the value of 4.2 kJ/(K.kg), the field measurable parameter processed in the third step and the refrigerating capacity calculated by the processed measurable parameter are taken as the known parameters, the model factor items formed by the known parameters are selected from the model factor item set to construct the initial expression of the COP of the energy efficiency model of the water chilling unit, if T eo 、T ei M is as follows e If the parameters are not field measurable parameters, the field measurable parameters processed in the third step are taken as the field measurable parametersThe known parameters construct an initial expression of COP:
COP＝β 0 X 0 +β 1 X 1 +…+β l X l
wherein l is the number of factor items selected from the model factor item set; beta 0 ,β 1 ,…,β l Fitting coefficients corresponding to the factor items; x is X 0 ,X 1 ,…,X l Respectively the field measurable parameters processed in the third step;
step five, screening each model factor according to the contribution rate of each model factor independent variable in the water chiller energy efficiency model to the COP, and removing model factor with insignificant influence on the COP, so that the influence of each factor in the final regression equation on the COP is significant;
solving fitting coefficients before each model factor item by adopting multiple linear regression, multiplying each model factor after screening by a corresponding coefficient, and summing to obtain an optimal energy efficiency model of the water chiller energy efficiency model COP;
and step seven, collecting measurable parameters related to each model factor item in the COP optimal energy efficiency model under the current running condition of the water chilling unit, preprocessing data of each measurable parameter by adopting the steps (2) - (3), and substituting the measurable parameters preprocessed by the data into the COP optimal energy efficiency model to predict the running performance of the water chilling unit.
2. The stepwise regression-based general modeling method for a water chilling unit energy efficiency model of claim 1, wherein: and calculating contribution coefficients of each model factor item by adopting a dispersion matrix S of each variable in the stepwise regression method.
3. The stepwise regression-based general modeling method for a water chiller energy efficiency model according to claim 1 or 2, characterized by: the non-ideal data includes format error data, unreasonable data, and transient data.
4. The stepwise regression-based water chilling unit energy efficiency model general modeling method of claim 3, wherein: the format error data are data with the value less than or equal to 0 and non-numerical data;
the unreasonable data is data which cannot pass through parameter threshold value verification and simple energy balance relation verification;
the transient data is that the difference between the front and rear monitoring data values of the chilled water supply temperature is more than 0.5 ℃ or the percentage difference between the front and rear monitoring data of the compressor current is more than 10%, namely the front and rear monitoring data are judged to be transient data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811212546.8A CN109583008B (en) | 2018-10-18 | 2018-10-18 | General modeling method for water chiller energy efficiency model based on stepwise regression |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811212546.8A CN109583008B (en) | 2018-10-18 | 2018-10-18 | General modeling method for water chiller energy efficiency model based on stepwise regression |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109583008A CN109583008A (en) | 2019-04-05 |
| CN109583008B true CN109583008B (en) | 2023-09-15 |
Family
ID=65920589
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811212546.8A Active CN109583008B (en) | 2018-10-18 | 2018-10-18 | General modeling method for water chiller energy efficiency model based on stepwise regression |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109583008B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110851897B (en) * | 2019-10-15 | 2024-03-22 | 天津大学 | Aqueduct stress-strain prediction method under multi-factor association |
| CN111442480A (en) * | 2020-04-08 | 2020-07-24 | 广东美的暖通设备有限公司 | Operation control method and system for air conditioning equipment, air conditioning equipment and storage medium |
| CN112503789B (en) * | 2020-12-09 | 2021-09-03 | 西安交通大学 | Intermediate pressure control method for cascade refrigeration system |
| CN113124919B (en) * | 2021-04-09 | 2021-12-07 | 中移(上海)信息通信科技有限公司 | Abnormity monitoring method and device for reference station and related equipment |
| CN114279235B (en) * | 2021-12-29 | 2024-05-10 | 博锐尚格科技股份有限公司 | Cooling tower operation control method based on switching of black box model and ash box model |
| CN115456115B (en) * | 2022-11-14 | 2023-04-07 | 广东电网有限责任公司佛山供电局 | Cold station operation and maintenance multilayer energy-saving potential diagnosis method based on actual measurement subentry measurement data |
| CN117847844B (en) * | 2024-01-15 | 2024-06-07 | 华商博雅(北京)科技有限公司 | Temperature division control method and system for carbon dioxide heat pump |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014041566A (en) * | 2012-08-23 | 2014-03-06 | Nippon Telegr & Teleph Corp <Ntt> | Device, method, and program for linear regression model estimation |
| CN105627504A (en) * | 2015-05-28 | 2016-06-01 | 重庆大学 | Energy consumption estimation method for water chilling unit of variable-air-volume central air conditioning system based on support vector machine |
| CN108168030A (en) * | 2017-12-14 | 2018-06-15 | 南京师范大学 | A kind of intelligent control method based on refrigeration performance curve |
-
2018
- 2018-10-18 CN CN201811212546.8A patent/CN109583008B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014041566A (en) * | 2012-08-23 | 2014-03-06 | Nippon Telegr & Teleph Corp <Ntt> | Device, method, and program for linear regression model estimation |
| CN105627504A (en) * | 2015-05-28 | 2016-06-01 | 重庆大学 | Energy consumption estimation method for water chilling unit of variable-air-volume central air conditioning system based on support vector machine |
| CN108168030A (en) * | 2017-12-14 | 2018-06-15 | 南京师范大学 | A kind of intelligent control method based on refrigeration performance curve |
Non-Patent Citations (4)
| Title |
|---|
| 《A comparison of empirically based steady-state models for vapor-compression liquid chillers》;Derk J. Swider;《Applied Thermal Engineering》;第23卷(第5期);第539-556页 * |
| 《An evaluation of empirically-based models for predicting energy performance of vapor-compression water chillers》;Tzong-Shing Lee等;《Applied Energy》;第87卷(第11期);第3486-3493页 * |
| 《基于CPSO-SVR 的冷水机组运行能效预测模型研究》;李利文等;《科技与创新》(第15期);第78-79页 * |
| 《基于机器学习的冷水机组能耗模型辨识方法研究》;魏峥等;《建筑科学》;第34卷(第6期);第115-122页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109583008A (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109583008B (en) | General modeling method for water chiller energy efficiency model based on stepwise regression | |
| CN112001439B (en) | GBDT-based market building air conditioner cold load prediction method, storage medium and equipment | |
| Zhao et al. | A review on the prediction of building energy consumption | |
| CN1106543C (en) | System for monitoring outdoor heat exchanger coil | |
| Zhou et al. | Benchmarking air-conditioning energy performance of residential rooms based on regression and clustering techniques | |
| Qian et al. | Power consumption and energy efficiency of VRF system based on large scale monitoring virtual sensors | |
| Monfet et al. | Ongoing commissioning of water-cooled electric chillers using benchmarking models | |
| CN117459073B (en) | Intelligent management method for heat pump system operation data | |
| Reddy et al. | Evaluation of the suitability of different chiller performance models for on-line training applied to automated fault detection and diagnosis (RP-1139) | |
| CN107992991A (en) | Annual electricity sales amount Forecasting Methodology based on external environmental factor and Co-integration Theory | |
| CN117494546A (en) | Data-driven air conditioner set temperature correlation simulation method, equipment and medium | |
| Vering et al. | Digital twin design with on-line calibration for HVAC systems in buildings | |
| CN113432247B (en) | Water chilling unit energy consumption prediction method and system based on graph neural network and storage medium | |
| CN109711049B (en) | Hybrid subway air conditioner water chilling unit energy efficiency estimation method | |
| CN110688740A (en) | Modelica combined simulation optimization-based cold water machine room model calibration method | |
| CN114021461A (en) | XGboost-based electronic expansion valve mass flow characteristic prediction method | |
| CN109359780B (en) | Residential electricity consumption prediction method based on electrical appliance index | |
| CN115654684A (en) | Air handling unit control and model construction method, device, equipment and medium | |
| CN109840307A (en) | A kind of indoor temperature and humidity field quick predict system and method | |
| CN116151793A (en) | Power transformation equipment evaluation and overhaul method and system based on fuzzy comprehensive evaluation | |
| Zhou et al. | Optimization of Metro Central Air Conditioning Cold Source System Based on PCA-ANN Data Model | |
| KR20040042734A (en) | Method for analysis of efficiency in turbo refrigerator and apparatus thereof | |
| Li et al. | On-line models for use in automated fault detection and diagnosis for HVAC&R equipment | |
| Gao et al. | A Steady-State Modeling Method for Direct Expansion Air Conditioning Systems. | |
| CN117436007A (en) | Fault diagnosis method, system, medium and equipment for water chilling unit |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |