CN109767044A - A method for predicting thermal energy conversion efficiency of ground source heat pump system with operating condition input - Google Patents
A method for predicting thermal energy conversion efficiency of ground source heat pump system with operating condition input Download PDFInfo
- Publication number
- CN109767044A CN109767044A CN201910043545.3A CN201910043545A CN109767044A CN 109767044 A CN109767044 A CN 109767044A CN 201910043545 A CN201910043545 A CN 201910043545A CN 109767044 A CN109767044 A CN 109767044A
- Authority
- CN
- China
- Prior art keywords
- thermal energy
- time series
- conversion efficiency
- prediction
- energy conversion
- 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.)
- Granted
Links
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000001932 seasonal effect Effects 0.000 claims abstract description 25
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 15
- 238000007637 random forest analysis Methods 0.000 claims abstract description 14
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 230000000737 periodic effect Effects 0.000 claims description 13
- 238000012800 visualization Methods 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 230000010354 integration Effects 0.000 claims description 5
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims description 4
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000013213 extrapolation Methods 0.000 claims description 3
- 239000007788 liquid Substances 0.000 claims description 3
- 239000000463 material Substances 0.000 claims description 3
- 238000012856 packing Methods 0.000 claims description 3
- 238000005192 partition Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 abstract description 2
- 238000004140 cleaning Methods 0.000 abstract 1
- 238000007781 pre-processing Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 8
- 230000008901 benefit Effects 0.000 description 5
- 238000009499 grossing Methods 0.000 description 3
- 238000010438 heat treatment Methods 0.000 description 3
- 238000000714 time series forecasting Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000001816 cooling Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000611 regression analysis Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004378 air conditioning Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Investigating Or Analyzing Materials Using Thermal Means (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明涉及一种带有工况输入的对地源热泵系统热能转换效率进行预测的方法，包括步骤如下：A、热能数据预处理1)清洗热能数据；2)对热能转换效率的时间序列进行季节性分解和初步分析；B、建立带有外部工况参数输入的自回归模型3)构建带有外部工况参数的自回归模型框架；4)构造基本分类回归树，生成随机森林，训练不同外部工况参数下的预测模型，选取使误差不再变化的树的最少棵树；C、热能转换效率预测根据训练好的预测模型，将外部工况参数和热能转换效率的时间序列整合输入进行预测，输出预测结果和预测精度。本发明将工况参数引入到时间序列预测当中，这样在预测时针对不同的工况参数会得到特定的更精确的预测模型。
The invention relates to a method for predicting the thermal energy conversion efficiency of a ground source heat pump system with input of operating conditions, comprising the following steps: A. Thermal energy data preprocessing 1) Cleaning the thermal energy data; 2) Performing a time series on the thermal energy conversion efficiency Seasonal decomposition and preliminary analysis; B. Establish an autoregressive model with external working condition parameter input 3) Build an autoregressive model framework with external working condition parameters; 4) Construct a basic classification regression tree, generate random forests, and train different For the prediction model under the external working condition parameters, select the least tree that makes the error no longer change; C. Heat energy conversion efficiency prediction According to the trained prediction model, the external working condition parameters and the time series of heat energy conversion efficiency are integrated and inputted. Predict, output prediction results and prediction accuracy. The present invention introduces working condition parameters into time series prediction, so that a specific and more accurate prediction model can be obtained for different working condition parameters during prediction.
Description
Technical field
The present invention relates to it is a kind of with operating condition input the method that ground-source heat pump system thermal energy conversion efficiency is predicted,
Belong to multi-energy data digging technology field.
Background technique
Ground-source heat pump system (GSHP) is a kind of using underground shallow layer geothermal energy resources, realizes to building and provides heating, system
The air-conditioning technical system of cold and domestic hot-water service.Compared with general resource, heat pump system has the advantage that (1) is significant
Environment and economic benefit, no waste that (2) are pollution-free to water source, (3) in air conditioner refrigerating, for warming up, daily hot water is answered
With (4) maintenance cost is low and saves space.Due to its advantage with high-efficient energy-saving environment friendly, meet sustainable development idea, ground
Source heat pump system is widely used all over the world.In order to assess and predict the thermal energy conversion efficiency of the system,
Optimal parameter is found, to guarantee the stability high efficiency of the system, researchers have carried out various research, however, about
The forecast analysis of ground-source heat pump system performance is seldom, and still to be improved.
Currently, the performance prediction and feasibility analysis to heat pump system are mainly also rested on for extremely complex heat pump system
System itself such as changes system structure and parameter, or the built-up pattern that adjustment is different, Lai Jinhang performance compare and analyze.It is this
Method is extremely complex, and operation difficulty is big, and can not provide practical analysis result.Start with from data mining to carry out prediction point
Analysis, precision of prediction is high, and technology maturation is easy to operate.
Multi-energy data excavation is mainly used in recurrence and time series forecasting.Regression analysis is a kind of mathematical model, refers to and grinds
Study carefully one group of stochastic variable (Y1,Y2,…,Ym) and another group of stochastic variable (X1,X2,…,Xn) between relationship statistical analysis technique,
Also known as multiple regression analysis.Usually (Y1,Y2,…,Ym) it is dependent variable, (X1,X2,…,Xn) it is independent variable.For example classics is linear
Regression model is exactly the linear relationship established between dependent variable and independent variable, can fit straight line in corresponding space.
Time series is also time series, history plural number or dynamic series.It is by the numerical value of certain statistical indicator, in chronological sequence
Sequence, which is discharged to, is formed by ordered series of numbers.Time series forecasting is exactly to pass through establishment and analysis time sequence, according to time series institute
Development process, direction and the trend reflected, is analogized or is extended, so as to predicting in lower a period of time or several years later
The level being likely to be breached.As the superior performance of ground-source heat pump system is of interest by more and more people, can people to its performance
Carrying out more accurate prediction becomes a urgent problem to be solved.Especially if the design phase in earth source heat pump will
Stringent input parameter is determined to complete building delivery, and is ensured that good performance, then Accurate Prediction ground source heat
The performance of pumping system is most important.
Summary of the invention
In view of the deficiencies of the prior art, ground-source heat pump system thermal energy is turned with operating condition input the present invention provides a kind of
Change the method that efficiency is predicted.
Summary of the invention:
A method of with predicting ground-source heat pump system thermal energy conversion efficiency for operating condition input, including data are pre-
It handles, establish three parts of prediction model and thermal energy conversion efficiency prediction.
In order to guarantee the reliability of initial data, it is cleaned, the exceptional value for being mingled in the inside and illegal word are rejected
Symbol.In order to determine the stationarities features such as the trend terms of data, season, seasonal decomposition is carried out to initial data.In order to when
Between duty parameter is introduced on the basis of sequence, establish the Self-regression Forecast Model with the input of external operating condition.
Term is explained:
COP refers to Coefficient Of Performance, i.e. conversion ratio between energy and heat, referred to as makes
Hot Energy Efficiency Ratio.The energy of unit 1 is converted as the heat of unit 0.5, i.e. COP is 0.5.COP value is in ARI standard, about the summer in winter
Season cycle efficieny propose it is defined below: in winter when heat supply, the ratio of heating capacity and input power is defined as the circulation of heat pump
Coefficient of performance.In cooling in summer, the ratio of refrigerating capacity and input power is defined as the Energy Efficiency Ratio BER of heat pump.Not cause
Ambiguity, the Energy Efficiency Ratio expression-form of the winter heat pump cycle coefficient of performance and summer heat pump is all made of COP (Energy Efficiency Ratio) table by us
Show.
Technical scheme is as follows:
A method of with predicting ground-source heat pump system thermal energy conversion efficiency for operating condition input, including step is such as
Under:
A, thermal energy data pre-processes
1) thermal energy data is cleaned；
2) seasonal decomposition and preliminary analysis are carried out to the time series of thermal energy conversion efficiency；
B, the autoregression model with the input of external duty parameter is established
3) building has the autoregression model frame of external duty parameter；
4) basic classification regression tree is constructed, random forest, the prediction model under the different external duty parameters of training, choosing are generated
Minimum of the tree for changing error no longer is taken to set；So that model is most simple.
C, thermal energy conversion efficiency is predicted
According to trained prediction model, by the integration of the time series of external duty parameter and thermal energy conversion efficiency input into
Row prediction, exports prediction result and precision of prediction.
The present invention includes two parts to the prediction of ground-source heat pump system thermal energy conversion efficiency, and wherein first part will clean
Rear time series carries out seasonal decomposition, analyzes the seasonal feature and tendency feature year by year of thermal energy conversion efficiency；The
Two parts are by first part it is determined that being input to the autoregression model with duty parameter for the time series after steadily, completion is pre-
It surveys, exports prediction result and prediction accuracy.
It is preferred according to the present invention, the step A, thermal energy data pretreatment,
Thermal energy data includes the time series and outside duty parameter of thermal energy conversion efficiency；
The time series of thermal energy conversion efficiency includes: heat pump unit power consumption Y1, GSHP system power consumption Y2, unit COP
Y3, GSHP system COP Y4, Y1、Y2、Y3、Y4In each data be length be L time series, L > 24；
External duty parameter includes: distribution of boreholes form X1, boring radius X2, pipe laying depth X3, pipe laying longitudinal pitch X4、
Pipe laying horizontal spacing X5, pipe laying quantity X6, manage in packing material thermal coefficient X7, U-tube nominal outside diameter X8, U-tube spacing X9, it is remote
Hold ground temperature X10, thermal conductivity X11With circulating liquid type X12；A shared M initial data is set, that is, is shown in a matrix
It is M row, L*4+12 column；It comprises the following steps that
A, original M initial data is read, detects the value of NAN value and nonumeric form, deletes NAN value and nonumeric shape
Row where the value of formula；
B, Y is taken out respectively from the step a data that obtain that treated1、Y2、Y3、Y4Visualization display is done, according to profession
Micro-judgment goes out to have abnormal numerical value, the row where the numerical value of suppressing exception；For example, in addition to being in April and October
Heat pump unit power consumption (the Y of the system of closed state1) value be outside zero, the value in other months does not exceed 35000kwh, and
There are the obvious abnormal values of some 60000kwh or more, are rejected；
C, the Y for obtaining step b1、Y2、Y3、Y4Four values do seasonal decomposition respectively, resolve into trend term T, season
S, periodic term C, stochastic error I, for a time series Yj, j=1,2,3,4, it is expressed as { yj,t, t=1,2 ..., L }, L
It indicates the length of the time series of thermal energy conversion efficiency, indicates time series YjIn the occurrence of t moment thermal energy conversion efficiency；Season
Section property decomposition model is Yj=Tj+Sj+Cj+Ij, trend term TjIt is obtained by the way of sliding average, formula is as follows:
In formula (I), p is the lag item of sliding average, time series YjIt is T in the value of the trend term at t+1 momentj,t+1；
Season SjIt is obtained by trend extrapolation partition method, after the time series of thermal energy conversion efficiency is subtracted trend term
It is monthly averaged, then subtracts each other item by item with average value, just obtain season Sj；
Periodic term CjIt is obtained using period map method: by thermal energy conversion efficiency time series YjSubtract trend term and after season
Decentralization processing is carried out, Fourier expansion is then carried out, obtains its maximum sinusoidal component AjIt is with cosine component
Number Bj, next, obtaining t moment time series YjThe value of periodic termτ refers to week
Phase；Because the influence of periodic term is minimum, also it is directly merged with error term sometimes；
Trend term T is subtracted from original time seriesj, season Sj, periodic term CjAfterwards, stochastic error I is obtainedj；
D, seasonal decomposition result is obtained in analytical procedure c, tentatively judges the tendency of the time series of thermal energy conversion efficiency
Situation and seasonal characteristics.For example, by observing apparent seasonal variety, it may be said that the bright data by seasonal effect compared with
Greatly, there is very strong rule that can follow, convenient for modeling using time series models；By observing the available Y of trend term1、Y2、Y3、Y4
With the whole tendency situation in time, there can be an intuitive Trend judgement to power consumption and COP.
According to the present invention preferred, the step B, establish the autoregression model for having external duty parameter, comprising:
E, tentatively judge that the time series of thermal energy conversion efficiency includes seasonal and visible trend feature by step d
Afterwards, it establishes the autoregression model frame for having external duty parameter: d external duty parameter is indicated with a d dimensional vector x,
Random Forest model is indicated with f (), obtains the predicted value of t moment using historical data and external duty parameterSuch as formula
(II) shown in:
In formula (II), q is lag item, and q historical data is predicted before expression t moment；
F, the autoregression model frame obtained according to step e establishes the random forest for specifically having external duty parameter
Model；
M is randomly selected from all (d+1) × q+d attributes, then select from this m attribute optimal dividing attribute into
Line splitting, optimal dividing Attributes Splitting at two child nodes sum of variance it is minimum, set the property set for some node
G is closed, using one of attribute g as optimal dividing attribute, by this node split at D1,D2Two child nodes, then optimal dividing
Shown in attribute such as formula (III):
In formula (III), V (D1),V(D2) refer to D1,D2Variance, g*Refer to optimal Split Attribute；
It is iteratively generating subtree, until cannot divide, ultimately generates a regression tree；
K regression tree is generated, this obtained K regression result is averaged after forming random forest, has just obtained t
The predicted value at momentAs shown in formula (IV):
In formula (IV), fkThe regression result of () expression kth regression tree.
According to the present invention preferred, the step C, according to trained prediction model, by external duty parameter and heat
The time series integration input of energy transfer efficiency is predicted, is exported prediction result and precision of prediction, is comprised the following steps that
To be divided into length be L to the time series of thermal energy conversion efficiency for being L by length1Historical data and length be L2Not
Carry out concept of reality measured data, i.e., with preceding L1A historical data, L after prediction2A future concept of reality measured data, the result that prediction is obtained
Compared with true observation, precision of prediction is obtained, the standard of predictive metrics order of accuarcy includes formula (V), formula (VI), formula (VII):
Formula (V), formula (VI), in formula (VII),
MAE full name is Mean Absolute Error, i.e. mean absolute error, is all single observations and true value
Inclined absolute value of the difference be averaged；
MAPE full name is Mean Absolute Percentage Error, i.e. average absolute percentage error, is all lists
The absolute value of true value percentage shared by a observation and true value deviation is averaged；
RMSE full name is Root Mean Square Error, i.e. root-mean-square error is all single observations and true
The deviation of value square takes after mean value sqrt again.
N indicates the number of the time series of the thermal energy conversion efficiency to be predicted, L2Indicate the length of predicted time series
Degree,WithRespectively indicate the predicted value and true observation of i-th of time series of t moment；
It is L that the N number of time series that will be predicted, which is divided into length,1Historical data and length be L2Future true observation number
According to rear, by N × L1Matrix be sent into trained model after obtain N × L2A prediction data passes through formula (V), formula (VI), formula
(VII) precision of prediction is obtained.
The invention has the benefit that
1, duty parameter is introduced into time series forecasting by the present invention, is joined in this way in prediction for different operating conditions
Number can obtain specific more accurate prediction model.
2, the present invention is simple for heat pump system performance prediction technique, and operation is easy, and is not required to be appreciated that during prediction
With the system parameter for calculating complexity.
3, the present invention realizes that prediction model, this integrated learning approach can obtain accurately using the method for random forest
Prediction result, and predetermined speed is fast.
Detailed description of the invention
Fig. 1 is a kind of method predicted ground-source heat pump system thermal energy conversion efficiency with operating condition input of the present invention
Overall flow figure；
Fig. 2 is prediction model internal structure chart；
Fig. 3 (a) is Y1Trend term visualization figure；
Fig. 3 (b) is Y1Season visualization figure；
Fig. 3 (c) is Y1Periodic term visualization figure；
Fig. 3 (d) is Y1Random entry visualization figure；
Fig. 4 (a) is Y1Rear 60 months prediction results and other existing method prediction results comparison diagram；
Fig. 4 (b) is Y2The comparison diagram of the rear 60 months prediction results and other existing method prediction results of application；
Fig. 4 (c) is Y3Rear 60 months prediction results and other existing method prediction results comparison diagram；
Fig. 4 (d) is Y4Rear 60 months prediction results and other existing method prediction results comparison diagram；
Specific embodiment
The present invention is further qualified with embodiment with reference to the accompanying drawings of the specification, but not limited to this.
Embodiment 1
A method of with predicting ground-source heat pump system thermal energy conversion efficiency for operating condition input, as shown in Figure 1,
It comprises the following steps that
A, thermal energy data pre-processes
1) thermal energy data is cleaned；
2) seasonal decomposition and preliminary analysis are carried out to the time series of thermal energy conversion efficiency；
B, the autoregression model with the input of external duty parameter is established
3) building has the autoregression model frame of external duty parameter；
4) basic classification regression tree is constructed, random forest, the prediction model under the different external duty parameters of training, choosing are generated
Minimum of the tree for changing error no longer is taken to set；So that model is most simple.
C, thermal energy conversion efficiency is predicted
According to trained prediction model, by the integration of the time series of external duty parameter and thermal energy conversion efficiency input into
Row prediction, exports prediction result and precision of prediction.
The present invention includes two parts to the prediction of ground-source heat pump system thermal energy conversion efficiency, and wherein first part will clean
Rear time series carries out seasonal decomposition, analyzes the seasonal feature and tendency feature year by year of thermal energy conversion efficiency；The
Two parts are by first part it is determined that being input to the autoregression model with duty parameter for the time series after steadily, completion is pre-
It surveys, exports prediction result and prediction accuracy.
Embodiment 2
It is according to claim 1 a kind of pre- to the progress of ground-source heat pump system thermal energy conversion efficiency with operating condition input
The method of survey is distinguished and is,
The step A, thermal energy data pretreatment,
The thermal energy data of the ground coupling heat pump system of " star of underground heat " acquisition includes the time series of thermal energy conversion efficiency
With external duty parameter；
The time series of thermal energy conversion efficiency includes: heat pump unit power consumption Y1, GSHP system power consumption Y2, unit COP
Y3, GSHP system COP Y4, Y1、Y2、Y3、Y4In each data be length be L time series, L=360；
External duty parameter includes: distribution of boreholes form X1, boring radius X2, pipe laying depth X3, pipe laying longitudinal pitch X4、
Pipe laying horizontal spacing X5, pipe laying quantity X6, manage in packing material thermal coefficient X7, U-tube nominal outside diameter X8, U-tube spacing X9, it is remote
Hold ground temperature X10, thermal conductivity X11With circulating liquid type X12；A shared M initial data is set, that is, is shown in a matrix
It is M row, L*4+12 is arranged, and M=5000 is comprised the following steps that
A, original M initial data is read, detects the value of NAN value and nonumeric form, deletes NAN value and nonumeric shape
Row where the value of formula；
B, Y is taken out respectively from the step a data that obtain that treated1、Y2、Y3、Y4Visualization display is done, according to profession
Micro-judgment goes out to have abnormal numerical value, the row where the numerical value of suppressing exception；For example, in addition to being in April and October
Heat pump unit power consumption (the Y of the system of closed state1) value be outside zero, the value in other months does not exceed 35000kwh, and
There are the obvious abnormal values of some 60000kwh or more, are rejected；
C, the Y for obtaining step b1、Y2、Y3、Y4Four values do seasonal decomposition respectively, resolve into trend term T, season
S, periodic term C, stochastic error I, for a time series Yj, j=1,2,3,4, it is expressed as { yj,t, t=1,2 ..., L }, L
It indicates the length of the time series of thermal energy conversion efficiency, indicates time series YjIn the occurrence of t moment thermal energy conversion efficiency；Season
Section property decomposition model is Yj=Tj+Sj+Cj+Ij, trend term TjIt is obtained by the way of sliding average, formula is as follows:
In formula (I), p is the lag item of sliding average, time series YjIt is T in the value of the trend term at t+1 momentj,t+1；
Season SjIt is obtained by trend extrapolation partition method, after the time series of thermal energy conversion efficiency is subtracted trend term
It is monthly averaged, then subtracts each other item by item with average value, just obtain season Sj；
Periodic term CjIt is obtained using period map method: by thermal energy conversion efficiency time series YjSubtract trend term and after season
Decentralization processing is carried out, Fourier expansion is then carried out, obtains its maximum sinusoidal component AjIt is with cosine component
Number Bj, next, obtaining t moment time series YjThe value of periodic termτ refers to week
Phase；Because the influence of periodic term is minimum, also it is directly merged with error term sometimes；
Trend term T is subtracted from original time seriesj, season Sj, periodic term CjAfterwards, stochastic error I is obtainedj；
Fig. 3 (a) is Y1Trend term visualization figure；Abscissa indicates month, and ordinate indicates corresponding Y1Trend term
Value；
Fig. 3 (b) is Y1Season visualization figure；Abscissa indicates month, and ordinate indicates corresponding Y1Season
Value；
Fig. 3 (c) is Y1Periodic term visualization figure；Abscissa indicates month, and ordinate indicates corresponding Y1Periodic term
Value；
Fig. 3 (d) is Y1Random entry visualization figure；Abscissa indicates month, and ordinate indicates corresponding Y1Random entry
Value；
From Fig. 3 (a) to Fig. 3 (d) it is found that Y1Trend term there is a slight decline to become (between 30 years) at this 360 months
Gesture, this is consistent with practical engineering application, the cooling of heat pump unit or heating power really can with the increase of the service life and
Reduce.And Y1Season show Y1With obviously seasonal characteristics；Y1Periodic term reflect the sequence potential period
Property variation, it can be seen that this very little, influence less.Y1Random entry then reflect a series of uncertainty.It can from figure
To find out, very little is fluctuated at 0, compared with whole, the influence of stochastic error is smaller.According to seasonality decompose as a result,
Y1~Y4Four amounts are stationary time series, and explanation can carry out further modeling and forecasting research to this four amounts.
D, seasonal decomposition result is obtained in analytical procedure c, tentatively judges the tendency of the time series of thermal energy conversion efficiency
Situation and seasonal characteristics.For example, by observing apparent seasonal variety, it may be said that the bright data by seasonal effect compared with
Greatly, there is very strong rule that can follow, convenient for modeling using time series models；By observing the available Y of trend term1、Y2、Y3、Y4
With the whole tendency situation in time, there can be an intuitive Trend judgement to power consumption and COP.
The step B establishes the autoregression model for having external duty parameter, comprising:
E, tentatively judge that the time series of thermal energy conversion efficiency includes seasonal and visible trend feature by step d
Afterwards, it establishes the autoregression model frame for having external duty parameter: d external duty parameter is indicated with a d dimensional vector x,
Random Forest model is indicated with f (), obtains the predicted value of t moment using historical data and external duty parameterSuch as formula
(II) shown in:
In formula (II), q is lag item, and q historical data is predicted before expression t moment；
F, the autoregression model frame obtained according to step e establishes the random forest for specifically having external duty parameter
Model；
M is randomly selected from all (d+1) × q+d attributes, then select from this m attribute optimal dividing attribute into
Line splitting, optimal dividing Attributes Splitting at two child nodes sum of variance it is minimum, set the property set for some node
G is closed, using one of attribute g as optimal dividing attribute, by this node split at D1,D2Two child nodes, then optimal dividing
Shown in attribute such as formula (III):
In formula (III), V (D1),V(D2) refer to D1,D2Variance, g*Refer to optimal Split Attribute；
It is iteratively generating subtree, until cannot divide, ultimately generates a regression tree；
K regression tree is generated, this obtained K regression result is averaged after forming random forest, has just obtained t
The predicted value at momentAs shown in formula (IV):
In formula (IV), fkThe regression result of () expression kth regression tree.Fig. 2 is prediction model internal structure chart；
The step C, according to trained prediction model, by the time series of external duty parameter and thermal energy conversion efficiency
Integration input is predicted, is exported prediction result and precision of prediction, is comprised the following steps that
To be divided into length be L to the time series of thermal energy conversion efficiency for being L by length1Historical data and length be L2Not
Come concept of reality measured data, L1=300, L2=60, i.e., with preceding L1A historical data, L after prediction2A future concept of reality measured data, will
Obtained result is predicted compared with true observation, obtains precision of prediction, the standard of predictive metrics order of accuarcy include formula (V),
Formula (VI), formula (VII):
Formula (V), formula (VI), in formula (VII),
MAE full name is Mean Absolute Error, i.e. mean absolute error, is all single observations and true value
Inclined absolute value of the difference be averaged；
MAPE full name is Mean Absolute Percentage Error, i.e. average absolute percentage error, is all lists
The absolute value of true value percentage shared by a observation and true value deviation is averaged；
RMSE full name is Root Mean Square Error, i.e. root-mean-square error is all single observations and true
The deviation of value square takes after mean value sqrt again.
N indicates the number of the time series of the thermal energy conversion efficiency to be predicted, L2Indicate the length of predicted time series
Degree,WithRespectively indicate the predicted value and true observation of i-th of time series of t moment；
It is L that the N number of time series that will be predicted, which is divided into length,1Historical data and length be L2Future true observation number
According to rear, by N × L1Matrix be sent into trained model after obtain N × L2A prediction data passes through formula (V), formula (VI), formula
(VII) precision of prediction is obtained.
Fig. 4 (a) is Y1Rear 60 months prediction results and other existing method prediction results comparison diagram；Abscissa table
Show month, ordinate indicates corresponding square mean error amount；Fig. 4 (b) is Y2Rear 60 months prediction results of application are existing with other
There is the comparison diagram of method prediction result；Abscissa indicates month, and ordinate indicates corresponding square mean error amount；Fig. 4 (c) is Y3's
The comparison diagram of 60 months prediction results and other existing method prediction results afterwards；Abscissa indicates month, ordinate expression pair
The square mean error amount answered；Fig. 4 (d) is Y4Rear 60 months prediction results and other existing method prediction results comparison diagram；
Abscissa indicates month, and ordinate indicates corresponding square mean error amount；Mean square error is smaller, indicates that precision of prediction is higher.From Fig. 4
(a) to Fig. 4 (d) as can be seen that the prediction result of application method of the invention will significantly be better than exponential smoothing on the whole
(ES), autoregression model (AR), autoregressive moving-average model (ARMA) and autoregressive moving-average model (ARIMA) is integrated etc.
Method.
What table 1 indicated is the method for the present invention and exponential smoothing (ES), autoregression model (AR), autoregressive moving average mould
Type (ARMA) and integrate the Y that autoregressive moving-average model (ARIMA) obtains1、Y2、Y3、Y4MAE, MAPE, RMSE；
Table 1
Can be obtained by table 1, with exponential smoothing (ES), autoregression model (AR), autoregressive moving-average model (ARMA) and
It integrates the methods of autoregressive moving-average model (ARIMA) to compare, the present invention is missed in mean absolute error, average absolute percentage
Significant advantage is all had under the Measure Indexes of difference and root-mean-square error.
Claims (4)
1. a kind of method predicted ground-source heat pump system thermal energy conversion efficiency with operating condition input, which is characterized in that
It comprises the following steps that
A, thermal energy data pre-processes
1) thermal energy data is cleaned；
2) seasonal decomposition and preliminary analysis are carried out to the time series of thermal energy conversion efficiency；
B, the autoregression model with the input of external duty parameter is established
3) building has the autoregression model frame of external duty parameter；
4) basic classification regression tree is constructed, random forest, the prediction model under the different external duty parameters of training are generated, selection makes
The minimum tree for the tree that error no longer changes；
C, thermal energy conversion efficiency is predicted
According to trained prediction model, the time series of external duty parameter and thermal energy conversion efficiency is integrated into input and is carried out in advance
It surveys, exports prediction result and precision of prediction.
2. it is according to claim 1 it is a kind of with operating condition input ground-source heat pump system thermal energy conversion efficiency is predicted
Method, which is characterized in that the step A, thermal energy data pretreatment,
Thermal energy data includes the time series and outside duty parameter of thermal energy conversion efficiency；
The time series of thermal energy conversion efficiency includes: heat pump unit power consumption Y1, GSHP system power consumption Y2, unit COP Y3、
GSHP system COP Y4, Y1、Y2、Y3、Y4In each data be length be L time series, L > 24；
External duty parameter includes: distribution of boreholes form X1, boring radius X2, pipe laying depth X3, pipe laying longitudinal pitch X4, pipe laying it is horizontal
To spacing X5, pipe laying quantity X6, manage in packing material thermal coefficient X7, U-tube nominal outside diameter X8, U-tube spacing X9, distal end ground
Temperature X10, thermal conductivity X11With circulating liquid type X12；A shared M initial data is set, that is, shows in a matrix it is M row,
L*4+12 column；It comprises the following steps that
A, original M initial data is read, detects the value of NAN value and nonumeric form, deletes NAN value and nonumeric form
Row where being worth；
B, Y is taken out respectively from the step a data that obtain that treated1、Y2、Y3、Y4Visualization display is done, judges to exist abnormal
Numerical value, the row where the numerical value of suppressing exception；
C, the Y for obtaining step b1、Y2、Y3、Y4Four values do seasonal decomposition respectively, resolve into trend term T, season S, period
Item C, stochastic error I, for a time series Yj, j=1,2,3,4, it is expressed as { yj,t, t=1,2 ..., L }, L indicates heat
The length of the time series of energy transfer efficiency, indicates time series YjIn the occurrence of t moment thermal energy conversion efficiency；Seasonality point
Solution model is Yj=Tj+Sj+Cj+Ij, trend term TjIt is obtained by the way of sliding average, formula is as follows:
In formula (I), p is the lag item of sliding average, time series YjIt is T in the value of the trend term at t+1 momentj,t+1；
Season SjIt is obtained by trend extrapolation partition method, the time series of thermal energy conversion efficiency is subtracted after trend term monthly
It is averaged, then subtracts each other item by item with average value, just obtain season Sj；
Periodic term CjIt is obtained using period map method: by thermal energy conversion efficiency time series YjIt subtracts trend term and is carried out after season
Decentralization processing, then carries out Fourier expansion, obtains its maximum sinusoidal component AjWith the coefficient B of cosine componentj,
Next, obtaining t moment time series YjThe value of periodic termτ refers to the period；
Trend term T is subtracted from original time seriesj, season Sj, periodic term CjAfterwards, stochastic error I is obtainedj；
D, seasonal decomposition result is obtained in analytical procedure c, tentatively judges the tendency situation of the time series of thermal energy conversion efficiency
And seasonal characteristics.
3. it is according to claim 1 it is a kind of with operating condition input ground-source heat pump system thermal energy conversion efficiency is predicted
Method, which is characterized in that the step B, establish have external duty parameter autoregression model, comprising:
E, the autoregression model frame for having external duty parameter is established: by one d dimensional vector x table of d external duty parameter
Show, Random Forest model is indicated with f (), obtains the predicted value of t moment using historical data and external duty parameterSuch as formula
(II) shown in:
In formula (II), q is lag item, and q historical data is predicted before expression t moment；
F, the autoregression model frame obtained according to step e establishes the Random Forest model for specifically having external duty parameter；
M are randomly selected from all (d+1) × q+d attributes, then are selected optimal dividing attribute from this m attribute and divided
Split, optimal dividing Attributes Splitting at two child nodes sum of variance it is minimum, set the attribute set G for some node,
Using one of attribute g as optimal dividing attribute, by this node split at D1,D2Two child nodes, then optimal dividing attribute
As shown in formula (III):
In formula (III), V (D1),V(D2) refer to D1,D2Variance, g*Refer to optimal Split Attribute；
It is iteratively generating subtree, until cannot divide, ultimately generates a regression tree；
K regression tree is generated, this obtained K regression result is averaged after forming random forest, has just obtained t moment
Predicted valueAs shown in formula (IV):
In formula (IV), fkThe regression result of () expression kth regression tree.
4. it is according to claim 1 to 3 it is a kind of with operating condition input to ground-source heat pump system thermal energy conversion efficiency into
The method of row prediction, which is characterized in that the step C turns external duty parameter and thermal energy according to trained prediction model
The time series integration input for changing efficiency is predicted, is exported prediction result and precision of prediction, is comprised the following steps that
To be divided into length be L to the time series of thermal energy conversion efficiency for being L by length1Historical data and length be L2Future it is true
Real observation data, i.e., with preceding L1A historical data, L after prediction2A future concept of reality measured data, the result that prediction is obtained and true
Real observation compares, and obtains precision of prediction, and the standard of predictive metrics order of accuarcy includes formula (V), formula (VI), formula (VII):
Formula (V), formula (VI), in formula (VII),
MAE, that is, mean absolute error is being averaged for the inclined absolute value of the difference of all single observations and true value；
MAPE, that is, average absolute percentage error is the absolute value of true value percentage shared by all single observations and true value deviation
Be averaged；
RMSE, that is, root-mean-square error is that the deviation of all single observations and true value square takes after mean value sqrt again；
N indicates the number of the time series of the thermal energy conversion efficiency to be predicted, L2Indicate the length of predicted time series,
WithRespectively indicate the predicted value and true observation of i-th of time series of t moment；
It is L that the N number of time series that will be predicted, which is divided into length,1Historical data and length be L2The following concept of reality measured data after,
By N × L1Matrix be sent into trained model after obtain N × L2A prediction data is obtained by formula (V), formula (VI), formula (VII)
To precision of prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043545.3A CN109767044B (en) | 2019-01-17 | 2019-01-17 | Method for predicting heat energy conversion efficiency of ground source heat pump system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043545.3A CN109767044B (en) | 2019-01-17 | 2019-01-17 | Method for predicting heat energy conversion efficiency of ground source heat pump system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109767044A true CN109767044A (en) | 2019-05-17 |
| CN109767044B CN109767044B (en) | 2023-04-07 |
Family
ID=66454062
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910043545.3A Expired - Fee Related CN109767044B (en) | 2019-01-17 | 2019-01-17 | Method for predicting heat energy conversion efficiency of ground source heat pump system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109767044B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110795846A (en) * | 2019-10-29 | 2020-02-14 | 东北财经大学 | Construction method of boundary forest model, updating method of multi-working-condition soft computing model for complex industrial process and application of updating method |
| CN118199130A (en) * | 2024-05-17 | 2024-06-14 | 西安热工研究院有限公司 | Super capacitor energy storage capacity distribution method and system considering history error influence |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101438334A (en) * | 2006-03-03 | 2009-05-20 | 因瑞克斯有限公司 | Dynamic time series prediction of future traffic conditions |
| CN107895214A (en) * | 2017-12-08 | 2018-04-10 | 北京邮电大学 | A kind of multivariate time series Forecasting Methodology |
| CN108022001A (en) * | 2017-09-20 | 2018-05-11 | 河海大学 | Short term probability density Forecasting Methodology based on PCA and quantile estimate forest |
-
2019
- 2019-01-17 CN CN201910043545.3A patent/CN109767044B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101438334A (en) * | 2006-03-03 | 2009-05-20 | 因瑞克斯有限公司 | Dynamic time series prediction of future traffic conditions |
| CN108022001A (en) * | 2017-09-20 | 2018-05-11 | 河海大学 | Short term probability density Forecasting Methodology based on PCA and quantile estimate forest |
| CN107895214A (en) * | 2017-12-08 | 2018-04-10 | 北京邮电大学 | A kind of multivariate time series Forecasting Methodology |
Non-Patent Citations (2)
| Title |
|---|
| ZHAOYI ZHUANG 等: "Accurately predicting heat transfer performance of ground heat exchanger for ground-coupled heat pump systems using data mining methods", 《SPRINGER》 * |
| 庄兆意 等: "变频技术在污水源热泵系统中的应用研究", 《可再生能源》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110795846A (en) * | 2019-10-29 | 2020-02-14 | 东北财经大学 | Construction method of boundary forest model, updating method of multi-working-condition soft computing model for complex industrial process and application of updating method |
| CN118199130A (en) * | 2024-05-17 | 2024-06-14 | 西安热工研究院有限公司 | Super capacitor energy storage capacity distribution method and system considering history error influence |
| CN118199130B (en) * | 2024-05-17 | 2024-08-20 | 西安热工研究院有限公司 | Super capacitor energy storage capacity distribution method and system considering history error influence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109767044B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jing et al. | Distributed or centralized? Designing district-level urban energy systems by a hierarchical approach considering demand uncertainties | |
| CN110906571B (en) | Solar heat pump hot water system control strategy optimization method based on machine learning | |
| WO2021208383A1 (en) | Quantitative economic evaluation method for regional geothermal heating | |
| Fan et al. | A graph mining-based methodology for discovering and visualizing high-level knowledge for building energy management | |
| Mathioulakis et al. | Artificial neural networks for the performance prediction of heat pump hot water heaters | |
| Ahmed et al. | Optimal design, operational controls, and data-driven machine learning in sustainable borehole heat exchanger coupled heat pumps: Key implementation challenges and advancement opportunities | |
| Ozgener et al. | Monitoring of energy exergy efficiencies and exergoeconomic parameters of geothermal district heating systems (GDHSs) | |
| WO2024040608A1 (en) | Model training method for energy management system, apparatus, and storage medium | |
| CN109767044A (en) | A method for predicting thermal energy conversion efficiency of ground source heat pump system with operating condition input | |
| Wang et al. | A load-complementarity combined flexible clustering approach for large-scale urban energy-water nexus optimization | |
| Ahmed et al. | Data-driven modelling techniques for earth-air heat exchangers to reduce energy consumption in buildings: a review | |
| Wang et al. | Scientometric of nearly zero energy building research: A systematic review from the perspective of co-citation analysis | |
| CN114186640A (en) | Prediction method for overall power generation capacity of regional hydropower | |
| Zhang et al. | Generic visual data mining-based framework for revealing abnormal operation patterns in building energy systems | |
| Zhang et al. | Clustering based on dynamic time warping to extract typical daily patterns from long-term operation data of a ground source heat pump system | |
| Yue et al. | Machine learning-based multi-performance prediction and analysis of Earth-Air Heat Exchanger | |
| CN109816166A (en) | A performance prediction method for ground source heat pump system | |
| CN116013436A (en) | An intelligent screening method for low melting point and high stability molten salt heat storage materials | |
| Alghamdi et al. | Predicting energy consumption using stacked LSTM snapshot ensemble | |
| Schlueter et al. | Analysis of georeferenced building data for the identification and evaluation of thermal microgrids | |
| CN110991870B (en) | An integrated energy system data processing and calculation method with intensive characteristics | |
| CN117769718A (en) | Modeling method and device of intelligent energy management system and storage medium | |
| CN110309944B (en) | Regional energy supply cold and hot load prediction method | |
| CN112365082A (en) | Public energy consumption prediction method based on machine learning | |
| Hanuszkiewicz-DrapaŁa et al. | Operation characteristics of heat pump systems with ground heat exchangers |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20230407 |