CN109816166B - Ground source heat pump system performance prediction method - Google Patents
Ground source heat pump system performance prediction method Download PDFInfo
- Publication number
- CN109816166B CN109816166B CN201910043376.3A CN201910043376A CN109816166B CN 109816166 B CN109816166 B CN 109816166B CN 201910043376 A CN201910043376 A CN 201910043376A CN 109816166 B CN109816166 B CN 109816166B
- Authority
- CN
- China
- Prior art keywords
- data
- prediction
- decision tree
- gshp
- model
- 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
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a performance prediction method of a ground source heat pump system, which comprises the following steps: (1) Preprocessing and seasonally decomposing heat transfer performance data of the GSHP system; (2) Constructing a data structure with GSHP system working parameter input by adopting a seasonal monthly mode; (3) Constructing a decision tree set model which has external system working parameter input and can be used for time series regression prediction, namely a prediction model; (4) Performing data reconstruction on the test set data by adopting the same method as the step (2), and performing performance prediction on the decision tree set model obtained by training in the step (3) by adopting an autoregressive mode; (5) And comparing the prediction data obtained by the test set through the decision tree set model with the real data to measure the prediction effect of the decision tree set model. The prediction method is simple and easy to operate. The invention also solves the problem that the traditional time sequence analysis method can not predict the time sequence with external parameters.
Description
Technical Field
The invention relates to a performance prediction method of a ground source heat pump system, and belongs to the technical field of performance prediction of ground source heat pump systems.
Background
A Ground Source Heat Pump (GSHP) is a high-efficiency energy-saving environment-friendly air conditioning technology which takes underground shallow geothermal resources as cold and Heat sources and utilizes a Heat Pump technology to realize the purposes of providing winter heating, summer cooling and all-year-round life hot water for buildings. GSHP may achieve higher efficiency and more consistent performance than traditional air-source heat pumps. Therefore, ground source heat pumps are receiving increasing attention as a new technology for heating and cooling buildings that is environmentally friendly, energy saving and economically feasible.
In practical engineering application, the initial investment cost of the ground source heat pump system is very high, and unreasonable design not only can cause waste of capital, but also can influence the application and popularization of the GSHP system. Therefore, the accurate prediction of the heat transfer performance of the ground heat pump system buried pipe heat exchanger becomes a key for promoting the technical development and application process thereof. However, the factors affecting the GSHP system are many and complex, including not only the system design factors such as drilling parameters and U-tube parameters, but also the external factors such as soil heat transfer properties, and in different regions, the weather change conditions within one year will also have a large influence on the system, so it is difficult to design a complete prediction model considering all the factors.
Currently, most of research on the performance evaluation of the GSHP system or the optimal design of the system is realized by methods of physical system modeling and energy analysis. Among the representative methods of comparison are the energy analysis model and
and analyzing the model. Although such methods based on energy analysis have already had a great deal of theoretical basis and experimental results, the following disadvantages still remain. Firstly, in the process of establishing a physical model, some influence factors are simplified or even ignored to different degrees, so that the problem of heat imbalance caused by inaccurate parameters is likely to occur in practical application; secondly, most of the research is carried out aiming at specific systems and operation modes, so that the established model and corresponding conclusions can not be applicable to other examples; finally, when the method is used for performance prediction of the system, the processes of physical analysis and model establishment of the system are quite complex, and the universality and the practicability in the actual life are poor.
With the explosive development of machine learning and big data analysis techniques, some data mining based methods are also beginning to be applied in the prediction of heat transfer performance of GSHP systems. Unlike conventional methods, this type of method is based on data mining techniques rather than physical modeling and energy analysis. Although relatively few research efforts have been made to date, it has been shown that data mining based methods can be fully applied to heat transfer performance prediction for GSHP systems. Such as Partial Least Squares Regression (PLSR), support Vector Machines (SVMs), random Forest (RF), and M5 models have all been applied to predict heat transfer performance of GSHP systems. However, although the methods avoid a complex physical modeling analysis process, the method only considers the influence of historical data in the time series of the performance data of the GSHP system on the future performance of the system, neglects the influence of parameters (such as system operating condition parameters) of the system, and has a non-ideal effect of predicting the heat transfer performance of the GSHP system.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a performance prediction method of a ground source heat pump system.
Summary of the invention:
the invention considers the influence of the historical performance data of the GSHP system and the working parameters of the system on the prediction of the future performance at the same time, firstly, the working parameters of the system are introduced into the historical performance data of the GSHP system to form a system data structure with external parameter input; and then generating a decision tree set model with external parameter input by a greedy mode through a multi-iteration optimization objective function based on a gradient lifting decision tree algorithm. The model is composed of a plurality of decision trees, each iteration is to add an optimal decision tree on the basis of the original model, and learn the residual error between the real value and the sum of all previous tree conclusions. And continuously reducing the residual error between the true value and the predicted value through multiple times of optimization training so as to improve the performance of the model. And finally, realizing regression prediction of the performance time sequence of the GSHP system with the external working parameters by using the prediction model obtained by training in an autoregressive prediction mode.
Interpretation of terms:
1. seasonal-Year, for monthly time series data of consecutive years, the same month data of different years are normalized and combined in month unit, and prediction of the same month in the next Year is performed according to the normalized and combined month data (for example, the data of 12 months in the previous 25 years is used for predicting the data of 12 months in the 26 th Year).
2. Within-Yeast, for the time series data of months of consecutive years, for the prediction of data of a specific month, the prediction is realized by using all known month data of the same Year except the month (for example, the data of 12 months in 26 years is predicted by using the data of 11 months before 26 years).
3. The forward distribution algorithm is that only one basis function and the coefficient thereof are learned in each step from front to back and are commonly used for optimization training of an objective function, so that the complexity of optimization is simplified.
4. The time sequence is formed by arranging the numerical values of certain statistic of a certain phenomenon at different moments according to the time sequence. A time series { X (T), T =1,2,3,.. T } (where T denotes a specific time, X (T) denotes an observed value corresponding to time T, and T denotes a length of the time series), and if a mean value and a variance of the time series are constants independent of time T and a covariance of the time series is a constant independent of time T and only related to time interval k, the time series is called a stationary time series.
The technical scheme of the invention is as follows:
a performance prediction method for a ground source heat pump system comprises the following steps:
(1) Preprocessing and analyzing heat transfer performance data of a GSHP system, completing data cleaning, carrying out seasonal decomposition on the heat transfer performance data, and analyzing the stability of the heat transfer performance data;
(2) Dividing the heat transfer performance data of the GSHP system processed in the step (1) into a training set and a testing set, constructing a data structure with GSHP system working parameter input, and introducing 12 working parameters of the GSHP system, including a drilling hole distribution form X 1 Radius of borehole X 2 Depth X of pipe laying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal spacing X of buried pipe 6 Thermal conductivity coefficient X of filling material in tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Remote rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 Reconstructing historical performance data of the GSHP system in a seasonal monthly mode;
(3) Based on a gradient lifting decision tree algorithm, a decision tree set model which has external system working parameter input and can be used for time series regression prediction, namely a prediction model, is generated in a mode of constructing an objective function through multiple times of optimization iterative training, and the performance of the GSHP system is predicted;
(4) Performing data reconstruction on the test set data by adopting the same method as the step (2), and performing performance prediction on the decision tree set model obtained by training in the step (3) by adopting an autoregressive mode;
(5) And comparing the prediction data obtained by the test set through the decision tree set model with the real data to measure the prediction effect of the decision tree set model.
Preferably, in step (1), the preprocessing and analyzing of the heat transfer performance data of the GSHP system includes:
A. simulating and acquiring heat transfer performance data of the GSHP system; the heat transfer performance data of the GSHP system includes a plurality of sets of time series data, each set of time series data including 12 system operating parameters X 1 ～X 12 And Y 1 ～Y 4 Time series of (2), Y 1 The unit is kWh which is the power consumption of the heat pump unit; y is 2 The total electricity consumption of the GSHP system is expressed in kWh; y is 3 The heat pump unit is used for controlling the ratio of the heating capacity or the refrigerating capacity of the heat pump unit to the operating power of the heat pump unit; y is 4 The ratio of the heating capacity or the refrigerating capacity of the heat pump unit to the total power of the system is referred to;
B. detecting NAN values and non-numerical values of the heat transfer performance data obtained by simulation, and deleting the rows where the NAN values and the non-numerical values are located;
C. b, performing visualization operation on the heat transfer performance data of the GSHP system obtained in the step B, checking the overall trend of the GSHP system, judging abnormal numerical values according to professional theoretical knowledge, and deleting lines where the abnormal numerical values are located; for example: four conversion efficiencies Y 1 ～Y 4 Should be non-negative and should be rejected for some values of apparent anomaly less than zero; except for the months 4 and 10, the power consumption Y of the heat pump unit 1 Because the system is in a closed state and the value is zero, the values of other months cannot exceed 35000kwh, and the values which are obviously abnormal and are more than 60000kwh are removed;
D. considering that the heat transfer performance data of the GSHP system is strongly influenced by weather such as temperature and is a periodic time series containing seasonal factors, the time series data obtained after the processing of step C is subjected to seasonal decomposition, and the time series data is decomposed into a seasonal item S, a trend item T and a random error item I, and for a time series data Y, the seasonal decomposition model is as follows:
Y t ＝S t +T t +I t (I)
in the formula (I), t is a specific time, Y t Denotes the observed value corresponding to time t, S t Refers to a seasonal term reflecting the cyclic variation of a time series within one year, T t The trend item can capture the long-term change of the time series, I t The random error term can capture changes which cannot be explained by trend or seasonal effect, and formula (I) shows that the observed value corresponding to the time sequence t is equal to the sum of the trend value, the seasonal effect and the random effect at the moment. Here we use the LOESS smoothing method to decompose the time series into trend terms, seasonal terms, and random terms.
E. And D, analyzing the seasonal decomposition result obtained in the step D, judging the stationarity of the seasonal decomposition result, if the seasonal decomposition result is not stationarity, stabilizing the time sequence data in a differential mode, and if the seasonal decomposition result is stationarity, entering the step (2). Providing theoretical basis and guarantee for the following modeling prediction work.
Preferably, in step (2), the dividing of the training set and the test set to construct the data structure with the input of the GSHP system operating parameters includes:
F. e, dividing the heat transfer performance data of the GSHP obtained in the step E to obtain a training set and a testing set; for example, the previous 4000 sets of heat transfer performance data were used as a training set, and the remainder were used as a test set.
G. On the basis of step F, each set of time series data in the training set is represented as x i ＝{y i1 ,y i2 ,...,y it I refers to the ith group of time sequence data, and the value range of i is 1<i<Total number of training set samples, y it Finger x i The observed value corresponding to the t moment in (1);
introducing GSHP system working parameter c it And for each observed value y it Expanding to form a new observation value with parameters as
c it Denotes y it A corresponding one of the GSHP system parameter vectors,
to is that
Respectively representing borehole patterns X 1 Radius of borehole X 2 Depth X of pipe laying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal spacing X of buried pipe 6 Thermal conductivity coefficient X of filling material in tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Far end rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 These 12 system operating parameters; the expansion process is summarized as follows:
after introducing the system operating parameters, new time series data in the form of:
x i new ＝{s i1 ,s i2 ,s i3 ,...,s it }＝{(c i1 ,y i1 ),(c i2 ,y i2 ),...,(c it ,y it )} (Ⅲ)
in the formula (III), the compound represented by the formula (III),
the method comprises the steps that new time sequence data are obtained by introducing ith group of time sequence data into GSHP system working parameters in a training set and expanding the time sequence data;
H. on the basis of the step G, the seasonal characteristics of the performance data are considered more, a seasonal month-by-month method of Between-year is adopted for different month data, the whole data structure X is divided into 12 blocks by taking a month as a unit, and the whole data structure X is as follows:
in the formula (IV), the compound is shown in the specification,
g, new time series data obtained by expansion in the step G are referred to;
after dividing the overall data structure X into 12 blocks, X = [ B 1 B 2 ...B 12 ]；
Setting the reconstructed data of the mth month as B m M is more than or equal to 1 and less than or equal to 12, and all new time sequence data x in the training set i new In the m-th band, the observation value
Starting data, at a sampling rate of 12 (12 months in 1 year), for x i new Sampling for p times to obtain observation values with parameters of all m months
Forming a seasonal monthly time series
Finally obtaining a two-dimensional matrix B m The method comprises the following steps:
in the step (V), the reaction mixture is,
i refers to the number of training set samples, p refers to the number of samples, 1<p<All years of observation, R refers to the real number field; b is m Any one of the row vectors
From the ith set of new time series performance data x i new Observation value with reference of all m months in
Composition m 1 ～m p Finger pair x i new M months of (1) to p samples are taken at intervals of 12; b is m The expression comprises 1 to i groups of time sequence performance data x in the training set i new The belt parameter observation values of all m months in the period.
The reconstruction form changes the conventional method based on the Within-Yeast data structure, further considers the seasonal characteristics of system performance data, integrates the data corresponding to different years of each month respectively, and establishes a data structure with external parameter input.
Preferably, in step (3), the establishing a prediction model includes:
I. for different months, based on a gradient lifting decision tree algorithm, establishing an objective function containing a geothermal square loss term and a system normalization regularization term, wherein the specific form is as follows:
in the formula (VI), theta represents parameters in the prediction model and refers to the tree structure and the tree leaf node score of each decision tree; n represents the number of samples; y is i Representing the true value of the ith sample;
representing a predicted value of the ith sample; t represents the number of the decision trees; f. of t Representing a specific decision tree; omega (f) t ) The regularization item for expressing the normalization of the geothermal system comprises a penalty item for the node number of each decision tree and an L2 modular square of the leaf node fraction, and is used for restraining the complexity of the decision tree; the method avoids the overfitting of the trained model to the performance of the GSHP system under a certain group of specific working parameters, standardizes the generalization capability of the trained prediction model under different system working parameters, and has the following mathematical expression:
in the formula (VII), K, w j Respectively representing the leaf node number of each decision tree and the score on each leaf node;
lambda and gamma respectively represent punishment coefficients of the leaf node number and the leaf node fraction;
J. minimizing an objective function in an incremental training mode by adopting a forward distribution algorithm;
generating a decision tree set model with external system working parameter input through multiple optimization iterative training based on the data reconstructed in the step H; initializing a decision tree set model, then performing multiple rounds of iterative optimization, wherein in each round of iteration, an optimal decision tree is added to an original decision tree set model to continuously reduce residual errors of the original decision tree set model, and finally obtaining the decision tree set model with very strong learning capacity through multiple iterative optimization training. The specific training process is as follows:
in formula (VIII), x is different for different months i For reconstructing the corresponding blocks B in the overall data structure X m The ith row of(Vector)
Thus, through r rounds of incremental training, a decision tree set model which is composed of r decision trees and has external system working parameter input is obtained, and the decision tree set model simultaneously and fully considers the influence of GSHP system historical performance data and system working parameters on the future performance of the system, which is deficient in the traditional time series method. The mathematical expression of the final decision tree set model is as follows:
K. and (4) respectively establishing independent decision tree set models for different months by adopting the same method of the step I, J.
Preferably, in the step (4), the performance prediction is performed in an autoregressive manner, and includes:
l, performing data reconstruction on the test set data by adopting the same method as the step (2), and establishing a data structure with system working parameter input;
m, on the basis of the step K, for different months, performing regression prediction on the performance time series data of the GSHP system by adopting a nonlinear autoregressive mode to realize long-term and continuous prediction on the performance of the GSHP system, wherein the nonlinear autoregressive mode is as follows:
in the formula (X), the compound represented by the formula (X),
denotes the predicted value, x, of the mth month of the yth year (y-M.m) Represents 12 system parameters, y, corresponding to the mth month of the y-M year (y-M,m) The true value of the mth month in the y-M year is represented, M represents the number of lag terms, and f (-) is an arbitrary model technology; e.g. support vector machine, random forest, gradient boostingDecision trees, and the like.
Equation (X) shows that a prediction of the performance of a particular month of the next year can be made based on data from the same month in different years of the previous years.
And D, respectively predicting different months by using the decision tree set model obtained in the step K, integrating and reconstructing the performance data obtained by the first prediction to be used as the next model input, performing the second prediction, and completing the regression prediction of the performance time series data of the GSHP system through the sliding prediction in a nonlinear autoregressive mode.
The invention has the beneficial effects that:
1. the invention adopts a method of gradient lifting decision tree time series regression, and realizes long-term continuous prediction of the heat transfer performance of the ground source heat pump system. Compared with the traditional method, the method does not need complex physical system modeling, energy analysis and system parameter calculation, and has simple prediction method and easy operation.
2. According to the method, through reconstruction of system performance data and model training, a set decision tree prediction model with external parameter input is established, the model considers the influence of historical performance data and system working parameters on the system non-performance, the regression prediction of a time sequence is realized in a nonlinear autoregressive prediction mode, and the problem that the traditional time sequence analysis method cannot predict the time sequence with external parameters is solved.
3. The prediction model established by the invention has certain generalization capability, can obtain more accurate prediction results, and has high prediction speed. Meanwhile, a new idea is provided for a time series analysis method with external parameters, and more factors influencing the heat transfer performance of the ground source heat pump system can be taken into consideration.
Drawings
Fig. 1 is a general flowchart of a method for predicting performance of a ground source heat pump system according to the present invention;
FIG. 2 is a schematic diagram of seasonal decomposition (in Y) 1 For example);
FIG. 3 is a diagram of the internal structure of the prediction model established in the present invention;
FIG. 4 (a) is a graph comparing the effect of the method of the present invention with that of the prior art exponential smoothing method (ES);
FIG. 4 (b) is a graph comparing the effect of the method of the present invention and the prior autoregressive model (AR);
FIG. 4 (c) is a graph comparing the effectiveness of the method of the present invention with a prior art autoregressive moving average model (ARMA);
FIG. 4 (d) is a graph comparing the effectiveness of the method of the present invention with that of a prior art autoregressive integrated moving average model (ARIMA);
Detailed Description
The present invention is described in detail below with reference to the drawings and examples, but is not limited thereto.
Example 1
A performance prediction method of a ground source heat pump system is shown in FIG. 1, and comprises the following steps:
(1) Preprocessing and analyzing heat transfer performance data of a GSHP system, completing data cleaning, carrying out seasonal decomposition on the heat transfer performance data, and analyzing the stability of the heat transfer performance data;
(2) Dividing the heat transfer performance data of the GSHP system processed in the step (1) into a training set and a testing set, constructing a data structure with GSHP system working parameter input, and introducing 12 working parameters of the GSHP system, including a drilling hole distribution form X 1 Radius of borehole X 2 Depth X of pipe laying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal distance X between embedded pipes 6 Thermal conductivity coefficient X of filling material in tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Far end rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 Reconstructing historical performance data of the GSHP system in a seasonal monthly mode;
(3) Based on a gradient lifting decision tree algorithm, a decision tree set model which has external system working parameter input and can be used for time series regression prediction, namely a prediction model, is generated in a mode of constructing an objective function through multiple times of optimization iterative training, and the performance of the GSHP system is predicted;
(4) Performing data reconstruction on the test set data by adopting the same method as the step (2), and performing performance prediction on the decision tree set model obtained by training in the step (3) by adopting an autoregressive mode;
(5) And comparing the prediction data obtained by the test set through the decision tree set model with the real data to measure the prediction effect of the decision tree set model.
Example 2
The performance prediction method of the ground source heat pump system according to embodiment 1 is characterized in that:
step (1), preprocessing and analyzing heat transfer performance data of a GSHP system, wherein the preprocessing and analyzing comprises the following steps:
A. simulating and acquiring heat transfer performance data of the GSHP system; the heat transfer performance data of the GSHP system includes several sets of time series data, each set of time series data including 12 system operating parameters X 1 ～X 12 And Y 1 ～Y 4 Time series of (2), Y 1 The unit is kWh which means the power consumption of the heat pump unit; y is 2 The total electricity consumption of the GSHP system is expressed in kWh; y is 3 The heat pump unit is used for controlling the ratio of the heating capacity or the refrigerating capacity of the heat pump unit to the operating power of the heat pump unit; y is 4 The ratio of the heating capacity or the refrigerating capacity of the heat pump unit to the total power of the system is referred to; 5000 groups of system performance data of the ground source heat pump system in 30 years (360 months) are obtained through simulation of software 'terrestrial heat star'. The system operating parameters include: borehole pattern X 1 Radius of borehole X 2 And the depth X of pipe burying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal spacing X of buried pipe 6 And the heat conductivity coefficient X of the filling material in the tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Remote rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 . The thermal energy conversion efficiency includes: power consumption Y of heat pump unit 1 The unit: kWh; total power consumption Y of the system 2 The unit: kWh; the ratio Y of the heating or cooling capacity of the heat pump unit to the operating power of the heat pump unit 3 (ii) a The ratio Y of the heating or cooling capacity of the heat pump unit to the total power of the system 4 . 5000 pieces of original data in total, each piece of data comprises 12 system working parameters X 1 ～X 12 And Y 1 ～Y 4 Four time sequences of length 360, i.e., 5000 rows, 12+360 + 4 columns are shown in the matrix.
B. Detecting NAN values and non-numerical values of the heat transfer performance data obtained by simulation, and deleting the rows where the NAN values and the non-numerical values are located;
C. carrying out visual operation on the heat transfer performance data of the GSHP system obtained in the step B, checking the overall trend of the GSHP system, judging abnormal numerical values according to professional theoretical knowledge, and deleting the lines where the abnormal numerical values are located; for example: four conversion efficiencies Y 1 ～Y 4 Should be non-negative and should be rejected for some values of apparent anomaly less than zero; except for 4 months and 10 months, the heat pump set power consumption Y 1 The value of (1) is zero when the system is in a closed state, the values of other months cannot exceed 35000kwh, and the values with obvious abnormality above 60000kwh are removed;
D. considering that the heat transfer performance data of the GSHP system is strongly influenced by weather such as temperature and is a periodic time series containing seasonal factors, the time series data obtained after the processing of step C is subjected to seasonal decomposition, and the time series data is decomposed into a seasonal item S, a trend item T and a random error item I, and for a time series data Y, the seasonal decomposition model is as follows:
Y t ＝S t +T t +I t (I)
in the formula (I), t is a specific time, Y t Denotes the observed value corresponding to time t, S t Refers to a seasonal term reflecting the cyclic variation of a time series within one year, T t Long-term variation of time series that the trend term can capture, I t The random error term can capture changes which cannot be explained by trend or seasonal effect, and formula (I) shows that the observed value corresponding to the time sequence t is equal to the sum of the trend value, the seasonal effect and the random effect at the moment. Here we use the LOESS smoothing method to decompose the time series into trend terms, seasonal terms, and random terms.
E. And D, analyzing the seasonal decomposition result obtained in the step D, judging the stationarity of the seasonal decomposition result, if the seasonal decomposition result is not stationarily, stabilizing the time sequence data in a differential mode, and if the seasonal decomposition result is stationarily, entering the step (2). Providing theoretical basis and guarantee for the following modeling prediction work.
Example 3
The performance prediction method of the ground source heat pump system according to embodiment 2 is characterized in that:
the step (2) of dividing the training set and the test set and constructing the data structure with the GSHP system working parameter input comprises the following steps:
F. e, dividing the heat transfer performance data of the GSHP obtained in the step E to obtain a training set and a testing set; for example, the previous 4000 sets of heat transfer performance data were used as a training set, and the remainder were used as a test set.
G. On the basis of step F, each set of time series data in the training set is represented as x i ＝{y i1 ,y i2 ,...,y it I refers to the ith group of time sequence data, and the value range of i is 1<i<Total number of training set samples, y it Finger x i The observed value corresponding to the t moment in (1);
introducing GSHP system working parameter c it For each observed value y it Expanding to form new observation value with parameter
c it Denotes y it A corresponding one of the GSHP system parameter vectors,
to
Respectively representing the borehole pattern X 1 Radius of borehole X 2 Depth X of pipe laying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal spacing X of buried pipe 6 Thermal conductivity coefficient X of filling material in tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Far end rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 These 12 system operating parameters; the expansion process is summarized as follows:
after introducing the system operating parameters, new time series data in the form of:
x i new ＝{s i1 ,s i2 ,s i3 ,...,s it }＝{(c i1 ,y i1 ),(c i2 ,y i2 ),...,(c it ,y it )} (Ⅲ)
in the formula (III), the reaction mixture is,
the method comprises the steps that new time sequence data are obtained by introducing ith group of time sequence data into GSHP system working parameters in a training set and expanding the time sequence data;
H. on the basis of the step G, the seasonal characteristics of the performance data are considered more, a seasonal month-by-month method of Between-year is adopted for different month data, the whole data structure X is divided into 12 blocks by taking a month as a unit, and the whole data structure X is as follows:
in the formula (IV), the compound is shown in the specification,
g, new time series data obtained by expansion in the step G are referred to;
after dividing the overall data structure X into 12 blocks, X = [ B = 1 B 2 ...B 12 ]；
Setting the reconstructed data of the mth month as B m M is more than or equal to 1 and less than or equal to 12, and all new time sequence data x in the training set i new In the first placem zones of observation
As starting data, with 12 (1 year 12 months) as sampling rate, for x i new Sampling for p times to obtain observation values with parameters of all m months
Forming a seasonal monthly time series
Finally, a two-dimensional matrix B is obtained m The method comprises the following steps:
(V) in the above-mentioned step (V),
i refers to the number of training set samples, p refers to the number of samples, 1<p<All years of observation, R refers to the real number field; b is m Any one of the row vectors
From the ith set of new time series performance data x i new Observation value with reference of all m months in
Composition m 1 ～m p Finger pair x i new M months of (1) to p samples are taken at intervals of 12; b is m The expression comprises 1 to i groups of time sequence performance data x in the training set i new The belt parameter observation values of all m months in the period.
The reconstruction form changes the conventional method based on the Within-Yeast data structure, further considers the seasonal characteristics of system performance data, integrates the data corresponding to different years of each month respectively, and establishes a data structure with external parameter input.
Example 4
The performance prediction method of the ground source heat pump system according to embodiment 3 is characterized in that:
in the step (3), establishing a prediction model includes:
I. for different months, based on a gradient lifting decision tree algorithm, establishing an objective function containing a geothermal square loss term and a system normalization regularization term, wherein the specific form is as follows:
in the formula (VI), theta represents parameters in the prediction model and refers to the tree structure and the tree leaf node score of each decision tree; n represents the number of samples; y is i Representing the true value of the ith sample;
representing a predicted value of the ith sample; t represents the number of the decision trees; f. of t Representing a specific decision tree; omega (f) t ) The regularization item for expressing the normalization of the geothermal system comprises a penalty item for the node number of each decision tree and an L2 modular square of the leaf node fraction, and is used for restraining the complexity of the decision tree; the performance of a model obtained by training under the condition that the model is over-fitted with a certain set of specific working parameters of the GSHP system is avoided, the generalization capability of a prediction model obtained by training under different system working parameters is standardized, and the mathematical expression of the generalization capability is as follows:
in the formula (VII), K, w j Respectively representing the leaf node number of each decision tree and the score on each leaf node; lambda and gamma respectively represent punishment coefficients of the leaf node number and the leaf node fraction;
J. minimizing an objective function in an incremental training mode by adopting a forward distribution algorithm;
generating a decision tree set model with external system working parameter input through multiple optimization iterative training based on the data reconstructed in the step H; initializing a decision tree set model, then performing multiple rounds of iterative optimization, wherein in each round of iteration, an optimal decision tree is added to an original decision tree set model to continuously reduce residual errors of the original decision tree set model, and finally obtaining the decision tree set model with very strong learning capacity through multiple iterative optimization training. The specific training process is as follows:
in formula (VIII), x is different for different months i For reconstructing the corresponding blocks B in the overall data structure X m Ith row vector of
Thus, through r rounds of incremental training, a decision tree set model which is composed of r decision trees and has external system working parameter input is obtained, and the decision tree set model simultaneously and fully considers the influence of GSHP system historical performance data and system working parameters on the future performance of the system, which is deficient in the traditional time series method. The mathematical expression of the final decision tree set model is as follows:
K. and (4) respectively establishing independent decision tree set models for different months by adopting the same method of the step I, J.
Example 5
The performance prediction method of the ground source heat pump system according to embodiment 4 is characterized in that:
in the step (4), the performance prediction is performed in an autoregressive manner, and the performance prediction method includes:
l, performing data reconstruction on the test set data by adopting the same method as the step (2), and establishing a data structure with system working parameter input;
m, on the basis of the step K, for different months, performing regression prediction on the performance time series data of the GSHP system by adopting a nonlinear autoregressive mode to realize long-term and continuous prediction on the performance of the GSHP system, wherein the nonlinear autoregressive mode is as follows:
in the formula (X), the compound represented by the formula (X),
denotes the predicted value, x, of the mth month of the yth year (y-M.m) Represents 12 system parameters, y, corresponding to the mth month of the y-M year (y-M,m) The true value of the mth month in the y-M year is represented, M represents the number of lag terms, and f (-) is an arbitrary model technology; such as support vector machines, random forests, gradient boosting decision trees, and the like.
Equation (X) shows that a prediction of the performance of a particular month of the next year can be made based on data from the same month in different years of the previous years.
And D, respectively predicting different months by using the decision tree set model obtained in the step K, integrating and reconstructing the performance data obtained by the first prediction to be used as the next model input, performing the second prediction, and completing the regression prediction of the performance time series data of the GSHP system through the sliding prediction in a nonlinear autoregressive mode.
Example 6
The performance prediction method of the ground source heat pump system according to embodiment 5 is characterized in that:
step (1): taking Y1 as an example, the time series seasonal decomposition results are shown in fig. 2.
FIG. 2 shows Y 1 A trend term, a seasonal factor term, and a random error term. As can be seen from FIG. 2, Y 1 There was a slight decline in the trend term of (1) during the 360 months (30 years)The trend is consistent with the practical engineering application, and the cooling or heating power of the heat pump unit is actually reduced along with the increase of the service life. And its season term indicates Y 1 Has very obvious seasonal cycle characteristics; the random error term reflects a range of uncertainties. It can be seen from the figure that it fluctuates little at 0, and the influence of the random error term is small compared to the whole. According to the result of the seasonal decomposition, Y 1 ～Y 4 The four quantities are all stationary time series, which shows that the four quantities can be further subjected to modeling prediction research.
And (2) to (4) constructing a data structure with system working parameter input by taking the previous 4000 groups of time sequences as a training set and taking the next 1000 groups as a test set, establishing a prediction model in a form of optimizing a training objective function, and finally predicting the heat transfer performance of the GSHP system in the next 5 years (60 months) by adopting a nonlinear autoregressive prediction method.
And (5) comparing the predicted data obtained by the test set through the prediction model with the real data. Aiming at the actual demand of the regression prediction of the heat transfer performance of the GSHP system, the average absolute error (MAE), the average absolute error (MAPE) and the Root Mean Square Error (RMSE) are respectively calculated to measure the prediction performance of the established model:
n represents the number of observation values included in each time series, T represents the length of the time series to be predicted,
and
respectively representing the predicted value and the real value of the ith observation value at the time t. To measure the overall prediction accuracy of the future 60 months, Y is measured separately 1 ～Y 4 The prediction error was averaged over 60 months and the results are shown in table 1. In which fig. 4 shows the MAE results of the prediction method of example 2 and a comparison with other methods. FIG. 4 (a) is a comparison graph of the effect of the method of this embodiment and the effect of the prior art exponential smoothing method (ES); FIG. 4 (b) is a graph comparing the effect of the method of the present embodiment with that of the conventional autoregressive model (AR); FIG. 4 (c) is a graph comparing the effect of the method of the present embodiment with that of the prior art autoregressive moving average (ARMA); FIG. 4 (d) is a graph comparing the effect of the method of the present embodiment with that of the prior autoregressive integrated moving average model (ARIMA); with Y 1 For example, Y1-ARX is the method of this example;
TABLE 1
As can be seen from table 1, the method of the present invention has significant advantages of minimum mean absolute error, minimum mean percentage absolute error and minimum root mean square error, compared with Exponential Smoothing (ES), autoregressive model (AR), autoregressive moving average model (ARMA), and autoregressive integrated moving average model (ARIMA).
In practical engineering applications, four energy conversion efficiencies representing the heat transfer performance of the system may exhibit different performance at different operating parameters. Therefore, ignoring the prediction of these parameters, the result is general, as it does not take into account the specificity of the different systems. The invention establishes a set decision tree prediction model with external parameter input by reconstructing system performance data and multi-round model training, and the model considers the influence of historical performance data and system working parameters on the system non-performance on the basis of considering more seasonal characteristics of the system performance data, which is lacking in the traditional time series analysis method. The model realizes regression prediction of the time sequence in a nonlinear autoregressive prediction mode, and overcomes the defect that the traditional time sequence analysis method cannot predict the time sequence with external parameters.
Meanwhile, the prediction model is established by using an algorithm based on a gradient lifting decision tree, and the integrated machine learning method has stronger learning capacity and generalization capacity. Experimental results show that the performance prediction method of the ground source heat pump system based on the gradient lifting decision tree time sequence regression can effectively improve the prediction precision of the time sequence.
Claims (4)
1. A performance prediction method for a ground source heat pump system is characterized by comprising the following steps:
(1) Preprocessing and analyzing heat transfer performance data of a GSHP system, completing data cleaning, carrying out seasonal decomposition on the heat transfer performance data, and analyzing the stability of the heat transfer performance data;
(2) Dividing the heat transfer performance data of the GSHP system processed in the step (1) into a training set and a testing set, constructing a data structure with GSHP system working parameter input, and introducing 12 working parameters of the GSHP system, including a drilling hole distribution form X 1 Radius of borehole X 2 Depth X of pipe laying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal spacing X of buried pipe 6 Thermal conductivity coefficient X of filling material in tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Far end rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 Reconstructing historical performance data of the GSHP system in a seasonal monthly mode;
the step (2) of dividing the training set and the test set and constructing the data structure with the GSHP system working parameter input comprises the following steps:
F. e, dividing the heat transfer performance data of the GSHP obtained in the step E to obtain a training set and a testing set;
G. each set of time series data in the training set is represented as x i ＝{y i1 ,y i2 ,...,y it I refers to the ith group of time sequence data, and the value range of i is 1<i<Total number of training set samples, y it Finger x i The observed value corresponding to the t moment in (1);
introducing GSHP system working parameter c it For each observed value y it Expanding to form a new observation value with parameters as
c it Denotes y it A corresponding one of the GSHP system parameter vectors,
to
Respectively representing borehole patterns X 1 Radius of borehole X 2 Depth X of pipe laying 3 X number of buried pipes 4 Transverse spacing X of buried pipes 5 Longitudinal spacing X of buried pipe 6 Thermal conductivity coefficient X of filling material in tube 7 Nominal outer diameter X of U-shaped tube 8 Distance X between U-shaped tubes 9 Far end rock temperature X 10 Thermal conductivity X 11 And circulating liquid type X 12 These 12 system operating parameters; the expansion process is summarized as follows:
after introducing the system operating parameters, new time series data in the form of:
x i new ＝{s i1 ,s i2 ,s i3 ,...,s it }＝{(c i1 ,y i1 ),(c i2 ,y i2 ),...,(c it ,y it )} (Ⅲ)
in the formula (III), the compound represented by the formula (III),
the method comprises the steps that new time sequence data are obtained by introducing ith group of time sequence data into GSHP system working parameters in a training set and expanding the time sequence data;
H. adopting a Between-year seasonal month-by-month method for different month data, and dividing the whole data structure X into 12 blocks by taking a month as a unit, wherein the whole data structure X is as follows:
in the formula (IV), the compound is shown in the specification,
the new time sequence data expanded in the step G is referred to; i is more than 1 and less than the number of training samples;
after dividing the overall data structure X into 12 blocks, X = [ B = 1 B 2 ... B 12 ]；
Setting the reconstructed data of the mth month as B m M is more than or equal to 1 and less than or equal to 12, and all new time sequence data x in the training set i new In the m-th band, the observation value
As initial data, at a sampling rate of 12, for x i new Sampling for p times to obtain observation values with parameters of all m months
Forming a seasonal monthly time series
Finally obtaining a two-dimensional matrix B m The method comprises the following steps:
in (V), B m ∈R i*p I denotes the number of training set samples, p denotes the number of samples, 1<p<All years of observation, R refers to the real number field; b is m Any one of the row vectors
From the ith set of new time series performance data x i new Observation value with reference of all m months in
Structure (m) of 1 ～m p Finger pair x i new M months of (c) are sampled 1-p times at intervals of 12; b is m The expression comprises 1 to i groups of time sequence performance data x in the training set i new The belt parameter observed values of all m months;
(3) Based on a gradient lifting decision tree algorithm, a decision tree set model which has external system working parameter input and can be used for time series regression prediction, namely a test module, is generated in a mode of constructing an objective function through multiple times of optimization iterative training, and the performance of the GSHP system is predicted;
(4) Performing data reconstruction on the test set data by adopting the same method as the step (2), and performing performance prediction on the decision tree set model obtained by training in the step (3) by adopting an autoregressive mode;
(5) And comparing the prediction data obtained by the test set through the decision tree set model with the real data to measure the prediction effect of the decision tree set model.
2. The method for predicting the performance of the ground source heat pump system according to claim 1, wherein the step (1) of preprocessing and analyzing the heat transfer performance data of the GSHP system comprises the following steps:
A. simulating and acquiring heat transfer performance data of the GSHP system; heat transfer performance data for GSHP systems include ifDry group of time sequence data, each group of time sequence data comprises 12 system working parameters X 1 ～X 12 And Y 1 ～Y 4 Time series of (2), Y 1 The unit is kWh which is the power consumption of the heat pump unit; y is 2 The total electricity consumption of the GSHP system is expressed in kWh; y is 3 The heat pump unit heating capacity or cooling capacity is compared with the operating power of the heat pump unit; y is 4 The ratio of the heating capacity or the refrigerating capacity of the heat pump unit to the total power of the system is referred to;
B. detecting NAN values and non-numerical value form values of the heat transfer performance data obtained by simulation, and deleting the rows where the NAN values and the non-numerical value form values are located;
C. carrying out visual operation on the heat transfer performance data of the GSHP system obtained in the step B, checking the overall trend of the GSHP system, judging abnormal numerical values according to professional theoretical knowledge, and deleting the lines where the abnormal numerical values are located;
D. c, performing seasonal decomposition on the time series data obtained after the processing of the step C, decomposing the time series data into a seasonal item S, a trend item T and a random error item I, and regarding a time series data Y, a seasonal decomposition model is as follows:
Y t ＝S t +T t +I t (Ⅰ)
in formula (I), t is a specific time, Y t The observed value corresponding to the moment t, S t Refers to a seasonal term reflecting the cyclic variation of a time series within one year, T t Long-term variation of time series captured by the trend term, I t The random error item captures changes which cannot be explained by trend or seasonal effect, and the formula (I) shows that an observed value corresponding to the time sequence t moment is the sum of the trend value, the seasonal effect and the random influence at the moment;
E. and D, analyzing the seasonal decomposition result obtained in the step D, judging the stationarity of the seasonal decomposition result, if the seasonal decomposition result is not stationarily, stabilizing the time sequence data in a differential mode, and if the seasonal decomposition result is stationarily, entering the step (2).
3. The method for predicting the performance of the ground source heat pump system according to claim 1, wherein in the step (3), a prediction model is established, and the method comprises the following steps:
I. for different months, based on a gradient lifting decision tree algorithm, establishing an objective function containing a geothermal square loss term and a system normalization regularization term, wherein the specific form is as follows:
in the formula (VI), theta represents parameters in the prediction model and refers to the tree structure and the tree leaf node score of each decision tree; n represents the number of samples; y is i Representing the true value of the ith sample;
representing a predicted value of the ith sample; t represents the number of the decision trees; f. of t Representing a specific decision tree; omega (f) t ) The regularization item for expressing the normalization of the geothermal system comprises a penalty item for the node number of each decision tree and an L2 modular square of the leaf node fraction, and is used for restraining the complexity of the decision tree; the performance of a model obtained by training under the condition that the model is over-fitted with a certain set of specific working parameters of the GSHP system is avoided, the generalization capability of a prediction model obtained by training under different system working parameters is standardized, and the mathematical expression of the generalization capability is as follows:
in the formula (VII), K, w j Respectively representing the leaf node number of each decision tree and the score on each leaf node; lambda and gamma respectively represent punishment coefficients of the leaf node number and the leaf node fraction;
J. minimizing an objective function in an incremental training mode by adopting a forward distribution algorithm;
generating a decision tree set model with external system working parameter input through multiple optimization iterative training based on the data reconstructed in the step H; initializing a decision tree set model, then performing multiple rounds of iterative optimization, wherein in each round of iteration, an optimal decision tree is added to an original decision tree set model to continuously reduce residual errors of the original decision tree set model, and the decision tree set model is finally obtained through multiple iterative optimization training, wherein the specific training process is as follows:
...
in formula (VIII), x is different for different months i For reconstructing the corresponding blocks B in the overall data structure X m Ith row vector of
Through r rounds of incremental training, a decision tree set model with external system working parameter input and formed by r decision trees is obtained, and the mathematical expression of the final decision tree set model is as follows:
K. and (4) respectively establishing independent decision tree set models for different months by adopting the same method of the step I, J.
4. A performance prediction method for a ground source heat pump system according to any one of claims 1-3, characterized in that in the step (4), the performance prediction is performed in an autoregressive manner, and comprises:
l, performing data reconstruction on the test set data by adopting the same method as the step (2), and establishing a data structure with system working parameter input;
m, on the basis of the step K, for different months, performing regression prediction on the performance time series data of the GSHP system by adopting a nonlinear autoregressive mode to realize long-term and continuous prediction on the performance of the GSHP system, wherein the nonlinear autoregressive mode is as follows:
in the formula (X),
denotes the predicted value, x, of the mth month of the yth year (y-M.m) Represents 12 system parameters, y, corresponding to the mth month of the y-M year (y-M,m) The true value of the mth month in the y-M year is represented, M represents the number of lagged terms, and f (-) is an arbitrary model technology;
and D, respectively predicting different months by using the decision tree set model obtained in the step J, integrating and reconstructing the performance data obtained by the first prediction to be used as the next model input, performing the second prediction, and completing the regression prediction of the performance time series data of the GSHP system through the sliding prediction in a nonlinear autoregressive mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043376.3A CN109816166B (en) | 2019-01-17 | 2019-01-17 | Ground source heat pump system performance prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910043376.3A CN109816166B (en) | 2019-01-17 | 2019-01-17 | Ground source heat pump system performance prediction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109816166A CN109816166A (en) | 2019-05-28 |
| CN109816166B true CN109816166B (en) | 2022-11-29 |
Family
ID=66604520
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910043376.3A Active CN109816166B (en) | 2019-01-17 | 2019-01-17 | Ground source heat pump system performance prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109816166B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457296B (en) * | 2019-07-05 | 2021-12-03 | 中国船舶重工集团公司第七一九研究所 | Cleaning method for running data of constant speed pump |
| CN110570322B (en) * | 2019-09-17 | 2023-09-12 | 西北农林科技大学 | Agricultural meteorological index insurance rate calibrating method based on time sequence simulation |
| CN113326557A (en) * | 2021-06-16 | 2021-08-31 | 西北农林科技大学 | Ground source heat pump long-term performance prediction method |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678322A (en) * | 2012-09-03 | 2014-03-26 | 阿里巴巴集团控股有限公司 | System and method for sample data integration |
| CN103912966A (en) * | 2014-03-31 | 2014-07-09 | 武汉科技大学 | Optimal control method for ground source heat pump refrigerating system |
| CN105488612A (en) * | 2015-11-24 | 2016-04-13 | 武汉科技大学 | Monte-Carlo method based reliability analysis method for buried vertical pipe of ground source heat pump |
| CN105937823A (en) * | 2016-03-31 | 2016-09-14 | 中国农业大学 | Control method and system for ground source heat pump |
| CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
| CN107122569A (en) * | 2017-05-25 | 2017-09-01 | 山东建筑大学 | Differential parallel HCMAC neutral net bilateral variable-flow earth-source hot-pump system power consumption Forecasting Methodologies |
| 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 |
| CN108304966A (en) * | 2018-01-18 | 2018-07-20 | 沈阳工程学院 | A kind of short-term monthly power demand prediction technique |
| CN109086560A (en) * | 2018-09-30 | 2018-12-25 | 安徽工业大学 | The vertical single U-shaped underground pipe fluid temperature (F.T.) distribution forecasting method of earth source heat pump under variable working condition |
-
2019
- 2019-01-17 CN CN201910043376.3A patent/CN109816166B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678322A (en) * | 2012-09-03 | 2014-03-26 | 阿里巴巴集团控股有限公司 | System and method for sample data integration |
| CN103912966A (en) * | 2014-03-31 | 2014-07-09 | 武汉科技大学 | Optimal control method for ground source heat pump refrigerating system |
| CN105488612A (en) * | 2015-11-24 | 2016-04-13 | 武汉科技大学 | Monte-Carlo method based reliability analysis method for buried vertical pipe of ground source heat pump |
| CN105937823A (en) * | 2016-03-31 | 2016-09-14 | 中国农业大学 | Control method and system for ground source heat pump |
| CN106971240A (en) * | 2017-03-16 | 2017-07-21 | 河海大学 | The short-term load forecasting method that a kind of variables choice is returned with Gaussian process |
| CN107122569A (en) * | 2017-05-25 | 2017-09-01 | 山东建筑大学 | Differential parallel HCMAC neutral net bilateral variable-flow earth-source hot-pump system power consumption Forecasting Methodologies |
| 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 |
| CN108304966A (en) * | 2018-01-18 | 2018-07-20 | 沈阳工程学院 | A kind of short-term monthly power demand prediction technique |
| CN109086560A (en) * | 2018-09-30 | 2018-12-25 | 安徽工业大学 | The vertical single U-shaped underground pipe fluid temperature (F.T.) distribution forecasting method of earth source heat pump under variable working condition |
Non-Patent Citations (3)
| Title |
|---|
| Accurately predicting heat transfer performance of ground heat exchanger for ground-coupled heat pump systems using data mining methods;Zhaoyi Zhuang;《Springerlink》;20171231;全文 * |
| 公共建筑用土壤源热泵机组运行性能测试分析及其模拟预测;张信树;《CNKI》;20080115;全文 * |
| 喷气增焓空气源热泵热性能评价及预测;张东;《化工学报》;20141231;第65卷(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109816166A (en) | 2019-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xue et al. | Multi-step ahead forecasting of heat load in district heating systems using machine learning algorithms | |
| CN109816166B (en) | Ground source heat pump system performance prediction method | |
| Parhizkar et al. | Evaluation and improvement of energy consumption prediction models using principal component analysis based feature reduction | |
| Wang et al. | China’s natural gas production and consumption analysis based on the multicycle Hubbert model and rolling Grey model | |
| CN113962364A (en) | Multi-factor power load prediction method based on deep learning | |
| CN107609667B (en) | Heat supply load prediction method and system based on Box _ cox transformation and UFCNN | |
| CN111639823B (en) | Building cold and heat load prediction method constructed based on feature set | |
| CN110544180A (en) | Building energy-saving system based on energy consumption prediction and analysis diagnosis | |
| CN115409396A (en) | Comprehensive energy system multi-time scale scheduling method based on double-layer rolling optimization | |
| CN112906987B (en) | Photovoltaic power prediction method based on convolutional neural network and two-dimensional meteorological matrix | |
| CN112100911B (en) | Solar radiation prediction method based on depth BILSTM | |
| CN111401757A (en) | Multi-region power load time sequence scene modeling method considering time-space correlation | |
| CN105139080A (en) | Improved photovoltaic power sequence prediction method based on Markov chain | |
| CN108197425A (en) | A kind of intelligent grid data resolving method based on Non-negative Matrix Factorization | |
| Alanbar et al. | Energy Consumption Prediction Using Deep Learning Technique | |
| Zhao et al. | Short-term microgrid load probability density forecasting method based on k-means-deep learning quantile regression | |
| CN114239292A (en) | Comprehensive evaluation method and system for low-carbon economic operation-oriented multifunctional demand park | |
| Vasanthkumar et al. | Improving energy consumption prediction for residential buildings using Modified Wild Horse Optimization with Deep Learning model | |
| Wang et al. | Deep learning based closed-loop well control optimization of geothermal reservoir with uncertain permeability | |
| Liu et al. | Short-term forecast of multi-load of electrical heating and cooling in regional integrated energy system based on deep LSTM RNN | |
| CN114091776A (en) | K-means-based multi-branch AGCNN short-term power load prediction method | |
| CN110033126A (en) | Shot and long term memory network prediction technique based on attention mechanism and logistic regression | |
| Zhang et al. | A data transfer method based on one dimensional convolutional neural network for cross-building load prediction | |
| CN109767044B (en) | Method for predicting heat energy conversion efficiency of ground source heat pump system | |
| CN117113054A (en) | Multi-element time sequence prediction method based on graph neural network and transducer |
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 |