A method and device for training and predicting energy consumption of a water chiller

Abstract

The application discloses a kind of energy consumption prediction model training and energy consumption prediction method, device of water chilling unit.For obtaining the historical hourly operation data of water chilling unit and the first water chilling unit energy consumption prediction model of preset, and the historical hourly operation data is preprocessed, test data set and training data set are obtained;Test data is randomly extracted from the test data set in turn, the profile of the training data set is extracted, and the training data set is updated according to the test data and the profile;According to the updated training data set, by multiple linear regression fitting, the first water chilling unit energy consumption prediction model is iteratively fitted, until the accuracy of the second water chilling unit energy consumption prediction model of current iteration fitting reaches the preset threshold, and the second water chilling unit energy consumption prediction model is output;If the model accuracy does not reach the preset threshold after iteration, adjust the training parameters of fitting process.The application can improve the precision of model and the accuracy of energy consumption prediction.

Description

Technical Field

[0001] This invention relates to the field of energy-saving control technology for central air conditioning, and in particular to a method and apparatus for training an energy consumption prediction model and predicting energy consumption for a chiller unit. Background Technology

[0002] Chillers account for more than half of the total energy consumption of air conditioning systems. Chiller energy consumption analysis is fundamental to energy-saving control in air conditioning systems, and energy conservation and consumption reduction typically revolve around chiller energy consumption analysis. Chiller operating parameters are highly dimensional and tightly coupled. Parameters such as chiller power, cooling water inlet temperature, cooling water outlet temperature, chilled water inlet temperature, chilled water outlet temperature, evaporation temperature, condensation temperature, and output cooling capacity are all interconnected and influence each other. Current chiller energy consumption models are built by collecting operating parameters and then predicting energy consumption based on this model. However, the actual model is affected by the collected data and cannot accurately predict energy consumption across all ranges, nor can it guarantee the comprehensiveness of the collected data. Therefore, the description of the range of chiller energy consumption variations is unclear, leading to decreased model accuracy and affecting the accuracy of chiller operating energy consumption prediction. Summary of the Invention

[0003] This invention provides a training method and apparatus for energy consumption prediction model of chiller units, as well as an energy consumption prediction method and apparatus, to solve the technical problem of low accuracy of energy consumption prediction models.

[0004] To address the aforementioned technical problems, in a first aspect, embodiments of the present invention provide a method for training a chiller unit energy consumption prediction model, comprising:

[0005] Acquire historical hourly operating data of the chiller unit and a preset first chiller unit energy consumption prediction model, and preprocess the historical hourly operating data to obtain a test dataset and a training dataset.

[0006] Test data is randomly extracted from the test dataset in sequence, the contour surface of the training dataset is extracted, and the training dataset is updated according to the test data and the contour surface.

[0007] Based on the updated training dataset, the energy consumption prediction model of the first chiller unit is iteratively fitted using multiple linear regression until the accuracy of the energy consumption prediction model of the second chiller unit fitted in the current iteration reaches a preset threshold, and the energy consumption prediction model of the second chiller unit is output; wherein, after each iteration, if the accuracy does not reach the preset threshold, the training parameters of the multiple linear regression fitting process are adjusted.

[0008] Compared with existing technologies, the above embodiments have the following beneficial effects: by extracting contour surfaces and continuously updating the training set, firstly, the generalization ability of subsequent models is improved, enabling them to perform well in different operating scenarios; secondly, the contour surfaces can clearly reflect the applicable range of the model, avoiding erroneous predictions caused by input data exceeding the predictive capability range of the model; and finally, the accuracy and robustness of the prediction model are improved through multiple fitting methods.

[0009] In one embodiment of the first aspect, the step of extracting the contour surface includes:

[0010] The first data point in the training dataset is sequentially traversed, and all second data points in the training dataset whose distance from the first data point is less than a first threshold are obtained to generate a first point set corresponding to each training data point.

[0011] Iterate through each second data point in the first point set corresponding to each first data point, and calculate the coordinates of the first center and the second center of each second data point.

[0012] Based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set, determine whether the corresponding first data point is a contour point;

[0013] The contour surface is generated based on all the contour points.

[0014] Compared with existing technologies, the above embodiments have the following beneficial effects: by determining whether the points in the point set are edge contour points by the geometric relationship between data points, the accuracy of the final determination of the usable range of the model can be improved; at the same time, the range determined by the contour surface can also help to further supplement the data in the training dataset and improve the accuracy of the model.

[0015] In one embodiment of the first aspect, the method for calculating the coordinates of the first center and the coordinates of the second center includes:

[0016] The formula for calculating the coordinates of the first circle center specifically includes:

[0017]

[0018]

[0019] Where x o1 The x-axis coordinate of the first circle's center; y o1 x is the ordinate of the first center point; x is the abscissa of the first data point; y is the ordinate of the first data point; x1 is the abscissa of the second data point; y1 is the ordinate of the second data point; α is half of the first threshold.

[0020] The formula for calculating the coordinates of the second center of the circle specifically includes:

[0021]

[0022]

[0023] Where x o2 The x-axis coordinate of the second circle center; y o2 x is the ordinate of the second center point; x is the abscissa of the first data point; y is the ordinate of the first data point; x1 is the abscissa of the second data point; y1 is the ordinate of the second data point; α is half of the first threshold.

[0024] Compared with the prior art, the above embodiments have the following beneficial effects: by calculating the center point, the relative positional relationship between other data points and the first data point can be reflected according to the geometric relationship, thereby improving the accuracy of contour point judgment.

[0025] In one embodiment of the first aspect, determining whether a corresponding first data point is a contour point based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set includes:

[0026] Calculate sequentially the first distance from the coordinates of the first circle center and the second distance from the coordinates of the second circle center for each data point in the first point set, excluding the second data point.

[0027] When there exists a second data point in the first set of points corresponding to the first data point that satisfies the condition that the first distance or second distance corresponding to other data points, excluding the second data point, is greater than the second threshold, the first data point is determined to be a contour point.

[0028] Compared with the prior art, the above embodiments have the following beneficial effects: Based on the relative relationship between other data points calculated through the coordinates of the center of the circle and the first data point, if the first data point is an edge contour point, then other points in the point set will definitely not have two points appearing on both sides of the first data point at the same time. Based on the above principle, the coordinates of the two center points can be easily calculated, and the contour point can be accurately determined based on the distance between the two center coordinates and other distances, thereby improving the accuracy of generating the contour surface.

[0029] In one embodiment of the first aspect, updating the training dataset based on the test data and the contour surface includes: adding the test data to the training dataset and removing the test data from the test dataset when the test data is not included in the contour surface.

[0030] Compared with the prior art, the above embodiments have the following beneficial effects: by using the contour surface, feature data that is not included in the training dataset is identified and then supplemented into the training dataset, thereby improving the comprehensiveness of the training dataset.

[0031] In one embodiment of the first aspect, adjusting the training parameters of the multiple linear regression fitting process includes: adjusting the minimum error parameter of the multiple linear regression fitting and adjusting the sensitive data proportion of the historical hourly running data.

[0032] Compared with the prior art, the above embodiments have the following beneficial effects: by adjusting the minimum error parameter and the proportion of sensitive data, the fitting effect of the final model can be improved.

[0033] In one embodiment of the first aspect, the preprocessing of the historical hourly running data to obtain a test dataset and a training dataset includes: cleaning the historical hourly running data, dividing the cleaned historical hourly running data according to a preset ratio, and obtaining the test dataset and the training dataset.

[0034] Compared with existing technologies, the above embodiments have the following beneficial effects: by cleaning the data, abnormal data such as garbled characters, negative numbers, and maximum values ​​in the dataset are removed, thereby improving the quality of the data.

[0035] Secondly, embodiments of the present invention also provide a method for predicting the energy consumption of a chiller unit, including:

[0036] A chiller unit energy consumption prediction model and a contour surface extracted from a training dataset are obtained; wherein the chiller unit energy consumption prediction model is trained according to the chiller unit energy consumption prediction model training method described in the above embodiments; and the contour surface is extracted according to the chiller unit energy consumption prediction model training method described in the above embodiments.

[0037] The current operating data of the chiller unit is obtained. When the operating data is contained in the contour surface, the operating data is input into the energy consumption prediction model to obtain the energy consumption prediction result.

[0038] Compared with the prior art, the above embodiments have the following beneficial effects: the contour surface can accurately identify the applicable range of the model, avoiding the use of the model to obtain incorrect results due to the input data not being within the range; at the same time, based on the contour surface, the robustness of the existing model can also be judged, and when the robustness is insufficient, the prediction model can be updated in time to ensure the accuracy of energy consumption prediction.

[0039] Thirdly, embodiments of the present invention provide a training device for a chiller unit energy consumption prediction model, comprising: a data processing module, a training dataset update module, and a prediction model fitting module;

[0040] The data processing module is used to acquire historical hourly operating data of the chiller unit and a preset first chiller unit energy consumption prediction model, and preprocess the historical hourly operating data to acquire a test dataset and a training dataset.

[0041] The training dataset update module is used to randomly extract test data from the test dataset, extract the contour surface of the training dataset, and update the training dataset according to the test data and the contour surface.

[0042] The prediction model fitting module is used to iteratively fit the first chiller unit energy consumption prediction model according to the updated training dataset through multiple linear regression until the accuracy of the second chiller unit energy consumption prediction model fitted in the current iteration reaches a preset threshold, and then output the second chiller unit energy consumption prediction model; wherein, after each iteration, if the accuracy does not reach the preset threshold, the training parameters of the multiple linear regression fitting process are adjusted.

[0043] In one embodiment of the third aspect, the training dataset update module includes: a first point set acquisition unit, a circle center coordinate calculation unit, a contour point judgment unit, and a contour surface generation unit;

[0044] The first point set acquisition unit is used to sequentially traverse the first data point in the training dataset, acquire all second data points in the training dataset whose distance from the first data point is less than a first threshold, and generate a first point set corresponding to each training data.

[0045] The circle center coordinate calculation unit is used to sequentially traverse each second data point in the first point set corresponding to each first data point, and calculate the first circle center coordinate and the second circle center coordinate corresponding to each second data point;

[0046] The contour point determination unit is used to determine whether the corresponding first data point is a contour point based on the first center coordinates and the second center coordinates of each second data point in the first point set.

[0047] The contour surface generation unit is used to generate the contour surface based on all the contour points.

[0048] In one embodiment of the third aspect, the method for calculating the coordinates of the first center and the coordinates of the second center includes:

[0049] The formula for calculating the coordinates of the first circle center specifically includes:

[0050]

[0051]

[0052] Where, x o1 The x-axis coordinate of the first circle's center; y o1 x is the ordinate of the first center point; x is the abscissa of the first data point; y is the ordinate of the first data point; x1 is the abscissa of the second data point; y1 is the ordinate of the second data point; α is half of the first threshold.

[0053] The formula for calculating the coordinates of the second center of the circle specifically includes:

[0054]

[0055]

[0056] Where, x o2 The x-axis coordinate of the second circle center; y o2 x is the ordinate of the second center point; x is the abscissa of the first data point; y is the ordinate of the first data point; x1 is the abscissa of the second data point; y1 is the ordinate of the second data point; α is half of the first threshold.

[0057] In one embodiment of the third aspect, the training dataset update module is further configured to update the training dataset according to the test data and the contour surface, including: adding the test data to the training dataset and removing the test data from the test dataset when the test data is not included in the contour surface.

[0058] In one embodiment of the third aspect, the contour point determination unit is configured to determine whether a corresponding first data point is a contour point based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set, including:

[0059] Calculate sequentially the first distance from the coordinates of the first circle center and the second distance from the coordinates of the second circle center for each data point in the first point set, excluding the second data point.

[0060] When there exists a second data point in the first set of points corresponding to the first data point that satisfies the condition that the first distance or second distance corresponding to other data points, excluding the second data point, is greater than the second threshold, the first data point is determined to be a contour point.

[0061] In one embodiment of the third aspect, the prediction model fitting module is used to adjust the training parameters of the multiple linear regression fitting process, including: adjusting the minimum error parameter of the multiple linear regression fitting and adjusting the sensitive data ratio of the historical hourly running data.

[0062] In one embodiment of the third aspect, the data processing module is used to preprocess the historical hourly running data to obtain a test dataset and a training dataset, including: cleaning the historical hourly running data, dividing the cleaned historical hourly running data according to a preset ratio, and obtaining the test dataset and the training dataset.

[0063] Fourthly, embodiments of the present invention also provide a chiller unit energy consumption prediction device, including: a model acquisition module and an energy consumption calculation module;

[0064] The model acquisition module is used to acquire a chiller unit energy consumption prediction model and a contour surface extracted from a training dataset; wherein the chiller unit energy consumption prediction model is trained using the chiller unit energy consumption prediction model training device described in the above embodiment; and the contour surface is acquired using the chiller unit energy consumption prediction model training device described in the above embodiment.

[0065] The energy consumption calculation module is used to obtain the current operating data of the chiller unit. When the operating data is included in the contour surface, the operating data is input into the energy consumption prediction model to obtain the energy consumption prediction result. Attached Figure Description

[0066] Figure 1 A flowchart illustrating a method for training a chiller unit energy consumption prediction model, provided in an embodiment of the present invention;

[0067] Figure 2 A flowchart illustrating another method for training a chiller unit energy consumption prediction model provided in an embodiment of the present invention;

[0068] Figure 3 A schematic flowchart of a chiller unit energy consumption prediction method provided in an embodiment of the present invention;

[0069] Figure 4 This is a schematic diagram of the training results of a chiller unit energy consumption prediction model provided in an embodiment of the present invention;

[0070] Figure 5 This is a schematic diagram of the structure of a training device for a chiller unit energy consumption prediction model provided in an embodiment of the present invention;

[0071] Figure 6 This is a schematic diagram of a chiller unit energy consumption prediction device provided in an embodiment of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Example 1

[0074] Please refer to Figure 1 The present invention provides a method for training a chiller unit energy consumption prediction model, comprising steps S101 to S103, specifically:

[0075] S101: Obtain historical hourly operating data of the chiller unit and a preset first chiller unit energy consumption prediction model, and preprocess the historical hourly operating data to obtain a test dataset and a training dataset.

[0076] Furthermore, the energy consumption prediction model for the first chiller unit can be constructed by combining the chilled water outlet temperature, cooling water inlet temperature, instantaneous output cooling capacity, and instantaneous power with the ASHRAE chiller unit energy consumption empirical model, specifically as follows:

[0077] P chiller =intercept+COf1T chw，out +COf2T cdw，in +COf3Q+COf4T chw，out 2

[0078] +COf5T chw，out T cdw，in +COf6T chw，out Q+COf7T cdw，in 2 +COf8T cdw，in Q+COf9Q 2

[0079] Where COf1 to COf9 are the fitting coefficients; T chw，out T represents the outlet temperature of the chilled water. cdw，in P is the inlet temperature of the chilled water; chiller is the energy consumption of the chiller unit; intercept is a constant term; Q is the instantaneous output cooling capacity.

[0080] Furthermore, the preprocessing of the historical hourly running data to obtain the test dataset and the training dataset includes: cleaning the historical hourly running data, dividing the cleaned historical hourly running data according to a preset ratio, and obtaining the test dataset and the training dataset.

[0081] Preferably, data cleaning can be implemented by removing abnormal data such as garbled characters, negative numbers, and maximum values ​​from the dataset.

[0082] The above operations can remove noise from the dataset, ensure the accuracy of model training, and improve model precision.

[0083] S102: Randomly sample test data from the test dataset, extract the contour surface of the training dataset, and update the training dataset based on the test data and the contour surface.

[0084] Furthermore, the extraction of the contour surface from the training dataset specifically includes: sequentially traversing the first data points in the training dataset, obtaining all second data points in the training dataset whose distance from the first data point is less than a first threshold, and generating a first point set corresponding to each training data point; sequentially traversing each second data point in the first point set corresponding to each first data point, calculating the first center coordinates and the second center coordinates corresponding to each second data point; sequentially determining whether the corresponding first data point is a contour point based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set; and generating the contour surface based on all the contour points.

[0085] Furthermore, the method for calculating the coordinates of the first center and the second center includes:

[0086] The formula for calculating the coordinates of the first circle center specifically includes:

[0087]

[0088]

[0089] The formula for calculating the coordinates of the second center of the circle specifically includes:

[0090]

[0091]

[0092] Where, x o1 The x-axis coordinate of the first circle's center; y o1 x is the ordinate of the first circle's center; x is the abscissa of the first data point. o2 The x-axis coordinate of the second circle center; y o2 x is the ordinate of the second center point; x is the abscissa of the first data point; y is the ordinate of the first data point; x1 is the abscissa of the second data point; y1 is the ordinate of the second data point; α is half of the first threshold.

[0093] Furthermore, determining whether the corresponding first data point is a contour point based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set includes: sequentially calculating the first distance and the second distance from the first center coordinates to the other data points in the first point set, excluding the second data point; when there is a second data point in the first point set corresponding to the first data point that satisfies that the first distance or the second distance corresponding to the other data points, excluding the second data point, is greater than the second threshold, the first data point is determined to be a contour point.

[0094] Preferably, the above-described embodiments for obtaining the contour surface can also be implemented through the following steps:

[0095] Step 1: Randomly select any data point in the training set, denoted as point P(x,y). Based on the pre-set rolling circle radius α, search for all points in the projected point cloud that are within 2α of point P, and denoted as point set A.

[0096] Step 2: Randomly select any data point in point set A, denoted as point P1(x1,y1). Based on P(x,y), P1(x1,y1), and the radius α of the rolling circle, calculate the coordinates of the two center points o1(x1,y1). o1 ,y o1 ) and o2(x o2 ,y o2 The center coordinates can be obtained using the center coordinate calculation formula in the above embodiment;

[0097] Step 3: For all points in point set A except P1(x1,y1), calculate the distance to the center o1(x1,y1) of each point. o1 ,y o1 ) and o2(x o2 ,y o2 The distance from o1 to o2 is calculated. If the distance from all points to o1 or o2 is greater than α, then mark point P as a contour point and terminate the judgment for that point.

[0098] Step 4: If the distances from all points other than P1 to o1 and o2 are not all greater than α, then iterate through all points in point set A and use them as P1 in turn, recalculate the center coordinates, and judge point P according to Step 3. If there exists a point that marks point P as a contour point, then determine that point P is a contour point; otherwise, determine that point P is not a contour point.

[0099] Step 5: Based on all the obtained contour points, fit the contour edge lines to obtain the contour surface.

[0100] Using the above method, the relative relationship between other data points and the first data point can be calculated based on the coordinates of the center of the circle. If the first data point is an edge contour point, then other points in the point set will not have two points appearing on both sides of the first data point at the same time. Based on the above principle, the coordinates of the two center points can be easily calculated, and the contour points can be accurately determined based on the distance between the coordinates of the two center points and other distances, thus improving the accuracy of generating the contour surface.

[0101] Furthermore, updating the training dataset based on the test data and the contour surface includes: adding the test data to the training dataset and removing the test data from the test dataset when the test data is not included in the contour surface.

[0102] By combining the aforementioned contour surface extraction method, the training dataset is updated to identify feature data that was not included in the training dataset and then supplement it into the training dataset, thereby improving the comprehensiveness of the training dataset and thus improving the accuracy of the final model prediction.

[0103] S103: Based on the updated training dataset, the first chiller unit energy consumption prediction model is iteratively fitted using multiple linear regression until the accuracy of the second chiller unit energy consumption prediction model fitted in the current iteration reaches a preset threshold, and the second chiller unit energy consumption prediction model is output; wherein, after each iteration, if the accuracy does not reach the preset threshold, the training parameters of the multiple linear regression fitting process are adjusted.

[0104] Preferably, the accuracy of the energy consumption prediction model for the second chiller unit can be improved by R... 2 An assessment will be conducted, specifically as follows:

[0105]

[0106] in, This indicates an error arising from the use of energy consumption models for prediction. Indicates use Errors in prediction, R 2 The closer the value is to 1, the higher the accuracy of the model; 0 indicates the worst accuracy of the model.

[0107] Furthermore, when R 2 If the error is greater than 0.9, the fitting is complete; otherwise, adjust the minimum fitting error or remove sensitive data and refit.

[0108] By evaluating the accuracy of the model and iteratively fitting the model, the stability of the model was improved, and the accuracy of the model was also increased.

[0109] refer to Figure 2The following is a flowchart illustrating another method for training a chiller unit energy consumption prediction model according to an embodiment of the present invention:

[0110] Step 1: Obtain historical hourly operating data of the air conditioning system chiller unit, clean and classify the data, and obtain the test dataset and training dataset;

[0111] Step 2: Fit the chiller unit energy consumption prediction model based on the training dataset to obtain the chiller unit energy consumption model;

[0112] Step 3: Calculate Ri for the chiller unit's energy consumption model. 2 If R 2 If the value is ≤0.9, proceed to step 4; otherwise, proceed to step 2.

[0113] Step 4: Extract the contour surface of the training dataset and randomly extract data points from the test dataset. Determine whether the data point is inside the contour surface. If the data point is inside the contour surface, output the chiller unit energy consumption model. Otherwise, add the data point to the training dataset and jump to step 2.

[0114] In summary, it can be seen that by extracting contour surfaces and continuously updating the training set, this application firstly improves the generalization ability of subsequent models, enabling them to perform well in different operating scenarios; secondly, the contour surfaces clearly reflect the applicable range of the model, avoiding erroneous predictions caused by input data exceeding the model's predictive capabilities; and finally, by using multiple fitting methods, it improves the accuracy and robustness of the prediction model.

[0115] Example 2

[0116] Please refer to Figure 3 The present invention provides a method for predicting the energy consumption of a chiller unit, comprising steps S101 to S103, specifically:

[0117] S201: Obtain the chiller unit energy consumption prediction model and the contour surface extracted from the training dataset; wherein, the chiller unit energy consumption prediction model is trained according to the chiller unit energy consumption prediction model training method described in any one of Embodiment 1; the contour surface is extracted according to the contour surface extraction method in the chiller unit energy consumption prediction model training method described in Embodiment 1.

[0118] S202: Obtain the current operating data of the chiller unit. When the operating data is included in the contour surface, input the operating data into the energy consumption prediction model to obtain the energy consumption prediction result.

[0119] For details, please refer to Figure 4This is a schematic diagram of a contour surface extracted from a training dataset, provided by an embodiment of the present invention. The maximum and minimum values ​​of each dimension can be determined from the contour surface, where the chilled water outlet temperature T... chw，out The temperature range is 4.3℃~20.0℃, and the cooling water inlet temperature T cdw，in The temperature range is 25.2℃~33.4℃, and the output cooling capacity Q ranges from 0 to 1765.8kW.

[0120] Furthermore, Figure 4 The corresponding energy consumption prediction model for chiller units is:

[0121] P chiller =-320.299821-9.320944T chw，out +28.381406T cdw，in -0.331231Q

[0122] +0.353415T chw，out 2 -0.019052T chw，out T cdw，in -0.000012Q 2 +0.003694

[0123] T chw，out Q-0.517179T cdw，in 2 +0.015047T cdw，in Q

[0124] The contour surface can accurately identify the applicable range of the model, avoiding incorrect results obtained by using the model due to input data not being within the range; at the same time, the robustness of the existing model can also be judged based on the contour surface, and when the robustness is insufficient, the prediction model can be updated in time to ensure the accuracy of energy consumption prediction.

[0125] Example 3

[0126] Please refer to Figure 5 The present invention provides a training device for a chiller unit energy consumption prediction model, comprising: a data processing module 501, a training dataset update module 502, and a prediction model fitting module 503; wherein, the training dataset update module 502 comprises: a first point set acquisition unit 5021, a circle center coordinate calculation unit 5022, a contour point judgment unit 5023, and a contour surface generation unit 5024.

[0127] Specifically, the data processing module 501 is used to acquire historical hourly operating data of the chiller unit and a preset first chiller unit energy consumption prediction model, and preprocess the historical hourly operating data to acquire a test dataset and a training dataset; the training dataset update module 502 is used to randomly extract test data from the test dataset, extract the contour surface of the training dataset, and update the training dataset according to the test data and the contour surface; the prediction model fitting module 503 is used to iteratively fit the first chiller unit energy consumption prediction model according to the updated training dataset through multiple linear regression fitting until the accuracy of the second chiller unit energy consumption prediction model fitted in the current iteration reaches a preset threshold, and output the second chiller unit energy consumption prediction model; wherein, after each iteration, if the accuracy does not reach the preset threshold, the training parameters of the multiple linear regression fitting process are adjusted.

[0128] Furthermore, the first point set acquisition unit 5021 is used to sequentially traverse the first data points in the training dataset, acquire all second data points in the training dataset whose distance from the first data point is less than a first threshold, and generate a first point set corresponding to each training data point; the circle center coordinate calculation unit 5022 is used to sequentially traverse each second data point in the first point set corresponding to each first data point, and calculate the first circle center coordinate and the second circle center coordinate corresponding to each second data point; the contour point judgment unit 5023 is used to sequentially determine whether the corresponding first data point is a contour point based on the first circle center coordinate and the second circle center coordinate corresponding to each second data point in the first point set; the contour surface generation unit 5024 is used to generate the contour surface based on all the contour points.

[0129] Furthermore, the method for calculating the center coordinates of the circle 5022 includes:

[0130] The formula for calculating the coordinates of the first circle center specifically includes:

[0131]

[0132]

[0133] The formula for calculating the coordinates of the second center of the circle specifically includes:

[0134]

[0135]

[0136] Where, x o1 The x-axis coordinate of the first circle's center; y o1x is the ordinate value of the first circle's center. o2 The x-axis coordinate of the second circle center; y o2 x is the ordinate of the second center point; x is the abscissa of the first data point; y is the ordinate of the first data point; x1 is the abscissa of the second data point; y1 is the ordinate of the second data point; α is half of the first threshold.

[0137] Furthermore, the contour point determination unit 5023 is used to determine whether the corresponding first data point is a contour point based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set, including: sequentially calculating the first distance and the second distance from the first center coordinates to the other data points in the first point set, excluding the second data point; when there is a second data point in the first point set corresponding to the first data point that satisfies that the first distance or the second distance corresponding to the other data points, excluding the second data point, is greater than the second threshold, the first data point is determined to be a contour point.

[0138] Furthermore, the training dataset update module 502 is also used to update the training dataset according to the test data and the contour surface, including: when the test data is not included in the contour surface, adding the test data to the training dataset and removing the test data from the test dataset.

[0139] Furthermore, the prediction model fitting module 503 is used to adjust the training parameters of the multiple linear regression fitting process, including: adjusting the minimum error parameter of the multiple linear regression fitting and adjusting the sensitive data ratio of the historical hourly running data.

[0140] Furthermore, the data processing module 501 is used to preprocess the historical hourly running data to obtain a test dataset and a training dataset, including: cleaning the historical hourly running data, dividing the cleaned historical hourly running data according to a preset ratio, and obtaining the test dataset and the training dataset.

[0141] Example 4

[0142] Please refer to Figure 6 A chiller unit energy consumption prediction device provided in this embodiment of the invention includes: a model acquisition module 601 and an energy consumption calculation module 602.

[0143] Furthermore, the model acquisition module 601 is used to acquire the chiller unit energy consumption prediction model and the contour surface extracted from the training dataset; wherein, the chiller unit energy consumption prediction model is trained by the chiller unit energy consumption prediction model training device described in any of the above embodiments; the contour surface is obtained by the chiller unit energy consumption prediction model training device described in any of the above embodiments; the energy consumption calculation module 602 is used to acquire the current operating data of the chiller unit, and when the operating data is included in the contour surface, the operating data is input to the energy consumption prediction model to obtain the energy consumption prediction result.

[0144] Example 5

[0145] Based on the above-described embodiments of the chiller unit energy consumption prediction model training method and energy consumption prediction method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the chiller unit energy consumption prediction model training method and energy consumption prediction method of any embodiment of the present invention.

[0146] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0147] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0148] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device through various interfaces and lines. The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store the operating system, at least one application program required for a function, etc.; the data storage area can store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0149] Example 6

[0150] Based on the above embodiments of the chiller unit energy consumption prediction model training method and energy consumption prediction method, another embodiment of the present invention provides a storage medium, the storage medium including a stored computer program, wherein, when the computer program is running, the device where the storage medium is located controls the execution of the chiller unit energy consumption prediction model training method and energy consumption prediction method of any embodiment of the present invention.

[0151] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0152] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A training method for a chiller unit energy consumption prediction model, characterized in that, include: Acquire historical hourly operating data of the chiller unit and a preset first chiller unit energy consumption prediction model, and preprocess the historical hourly operating data to obtain a test dataset and a training dataset. The historical hourly operating data includes: chilled water outlet temperature, cooling water inlet temperature, instantaneous output cooling capacity, and instantaneous power. Test data is randomly extracted from the test dataset in sequence, the contour surface of the training dataset is extracted, and the training dataset is updated according to the test data and the contour surface. Based on the updated training dataset, the energy consumption prediction model of the first chiller unit is iteratively fitted through multiple linear regression until the accuracy of the energy consumption prediction model of the second chiller unit fitted in the current iteration reaches a preset threshold, and the energy consumption prediction model of the second chiller unit is output; wherein, after each iteration, if the accuracy does not reach the preset threshold, the training parameters of the multiple linear regression fitting process are adjusted. The step of extracting the contour surface includes: The first data point in the training dataset is sequentially traversed, and all second data points in the training dataset whose distance from the first data point is less than a first threshold are obtained to generate a first point set corresponding to each training data point. Iterate through each second data point in the first point set corresponding to each first data point, and calculate the coordinates of the first center and the second center of each second data point. Based on the first center coordinates and the second center coordinates corresponding to each second data point in the first point set, determine whether the corresponding first data point is a contour point; Generate the contour surface based on all the contour points; The step of updating the training dataset based on the test data and the contour surface includes: adding the test data to the training dataset when the test data is not included in the contour surface, and removing the test data from the test dataset; The method for calculating the coordinates of the first center and the second center includes: The formula for calculating the coordinates of the first circle center specifically includes: in, The x-axis coordinate value of the first circle center; The vertical coordinate value of the first circle's center. This represents the x-axis coordinate of the first data point; The vertical coordinate value of the first data point; This represents the x-axis coordinate of the second data point; This represents the ordinate value of the second data point; It is half of the first threshold; The formula for calculating the coordinates of the second center of the circle specifically includes: in, The x-axis coordinate value of the second center point; The vertical coordinate value of the second center point; This represents the x-axis coordinate of the first data point; The vertical coordinate value of the first data point; This represents the x-axis coordinate of the second data point; This represents the ordinate value of the second data point; It is half of the first threshold.

2. The method for training a chiller unit energy consumption prediction model as described in claim 1, characterized in that, The step of determining whether a corresponding first data point is a contour point based on the first center coordinates and the second center coordinates of each second data point in the first point set includes: Calculate sequentially the first distance from the coordinates of the first circle center and the second distance from the coordinates of the second circle center for each data point in the first point set, excluding the second data point. When there exists a second data point in the first set of points corresponding to the first data point that satisfies the condition that the first distance or second distance corresponding to other data points, excluding the second data point, is greater than the second threshold, the first data point is determined to be a contour point.

3. The method for training a chiller unit energy consumption prediction model as described in claim 1, characterized in that, The adjustment of training parameters in the multiple linear regression fitting process includes: adjusting the minimum error parameter of the multiple linear regression fitting and adjusting the proportion of sensitive data in the historical hourly running data.

4. The method for training a chiller unit energy consumption prediction model as described in claim 1, characterized in that, The preprocessing of the historical hourly running data to obtain the test dataset and the training dataset includes: cleaning the historical hourly running data, dividing the cleaned historical hourly running data according to a preset ratio, and obtaining the test dataset and the training dataset.

5. A method for predicting the energy consumption of a chiller unit, characterized in that, include: A chiller unit energy consumption prediction model and a contour surface extracted from a training dataset are obtained; wherein the chiller unit energy consumption prediction model is trained using the chiller unit energy consumption prediction model training method according to any one of claims 1 to 4; and the contour surface is extracted using the chiller unit energy consumption prediction model training method according to any one of claims 1 to 2. The current operating data of the chiller unit is obtained. When the operating data is contained in the contour surface, the operating data is input into the energy consumption prediction model to obtain the energy consumption prediction result.

6. A training device for a chiller unit energy consumption prediction model, characterized in that, include: The module includes a data processing module, a training dataset update module, and a prediction model fitting module. The data processing module is used to acquire historical hourly operating data of the chiller unit and a preset first chiller unit energy consumption prediction model, and preprocess the historical hourly operating data to acquire a test dataset and a training dataset; the historical hourly operating data includes: chilled water outlet temperature, cooling water inlet temperature, instantaneous output cooling capacity, and instantaneous power. The training dataset update module is used to randomly extract test data from the test dataset, extract the contour surface of the training dataset, and update the training dataset according to the test data and the contour surface. The prediction model fitting module is used to iteratively fit the first chiller unit energy consumption prediction model according to the updated training dataset through multiple linear regression until the accuracy of the second chiller unit energy consumption prediction model fitted in the current iteration reaches a preset threshold, and then output the second chiller unit energy consumption prediction model; wherein, after each iteration, if the accuracy does not reach the preset threshold, the training parameters of the multiple linear regression fitting process are adjusted. The training dataset update module includes: a first point set acquisition unit, a circle center coordinate calculation unit, a contour point judgment unit, and a contour surface generation unit; The first point set acquisition unit is used to sequentially traverse the first data point in the training dataset, acquire all second data points in the training dataset whose distance from the first data point is less than a first threshold, and generate a first point set corresponding to each training data. The circle center coordinate calculation unit is used to sequentially traverse each second data point in the first point set corresponding to each first data point, and calculate the first circle center coordinate and the second circle center coordinate corresponding to each second data point; The contour point determination unit is used to determine whether the corresponding first data point is a contour point based on the first center coordinates and the second center coordinates of each second data point in the first point set. The contour surface generation unit is used to generate the contour surface based on all the contour points; The training dataset update module is further configured to update the training dataset based on the test data and the contour surface, including: when the test data is not included in the contour surface, adding the test data to the training dataset and removing the test data from the test dataset; The method for calculating the coordinates of the first center and the second center includes: The formula for calculating the coordinates of the first circle center specifically includes: in, The x-axis coordinate value of the first circle center; The vertical coordinate value of the first circle's center. This represents the x-axis coordinate of the first data point; The vertical coordinate value of the first data point; This represents the x-axis coordinate of the second data point; This represents the ordinate value of the second data point; It is half of the first threshold; The formula for calculating the coordinates of the second center of the circle specifically includes: in, The x-axis coordinate value of the second center point; The vertical coordinate value of the second center point; This represents the x-axis coordinate of the first data point; The vertical coordinate value of the first data point; This represents the x-axis coordinate of the second data point; This represents the ordinate value of the second data point; It is half of the first threshold.

7. The chiller unit energy consumption prediction model training device as described in claim 6, characterized in that, The contour point determination unit is used to determine whether a corresponding first data point is a contour point based on the first center coordinates and the second center coordinates of each second data point in the first point set, including: Calculate sequentially the first distance from the coordinates of the first circle center and the second distance from the coordinates of the second circle center for each data point in the first point set, excluding the second data point. When there exists a second data point in the first set of points corresponding to the first data point that satisfies the condition that the first distance or second distance corresponding to other data points, excluding the second data point, is greater than the second threshold, the first data point is determined to be a contour point.

8. The chiller unit energy consumption prediction model training device as described in claim 6, characterized in that, The prediction model fitting module is used to adjust the training parameters of the multiple linear regression fitting process, including: adjusting the minimum error parameter of the multiple linear regression fitting and adjusting the sensitive data ratio of the historical hourly running data.

9. The chiller unit energy consumption prediction model training device as described in claim 6, characterized in that, The data processing module is used to preprocess the historical hourly running data and obtain a test dataset and a training dataset, including: cleaning the historical hourly running data, dividing the cleaned historical hourly running data according to a preset ratio, and obtaining the test dataset and the training dataset.

10. A chiller unit energy consumption prediction device, characterized in that, include: Model acquisition module and energy consumption calculation module; The model acquisition module is used to acquire a chiller unit energy consumption prediction model and a contour surface extracted from a training dataset; wherein the chiller unit energy consumption prediction model is trained using the chiller unit energy consumption prediction model training device according to any one of claims 6 to 9; and the contour surface is acquired using the chiller unit energy consumption prediction model training device according to any one of claims 6 to 7. The energy consumption calculation module is used to obtain the current operating data of the chiller unit. When the operating data is included in the contour surface, the operating data is input into the energy consumption prediction model to obtain the energy consumption prediction result.

Images