A cold and heat source host modeling method based on sample data derivation and incremental learning

Abstract

The application discloses a cold and heat source host modeling method based on sample data derivation and incremental learning, and has the characteristics that performance curves of the same type of cold and heat source host under different models have similar change trends, sample operation condition data of a certain model of the same type of cold and heat source host needing modeling are selected as a benchmark, data derivation is carried out on a target cold and heat source host, model fitting is carried out by using the derived data, finally, incremental learning is carried out on the model by using actual operation data of the target cold and heat source host through clustering to extract typical working conditions, and then the model suitable for the cold and heat source host is obtained. The method is easy to extend to modeling of cold and heat source hosts of different types and different models, and has a good practical engineering application prospect.

Description

Technical Field

[0001] This invention belongs to the field of heating, ventilation and air conditioning, and specifically relates to a method for modeling cold and heat source host based on sample data derivation and incremental learning. Background Technology

[0002] The chiller / heat source unit is a high-energy-consuming device in the HVAC system. In order to predict the operation of the chiller / heat source unit under different operating conditions, summarize the operating rules of the chiller / heat source unit, and optimize the energy-saving operation of the equipment, it is necessary to model the chiller / heat source unit to facilitate unified management of the equipment in the computer room.

[0003] In current engineering applications, the modeling of cold and heat source hosts typically relies on a period of operational data. When the amount and quality of data do not meet the modeling requirements, it is often difficult to obtain a model for the cold and heat source host. Furthermore, relying solely on actual operational data from the cold and heat source host for modeling results in overly concentrated data, as the operating conditions of the equipment change relatively little over a certain period. This can easily lead to overfitting of the model, making it unable to predict operating conditions outside the training scenarios. Therefore, there is an urgent need for a modeling method that can quickly and accurately build cold and heat source host models in scenarios with limited sample sizes, and that can continuously update and adjust the model based on real operating conditions. Summary of the Invention

[0004] To address the aforementioned issues, this invention discloses a cold and hot source host modeling method based on sample data derivation and incremental learning. This method can quickly and accurately establish cold and hot source host models in scenarios with few samples, and can continuously update and adjust its own model according to actual operating conditions. This ensures that the model not only conforms to the actual operating rules of cold and hot source hosts but can also predict data under all operating conditions.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A method for modeling hot and cold source hosts based on sample data derivation and incremental learning includes the following steps:

[0007] S1: Obtain the operating condition data of the sample unit;

[0008] S2: Derivation of sample data based on nominal operating condition data on the nameplate of the on-site heat and cold source unit;

[0009] S3: Use sample-derived data to fit the model and obtain the data-derived model;

[0010] S4: Collect the actual operating data of the on-site cold and heat source host, cluster them to obtain k clusters, and the cluster centers of the k clusters are the typical on-site operating conditions;

[0011] S5: Using the typical on-site working conditions obtained in S4, the data derived from S3 is incrementally learned to obtain the final prediction model.

[0012] Furthermore, in step S1:

[0013] The collected operating data includes the cooling or heating capacity of the chiller unit under different cooling water temperatures, different chilled water temperatures, and different load rates, as well as the cooling or heating efficiency (COP) values.

[0014] The range of variation of cooling water temperature, chilled water temperature, and load rate in the sample data needs to include all possible values ​​within the operating range of the heat source unit.

[0015] The sample unit needs to be of the same type as the target unit on site.

[0016] Further, step S2 includes:

[0017] S21: Calculate the ratio of the nominal operating condition of the on-site cold and heat source host to the nominal operating condition of the sample cold and heat source host, thereby scaling up or down all the cooling or heating capacity of the sample data.

[0018] S22: Calculate the COP ratio between the nominal operating conditions of the on-site cold and heat source host and the nominal operating conditions of the sample cold and heat source host, thereby scaling up or down all cooling or heating capacity of the sample data.

[0019] Further, step S3 includes

[0020] When performing data-derived simulation fitting in step S3, the formula for the model to be fitted is:

[0021] COP = a1 × T e，out +a2×T c，in +a3×T c，out +a4×Q+a5×Q×Q+a6×T e，out ×Q+a7×T c，in ×Q+a8×T c，out ×Q+a9;

[0022] Where a1, a2...a9 are the parameters to be fitted, COP is the cooling efficiency or heating efficiency, and T... e，out T represents the chilled water outlet temperature of the chiller / heat source unit. c，in T represents the inlet temperature of the cooling water for the main unit, which is the source of both heat and cold water. c，out The outlet temperature of the cooling water of the heat source unit is denoted as , and Q is the cooling capacity or heating capacity of the heat source unit. The regression fitting method used is Lasso regression.

[0023] Furthermore, when the sample data does not include the outlet water temperature of the cooling water from the heat source and cold source units, the following method is used to calculate and obtain sample data containing the outlet water temperature of the cooling water from the heat source and cold source units:

[0024] S31. Calculate the rated cooling water flow rate G of the cold and heat source host under the nominal operating conditions using the cooling capacity or heating capacity and power in the sample data under the nominal operating conditions;

[0025] S32. Assume that the cooling water supply and return water temperature difference is 5℃ in the sample data operating conditions, and then calculate the cooling water flow rate for all operating conditions in the sample data;

[0026] S33. When the cooling water flow rate calculated in step S32 is greater than 30% of the rated cooling water flow rate G of the heat source unit, the cooling water outlet temperature of the heat source unit is calculated using the following formula:

[0027] T c，out =T c，in +5;

[0028] When the cooling water flow rate calculated in step S32 is less than or equal to 30% of the rated cooling water flow rate G of the heat source unit, its flow rate is regarded as 30% of the rated flow rate, and then the cooling water outlet temperature of the heat source unit is calculated using the following formula:

[0029] T c，out = (Q + Q / COP) / c / (G × 0.3) + T c，in ;

[0030] Where c is the specific heat capacity of water.

[0031] Further, step S4 includes:

[0032] S41: Collect the start-up and shutdown status, chilled water inlet and outlet water temperatures, cooling water inlet and outlet water temperatures, load rate, cooling capacity or heating capacity and power data of the on-site chiller and heat source unit during operation, and calculate the cooling efficiency or heating efficiency of the chiller and heat source unit using the formula COP=Q / W, where Q is the cooling capacity or heating capacity of the chiller and heat source unit and W is the power of the chiller and heat source unit.

[0033] S42: Preprocess the data collected in S41, specifically including: converting non-numerical operating parameter data of the heat source host equipment into numerical values, such as: device on / off status, using 1 to represent on and 0 to represent off; removing and correcting outliers based on the actual possible range of operating parameters, such as: the collected temperature exceeding the actual possible operating temperature range of the heat source host; filling missing signal parameters with interpolation or extrapolation using data from before and after the missing time points.

[0034] S43: Use the K-means clustering method to perform clustering analysis on the dataset D processed by S42:

[0035] Further, step S43 includes:

[0036] S431: Randomly select k data points from dataset D as cluster centers;

[0037] S432: Then, for each data point in dataset D, find the nearest cluster center and assign the data point to the cluster corresponding to that cluster center;

[0038] S433: Update the cluster center of each cluster to the mean of all data points in that cluster;

[0039] S434: Repeat steps S432 and S433 until the cluster center position no longer changes.

[0040] S44: Define the k cluster centers obtained in step S43 as typical on-site working conditions.

[0041] Further, step S5 includes:

[0042] The data-derived model obtained in step S3 incorporates the k typical field conditions obtained in step S4 into the dataset for incremental learning, and corrects the fitting parameters a1, a2...a9 already obtained in the data-derived model to form the final prediction model.

[0043] The beneficial effects of this invention are:

[0044] This invention leverages the similar performance trends of different models of the same type of heat source / cooler unit. It selects sample operating data from a specific model of the heat source / cooler unit of the type to be modeled as a benchmark. Data is derived from the target heat source / cooler unit, and the derived data is used for model fitting. Finally, a small amount of actual operating data from the target heat source / cooler unit is used to extract typical operating conditions through clustering for incremental model learning, thereby obtaining a model adapted to that heat source / cooler unit. This method is easily extendable to modeling different types and models of heat source / cooler units, and has good prospects for practical engineering applications. Attached Figure Description

[0045] Figure 1 This is a flowchart of the present invention;

[0046] Figure 2 This is a flowchart illustrating the process of deriving sample data from the nominal operating condition data on the nameplate of the on-site cold and heat source host in the modeling method described in this invention.

[0047] Figure 3 This is a flowchart illustrating the cluster analysis using real-world data in the modeling method described in this invention.

[0048] Figure 4 This is a scatter plot of the COP prediction results of the final prediction model obtained in a specific embodiment of the modeling method described in this invention and the actual operating conditions. Detailed Implementation

[0049] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0050] like Figure 1 As shown, the cold and hot source host modeling method based on sample data derivation and incremental learning described in this invention includes:

[0051] S1: Obtain the operating condition data of the sample unit;

[0052] S2: Derivation of sample data based on nominal operating condition data on the nameplate of the on-site heat and cold source unit;

[0053] S3: Use sample-derived data to fit the model and obtain the data-derived model;

[0054] S4: Collect the actual operating data of the on-site cold and heat source host, cluster them to obtain k clusters, and the cluster centers of the k clusters are the typical on-site operating conditions;

[0055] S5: Using the typical on-site working conditions obtained in S4, incremental learning is performed on the data derived model in S3 to obtain the final prediction model.

[0056] In a specific embodiment, in step S1:

[0057] The collected operating data includes the cooling or heating capacity of the chiller unit under different cooling water temperatures, different chilled water temperatures, and different load rates, as well as the cooling or heating efficiency (COP) values.

[0058] The range of variation of cooling water temperature, chilled water temperature, and load rate in the sample data needs to include all possible values ​​within the operating range of the heat source unit.

[0059] The sample unit needs to be of the same type as the target unit on site.

[0060] In a specific embodiment, the sample unit data collected in step S1 comes from the Type 142 water-cooled cold and heat source host dataset in TRNSYS18.

[0061] Specifically, step S2 includes:

[0062] S21: Calculate the ratio of the nominal operating condition of the on-site cold and heat source host to the nominal operating condition of the sample cold and heat source host, thereby scaling up or down all the cooling or heating capacity of the sample data.

[0063] S22: Calculate the COP ratio between the nominal operating conditions of the on-site cold and heat source host and the nominal operating conditions of the sample cold and heat source host, thereby scaling up or down all cooling or heating capacity of the sample data.

[0064] Specifically, step S3 includes:

[0065] When performing data-derived simulation fitting, the Lasso regression fitting method is used, and the formula for the model to be fitted is:

[0066] COP = a1 × T e，out +a2×T c，in +a3×T c，out +a4×Q+a5×Q×Q+a6×T e，out ×Q+a7×T c，in ×Q+a8×T c，out ×Q+a9;

[0067] Where a1, a2...a9 are the parameters to be fitted, COP is the cooling efficiency or heating efficiency, and T... e，out T represents the chilled water outlet temperature of the chiller / heat source unit. c，in T represents the inlet temperature of the cooling water for the main unit, which is the source of both heat and cold water. c，out Q represents the outlet temperature of the cooling water from the heat source unit, and Q represents the cooling or heating capacity of the heat source unit.

[0068] Lasso regression is a regularized linear regression model that adds regular expressions to the least squares regression to constrain the parameters. In practice, the penalty term alpha is set to 0.005.

[0069] When the sample data does not include the outlet temperature of the cooling water from the heat source and cold source units, the following method is used to calculate and obtain sample data that includes the outlet temperature of the cooling water from the heat source and cold source units:

[0070] S31. Calculate the rated cooling water flow rate G of the cold and heat source host under the nominal operating conditions using the cooling capacity or heating capacity and power in the sample data under the nominal operating conditions;

[0071] S32. Assume that the cooling water supply and return water temperature difference is 5℃ in the sample data operating conditions, and then calculate the cooling water flow rate for all operating conditions in the sample data;

[0072] S33. When the cooling water flow rate calculated in step S32 is greater than 30% of the rated cooling water flow rate G of the heat source unit, the cooling water outlet temperature of the heat source unit is calculated using the following formula:

[0073] T c，out =T c，in +5;

[0074] When the cooling water flow rate calculated in step S32 is less than or equal to 30% of the rated cooling water flow rate G of the heat source unit, its flow rate is regarded as 30% of the rated flow rate, and then the cooling water outlet temperature of the heat source unit is calculated using the following formula:

[0075] T c，out = (Q + Q / COP) / c / (G × 0.3) + T c，in ;

[0076] Where c is the specific heat capacity of water.

[0077] As a specific embodiment, the model formula obtained by fitting the Type 142 water-cooled chiller host dataset in TRNSYS18 through step S3 is as follows:

[0078] COP = 9.162 × 10 -2 ×T e，out +1.802×10 -2 ×T c，in -1.606×10 -1 ×T c，out +6.543×10 -4 ×Q-8.206×10 81 ×Q×Q+1.708×10 -5 ×T e，out ×Q-1.548×10 -5 ×T c，in ×Q+1.016×10 -5 ×T c，out ×Q+10.269;

[0079] Specifically, step S4 includes:

[0080] S41: Collect the start-up and shutdown status, chilled water inlet and outlet water temperatures, cooling water inlet and outlet water temperatures, load rate, cooling capacity or heating capacity and power data of the on-site chiller and heat source unit during operation, and calculate the cooling efficiency or heating efficiency of the chiller and heat source unit using the formula COP=Q / W, where Q is the cooling capacity or heating capacity of the chiller and heat source unit and W is the power of the chiller and heat source unit.

[0081] S42: Preprocess the data collected in S41, specifically including: converting non-numerical operating parameter data of the heat source host equipment into numerical values, such as: device on / off status, using 1 to represent on and 0 to represent off; removing and correcting outliers based on the actual possible range of operating parameters, such as: the collected temperature exceeding the actual possible operating temperature range of the heat source host; filling missing signal parameters with interpolation or extrapolation using data from before and after the missing time points.

[0082] S43: Use the K-means clustering method to perform clustering analysis on the dataset D processed by S42:

[0083] Specifically, step S43 is as follows: Figure 3 As shown, it includes:

[0084] S431: Randomly select k data points from dataset D as cluster centers.

[0085] S432: Then, for each data point in dataset D, find the nearest cluster center and assign that data point to the cluster corresponding to that cluster center.

[0086] S433: Update the cluster center of each cluster to the mean of all data points in that cluster;

[0087] S434: Repeat steps S432 and S433 until the cluster center position no longer changes.

[0088] S44: Define the k cluster centers obtained in step S43 as typical field conditions.

[0089] In a specific embodiment, the three cluster centers obtained after step S4 of the operating data of a certain chiller are: [[8.15, 27.22, 4252.67, 30.16, 6.51], [8.18, 22.84, 3638.67, 25.27, 6.74], [8.04, 29.86, 4730.78, 32.97, 6.33]], where the first to fifth items are respectively: chilled water outlet temperature, cooling water inlet temperature, cooling capacity, cooling water outlet temperature, and COP.

[0090] In a specific embodiment, step S5 includes:

[0091] The data derived model obtained in step S3 sequentially incorporates the k typical on-site working conditions obtained in step S4 into the dataset for incremental learning, and corrects the fitting parameters a1, a2...a9 already obtained in the data derived model to form the final prediction model.

[0092] As a specific example, the final prediction model after parameter correction for a typical on-site operating condition of a chiller is as follows:

[0093] COP = 9.162 × 10 -2 ×T e，out +1.802×10 -2 ×T c，in -1.606×10 -1 ×T c，out +6.543×10 -4 ×Q-8.206×10 81 ×Q×Q+1.708×10 -5 ×T e，out ×Q-1.548×10 -5 ×T c，in ×Q+1.016×10 -5 ×T c，out ×Q+9.047;

[0094] As a specific embodiment, such as Figure 4 The model's root mean square error (RMSE) for predicting the COP of a chiller unit during actual operation is 0.499, demonstrating that the model has predictive ability and high accuracy.

[0095] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A method for modeling hot and cold source hosts based on sample data derivation and incremental learning, characterized in that, include: S1: Obtain the operating condition data of the sample unit; S2: Derivation of sample data based on nominal operating condition data on the nameplate of the on-site heat and cold source unit; Specifically, it includes: S21: Calculate the ratio A between the nominal operating conditions of the on-site cold and heat source host and the nominal operating conditions of the sample cold and heat source host, and multiply the cooling or heating data in the operating conditions data of the sample unit by the ratio A, thereby scaling all cooling or heating data in the sample data. S22: Calculate the ratio B between the nominal operating conditions of the on-site chiller / heater unit and the nominal operating conditions of the sample chiller / heater unit, and multiply all the chiller / heater efficiency data in the sample unit's operating data by the ratio B, thereby scaling all chiller / heater efficiency data in the sample data. S3: Use the sample-derived data obtained in S2 to perform model fitting and obtain a data-derived model; When performing data-derived simulation fitting, the formula for the model to be fitted is: ； in, Here are the parameters to be fitted, and COP is the cooling efficiency or heating efficiency. The chilled water outlet temperature of the cold and heat source unit. The inlet temperature of the cooling water for the main unit, which is the source of both cold and heat sources. The outlet temperature of the cooling water for the main unit, which is the source of both cold and heat sources. For the cooling or heating capacity of the cold / heat source unit, the regression fitting method used is Lasso regression. When the sample data does not include the inlet water temperature of the cold and heat source main unit, the following method is used to calculate and obtain sample data that includes the inlet water temperature of the cold and heat source main unit: S31. Calculate the rated cooling water flow rate of the chiller / heat source unit under nominal operating conditions using the cooling capacity or heating capacity and power in the sample data. ; S32. Assume that the cooling water supply and return water temperature difference is 5℃ in the sample data operating conditions, and then calculate the cooling water flow rate for all operating conditions in the sample data; S33. When the cooling water flow rate calculated in step S32 is greater than the rated cooling water flow rate of the heat source unit. When the temperature is 30% of the rated value, the outlet temperature of the cooling water from the heat source unit is calculated using the following formula: ； When the cooling water flow rate calculated in step S32 is less than or equal to the rated cooling water flow rate of the heat source unit When the flow rate is 30% of the rated flow rate, it is considered as 30% of the rated flow rate. The outlet temperature of the cooling water for the heat source / cold source unit is then calculated using the following formula: ； in, This is the specific heat capacity of water; S4: Collect actual operating data of the on-site cold and heat source host, cluster the actual operating data of the on-site cold and heat source host to obtain k clusters, and the cluster center of the k clusters is the typical on-site operating condition. S5: The data-derived model obtained in step S3 incorporates the k typical on-site working conditions obtained in step S4 into the dataset for incremental learning, and applies the data already obtained in the data-derived model... The fitting parameters are then corrected to form the final prediction model.

2. The method for modeling hot and cold source hosts based on sample data derivation and incremental learning as described in claim 1, characterized in that: The collected operating data includes the cooling or heating capacity of the chiller unit under different cooling water temperatures, different chilled water temperatures, and different load rates, as well as the values ​​of cooling or heating efficiency. The range of variation of cooling water temperature, chilled water temperature, and load rate in the sample data needs to include all possible values ​​within the operating range of the heat source unit. The sample unit needs to be of the same type as the target unit on site.

3. The method for modeling hot and cold source hosts based on sample data derivation and incremental learning as described in claim 1, characterized in that, Step S4 includes: S41: Collect data on the start / stop status, chilled water inlet and outlet temperatures, cooling water inlet and outlet temperatures, load rate, cooling capacity or heating capacity, and power during the operation of the on-site chiller / heater unit, and use the formula... Calculate the cooling or heating efficiency of the cold / heat source unit, where... Power of the heat source / cold source unit; S42: Preprocess the data collected in S41, specifically including: converting the non-numerical operating parameter data of the cold and heat source host equipment into numerical data; and removing and correcting outliers based on the actual possible range of the operating parameters. S43: Use the K-means clustering method to perform clustering analysis on the dataset D processed in S42. The specific steps are as follows: S431: Randomly select k data points from dataset D as cluster centers; S432: Then, for each data point in dataset D, find the nearest cluster center and assign the data point to the cluster corresponding to that cluster center; S433: Update the cluster center of each cluster to the mean of all data points in that cluster; S434: Repeat steps S432 and S433 until the cluster center position no longer changes; S44: Define the k cluster centers obtained in step S43 as typical working conditions in the field.

Images