Air conditioning system energy efficiency optimization method based on load prediction

Abstract

The invention provides an air conditioning system energy efficiency optimization method based on load prediction, and the method comprises the steps: collecting air conditioning load related data, building a load prediction model which employs a support vector regression model, and training the support vector regression model through the air conditioning load related data; constructing a model function by adopting a mixed kernel function of a linear kernel and a Gaussian kernel in the support vector regression model, and obtaining an air conditioner load prediction result according to the load prediction model; an energy efficiency optimization model is constructed, the energy efficiency optimization model adopts a mixed integer linear programming model, the mixed integer linear programming model takes the total energy consumption minimization of the air conditioning system as a target function, energy consumption of various devices in the air conditioning system is considered, and an energy consumption function is established; the energy consumption function represents the energy consumption under the given control variable and the air conditioner load prediction result. The operation state and parameters of the air conditioning system are adjusted in real time according to the prediction load and the result of the optimization model, and energy-saving operation of the system is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of air - conditioner energy - efficiency optimization control, and more particularly, to an energy - efficiency optimization method for an air - conditioner system based on load prediction. Background Art

[0002] Currently, with the growth of global energy demand and the enhancement of environmental protection awareness, improving energy utilization efficiency has become an important topic. Analyzing the composition of building energy consumption, it can be found that as the main component of building energy consumption, the air - conditioner system, with the increasing demand for building environmental control, accounts for an increasing share of the total energy consumption. Therefore, the improvement of its energy efficiency has become a research hotspot.

[0003] Accurately predicting the air - conditioner load and optimizing the system operation accordingly is the key to reducing air - conditioner energy consumption. Traditional air - conditioner load prediction methods (such as time - series models) usually rely on simple statistical models or empirical formulas, which are difficult to accurately capture the relationship between complex environmental changes and dynamic loads, have poor adaptability to non - linear load data, and have insufficient prediction accuracy.

[0004] In recent years, machine - learning technologies have developed rapidly and are widely used in engineering modeling to predict various engineering parameters. However, limited by the performance of hardware devices, deep - learning models (such as LSTM) rely on massive data and have high computational costs, making it difficult to be deployed in medium - sized and small - sized scenarios. Moreover, existing optimization models are not deeply coupled with load prediction, do not fully consider the uncertainty of load prediction, mainly focus on static optimization, lack collaborative optimization of multiple objectives, and it is difficult to balance real - time performance and computational complexity in the actual operation of the system, resulting in poor energy - saving effects.

[0005] Support Vector Regression (SVR), a regression analysis method based on support vector machines, performs stably in small - sample and non - linear data, is suitable for load - fluctuation scenarios and shows good performance in load prediction. However, how to effectively apply these prediction results to the overall system energy - efficiency optimization remains a challenge. Summary of the Invention

[0006] The present invention aims to solve at least one of the above - mentioned technical problems in the prior art.

[0007] To this end, the present invention provides an energy - efficiency optimization method for an air - conditioner system based on load prediction.

[0008] The energy - efficiency optimization method for an air - conditioner system based on load prediction provided by the present invention includes:

[0009] Collecting air - conditioner load - related data, where the air - conditioner load - related data includes outdoor meteorological data, indoor environmental parameters, and historical air - conditioner load data;

[0010] Constructing a load forecasting model, wherein the load forecasting model adopts a support vector regression model and is trained using the air conditioning load related data; constructing a model function using a hybrid kernel function of a linear kernel and a Gaussian kernel in the support vector regression model, wherein the linear kernel is used to capture the air conditioning load trend and the Gaussian kernel is used to capture local fluctuations; and obtaining an air conditioning load forecast result according to the load forecasting model;

[0011] An energy efficiency optimization model is constructed. The energy efficiency optimization model adopts a mixed integer linear programming model. The mixed integer linear programming model takes minimizing the total energy consumption of the air-conditioning system as the objective function, considers the energy consumption of various equipment in the air-conditioning system, and establishes an energy consumption function; the energy consumption function represents the energy consumption under given control variables and air-conditioning load prediction results.

[0012] The air conditioning system energy efficiency optimization method based on load forecasting according to the above technical solution of the present invention may also have the following additional technical features:

[0013] In the above technical solution, the model function of the support vector regression model is:

[0014]

[0015] Where N is the number of samples; i is the sample number; α i represents the Lagrange multiplier; represents the dual form of the Lagrange multiplier; x represents the input data; x i represents the sample of input training data; b represents the bias term; K mix (x,x i ) represents the mixed kernel function;

[0016] The expression of the mixed kernel function is as follows:

[0017] K mix (x,x i )=αK lin (x,x i )+(1-α)K rbf (x,x i )

[0018] K lin (x,x i )=x T x i

[0019] K rbf (x,x i )=exp(-γ||xx i || 2 )

[0020] Among them, Klin (x, x i ) represents a linear kernel function; K rbf (x, x i ) represents a Gaussian kernel function; α represents the weight coefficient of the mixed kernel function, which is used to balance the contributions of the linear kernel function and the Gaussian kernel function; γ represents the width coefficient of the Gaussian function.

[0021] In the above technical solution, the selection of the weight coefficient of the mixed kernel function is related to the humidity parameter in the historical air-conditioning load data. The higher the humidity control requirement, the smaller the value of the weight coefficient of the mixed kernel function.

[0022] In the above technical solution, in the load prediction model, the defined input data includes outdoor meteorological data, indoor environmental parameters, and air-conditioning equipment power in the air-conditioning load-related data, and the defined output data includes air-conditioning load data in the air-conditioning load-related data;

[0023] The method for training the load prediction model includes:

[0024] Dividing the air-conditioning load-related data into a training set and a test set according to a set ratio;

[0025] Training the load prediction model based on the training set data, where a hyperparameter optimization method is used to traverse all parameter combinations in combination with cross-validation to find the best parameter combination;

[0026] After the training is completed, use the test set data to evaluate the model, and adjust the feature selection and parameters of the load prediction model according to the evaluation results.

[0027] In the above technical solution, the construction of the energy efficiency optimization model includes:

[0028] Constructing energy consumption models for various devices in the air-conditioning system, where the various devices in the air-conditioning system at least include chillers, circulating water pumps, cooling towers, and terminal air handling units;

[0029] Integrating the energy consumption models of various devices in the air-conditioning system to obtain the total energy consumption function of the air-conditioning system;

[0030] Integrating the air-conditioning load prediction result output by the load prediction model into the energy efficiency optimization model, taking the minimization of the total energy consumption of the air-conditioning system as the objective function, defining control variables and setting constraint conditions, and completing the modeling of the energy efficiency optimization model; among them, linearize the non-linear parts in the constraint conditions and the objective function.

[0031] In the above technical solution, the total energy consumption function is:

[0032]

[0033] Among them, represents the total energy consumption function, where represents the predicted air-conditioning load result output by the load prediction model; u t represents the control variable vector; E chiller (Q c , T cw , T ww ) represents the energy consumption model of the chiller, Q c represents the refrigeration load, T cw represents the cooling water temperature, T ww represents the chilled water temperature; E pump (Q p , H p ) represents the energy consumption model of the circulating water pump, Q p represents the pump flow rate, H p represents the pump head; E coolingtower (Q f ) represents the energy consumption model of the cooling tower, Q f represents the cooling tower fan air volume; E FCU (Q fc , P fc ) represents the energy consumption model of the fan coil unit, Q fc represents the fan coil unit air volume; P fc represents the pressure difference of the fan coil unit fan; E VAV (Q vav , P vav ) represents the energy consumption model of the variable air volume system, Q vav represents the variable air volume system air volume; P vav represents the pressure difference of the variable air volume system fan.

[0034] In the above technical solution, the expression of the energy efficiency optimization model is:

[0035]

[0036] where T represents the time period; t represents the moment.

[0037] In the above technical solution, the control variables include:

[0038] The operating state of the chiller, the operating state of the water pump, the operating state of the cooling tower fan, the operating state of the fan coil unit, the operating state of the variable air volume system, the set temperature, the refrigeration load, the pump flow rate, the cooling tower fan air volume, the fan coil unit air volume, and the variable air volume system air volume. [[ID=……]]

[0039] In the above technical solution, the constraint conditions include: equipment operating state constraints, temperature constraints, and robust load constraints;

[0040] Among them, the robust load constraints include:

[0041]

[0042] Ensure that the refrigeration load meets the air-conditioning load prediction results output by the load prediction model within the allowable deviation range through robust load constraints;

[0043] Among them, δ represents the maximum allowable deviation value. Based on historical load data records and according to the requirements of prediction accuracy (such as daily / hourly / minutely), the fluctuation range of the load at different time scales can be statistically analyzed to determine the deviation value.

[0044] In the above technical solution, in the energy efficiency optimization model, considering the uncertainty Δx of the input features and the uncertainty Δθ of the model parameters, robust optimization is carried out by maximizing the objective function in the worst case;

[0045] The introduction of the uncertainty of the input features includes:

[0046] x′ i = x i + Δx i , Δx i = ||x i - x i0 || ∈ Ω

[0047] The introduction of the uncertainty of the model parameters includes:

[0048] Δθ = ||θ - θ0|| ∈ Θ

[0049] Among them, x i0 is the original input feature, Ω represents the set of uncertainties of the input features; θ0 is the initial parameter of the model, and Θ represents the set of uncertainties of the model parameters.

[0050] In summary, due to the adoption of the above technical features, the beneficial effects of the present invention are:

[0051] The present invention combines the support vector regression (SVR) load prediction with the mixed integer linear programming (MILP) energy efficiency optimization model. By predicting the air-conditioning cooling load demand in the future period and using the predicted value as the input parameter of the system energy efficiency optimization model, an "prediction-optimization" analysis and decision-making framework is jointly constructed. Further clarify the pre-data preprocessing and the main feature vector screening steps to improve the accuracy of the air-conditioning load prediction, and fully consider the uncertainty of the load prediction for robust optimization to ensure the overall optimal energy efficiency of the air-conditioning system under dynamic load demands.

[0052] Specifically, by adopting the SVR model and combining with the hybrid kernel function, the complex non-linear relationship between the air-conditioning load and various influencing factors can be effectively captured, the accuracy of air-conditioning load prediction can be improved, the air-conditioning load demand can be grasped more accurately, and a reliable basis for energy efficiency optimization can be provided. Modeling the energy consumption characteristics of chillers, pumps, cooling towers and terminal air handling units (such as fan coils and variable air volume systems) systematically encompasses the main energy-consuming equipment and improves the accuracy of energy consumption prediction. Introducing robust optimization fully considers the uncertainty of load prediction and enhances the reliability and adaptability of the energy efficiency optimization model. The analysis results can be used to control by optimizing equipment parameters to achieve system energy consumption reduction, or to evaluate the overall energy-saving effect of the system, compare the energy efficiency of the system operation before and after optimization, and provide a basis for formulating energy-saving strategies.

[0053] The additional aspects and advantages of the present invention will become apparent in the following description section or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0055] Figure 1 is a flowchart of an energy efficiency optimization method for an air-conditioning system based on load prediction according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments may be combined with each other.

[0057] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0058] The following refers to Figure 1 to describe an energy efficiency optimization method for an air-conditioning system based on load prediction according to some embodiments of the present invention.

[0059] Some embodiments of the present application provide an energy efficiency optimization method for an air-conditioning system based on load prediction.

[0060] As Figure 1 shown, the first embodiment of the present invention proposes an energy efficiency optimization method for an air-conditioning system based on load prediction, including the following steps S1-S3.

[0061] S1. Collect air-conditioning load-related data, where the air-conditioning load-related data includes outdoor meteorological data, indoor environmental parameters, and historical air-conditioning load data. Specifically, environmental data and usage data inside and outside the building, that is, air-conditioning load-related data, can be obtained from various sensors or databases.

[0062] In some embodiments, the outdoor meteorological data includes outdoor temperature information, humidity information, etc.; the indoor environmental parameters include indoor temperature, number of people, and equipment power, etc.

[0063] In some embodiments, step S1 further includes organizing the data, and organizing the air-conditioning load-related data into a data set in a multi-input single-output form. For example, data such as outdoor temperature, indoor temperature, and equipment power are input data, and air-conditioning load data is output data.

[0064] In some embodiments, before performing subsequent steps, it is also necessary to clean, normalize, and select features of the data to improve the effect of model training. Specifically, outliers and missing values are removed through data cleaning, and normalization processing is performed.

[0065] It should be noted that the data set should be continuous multiple source parameters, and the acquisition time period should match the system control response time interval. The load prediction and energy efficiency optimization period of this disclosure is considered to be 15 minutes; at the same time, unstable noise data when the air-conditioning system is just started and running needs to be excluded.

[0066] S2. Construct a load prediction model. The load prediction model uses a support vector regression model, and uses the air-conditioning load-related data to train the support vector regression model; in the support vector regression model, a mixed kernel function of a linear kernel and a Gaussian kernel is used to construct the model function, where the linear kernel is used to capture the air-conditioning load trend, and the Gaussian kernel is used to capture local fluctuations; obtain the air-conditioning load prediction result according to the load prediction model.

[0067] Specifically, in non-linear SVR, a kernel function is introduced to map the input data to a high-dimensional feature space, so that a linear regression model can be found in the high-dimensional space.

[0068] In some embodiments, the model function of the support vector regression model is:

[0069]

[0070] Among them, N represents the number of samples; i represents the sample number; α i represents the Lagrange multiplier; represents the dual form of the Lagrange multiplier; x represents the input data; x i represents the sample of the input training data; b represents the bias term; K mix (x, xi ) represents a mixed kernel function;

[0071] The expression of the mixed kernel function is as follows:

[0072] K mix (x, x i ) = αK lin (x, x i ) + (1 - α)K rbf (x, x i )

[0073] K lin (x, x i ) = x T x i

[0074] K rbf (x, x i ) = exp(-γ||x - x i || 2 )

[0075] Among them, K lin (x, x i ) represents a linear kernel function; K rbf (x, x i ) represents a Gaussian kernel function; α represents the weight coefficient of the mixed kernel function, which is used to balance the contributions of the linear kernel function and the Gaussian kernel function; γ represents the width coefficient of the Gaussian function, which is a positive parameter, and the value set is [0.001, 0.01, 0.1, 1, 10, 100].

[0076] In some embodiments, the selection of the weight coefficient of the mixed kernel function is related to the humidity parameter in the historical air-conditioning load data, that is, it is allocated according to the humidity parameter in the historical data. The higher the humidity control requirement, the smaller the value of the weight coefficient of the mixed kernel function.

[0077] In a specific embodiment, in the load prediction model, the defined input data includes outdoor meteorological data, indoor environmental parameters, and air-conditioning equipment power in the air-conditioning load-related data, and the defined output data includes air-conditioning load data in the air-conditioning load-related data; then the method for training the load prediction model includes:

[0078] Use the scikit-learn library in Python or other machine learning tools to train the load prediction model, and divide the air-conditioning load-related data into a training set and a test set according to a set ratio; usually, a ratio of 70%-30% or 80%-20% is adopted.

[0079] The load prediction model is trained based on the training set data. Among them, a hyperparameter optimization method is adopted to traverse all parameter combinations in combination with cross-validation to find the best parameter combination. The hyperparameter optimization method can be methods such as Grid Search or Random Search. The prediction accuracy of the model is improved through the above parameter optimization process. It should be noted that the model parameter automatic optimization method is a well-known method in the art and will not be elaborated here.

[0080] After the training is completed, the model is evaluated using the test set data, and the feature selection and parameters of the load prediction model are adjusted according to the evaluation results. For example, the prediction performance of the model is judged according to indicators such as the mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE). If the performance does not meet the requirements, the feature selection or model parameters are readjusted until a satisfactory prediction effect is achieved.

[0081] S3. Construct an energy efficiency optimization model. The energy efficiency optimization model adopts a mixed-integer linear programming model. The mixed-integer linear programming model takes the minimization of the total energy consumption of the air-conditioning system as the objective function, considers the energy consumption of various types of equipment in the air-conditioning system, and establishes an energy consumption function. The energy consumption function represents the energy consumption under given control variables and air-conditioning load prediction results.

[0082] In some embodiments, the construction of the energy efficiency optimization model includes:

[0083] S31. Construct an energy consumption model for various types of equipment in the air-conditioning system. The various types of equipment in the air-conditioning system at least include a chiller, a circulating water pump, a cooling tower, and a terminal air handling unit. The terminal air handling unit includes fan coils and a variable air volume system.

[0084] In some embodiments, the energy consumption models for various types of equipment in the constructed air-conditioning system are as follows.

[0085] The energy consumption model of the chiller is:

[0086] E chiller (Q c ,T cw ,T ww )=a1Q c +a2(T cw -T ww )+b

[0087] Where Q c represents the cooling load, T cw represents the cooling water temperature, and T ww represents the chilled water temperature; a1, a2, and b all represent coefficients determined according to experimental or manufacturer data.

[0088] The energy consumption model of the circulating water pump is as follows:

[0089]

[0090] Among them, Q p represents the water pump flow rate, and H p represents the water pump head; ρ is the density of water; g is the acceleration due to gravity; η is the efficiency of the water pump.

[0091] The energy consumption model of the cooling tower is as follows:

[0092]

[0093] Among them, Q f represents the air volume of the cooling tower fan, and ΔP f represents the pressure difference of the cooling tower fan; η f represents the efficiency of the cooling tower fan.

[0094] The energy consumption model of the fan coil unit is as follows:

[0095]

[0096] Among them, Q fc represents the air volume of the fan coil unit, and P fc represents the pressure difference of the fan coil unit fan; η fc represents the efficiency of the fan coil unit fan.

[0097] The energy consumption model of the variable air volume system is as follows:

[0098]

[0099] Among them, Q vav represents the air volume of the variable air volume system, and P vav represents the pressure difference of the variable air volume system fan; η vav represents the efficiency of the variable air volume system fan.

[0100] S32. Integrate the energy consumption models of various devices in the air conditioning system to obtain the total energy consumption function of the air conditioning system.

[0101] In a specific embodiment, the total energy consumption function is as follows:

[0102]

[0103] Among them, represents the total energy consumption function, represents the air conditioning load prediction result output by the load prediction model; u t represents the control variable vector.

[0104] S33. Integrate the air-conditioning load prediction result output by the load prediction model into the energy efficiency optimization model. Taking the minimum total energy consumption of the air-conditioning system as the objective function, define control variables and set constraint conditions to complete the modeling of the energy efficiency optimization model. Among them, linearize the non-linear parts in the constraint conditions and the objective function.

[0105] In a specific embodiment, the expression of the energy efficiency optimization model is:

[0106]

[0107] Among them, T represents the time period; t represents the moment.

[0108] In some embodiments, the defined control variables include: the operating state of the chiller, the operating state of the water pump, the operating state of the cooling tower fan, the operating state of the fan coil, the operating state of the variable air volume system, the set temperature, the cooling load, the water pump flow rate, the cooling tower fan air volume, the fan coil air volume, and the variable air volume system air volume.

[0109] Among them, the values of the operating state of the chiller, the operating state of the water pump, the operating state of the cooling tower fan, the operating state of the fan coil, and the operating state of the variable air volume system are 0 or 1. When the value is 0, it means not operating, and when the value is 1, it means operating.

[0110] In some embodiments, the set constraint conditions include: equipment operating state constraint, temperature constraint, and robust load constraint.

[0111] Among them, the equipment operating state constraint is:

[0112] Chiller operating state u c,t Constraint:

[0113] u c,t ∈{0,1}, Q c ≤u c,t ·Q max

[0114] Among them, Q max represents the maximum value of the cooling load.

[0115] Water pump operating state u p,t Constraint:

[0116] u p,t ∈{0,1}, Q p ≤u p,t ·Q p,max

[0117] Among them, Q p,max represents the maximum value of the water pump flow rate.

[0118] Operating status of the cooling tower u f,t Constraint:

[0119] u f,t ∈ {0, 1}, Q f ≤ u f,t ·Q f,max

[0120] Where, Q f,max represents the maximum value of the air volume of the cooling tower fan.

[0121] Operating status of the fan coil unit u fc,t Constraint:

[0122] u fc,t ∈ {0, 1}, Q fc ≤ u fc,t ·Q fc,max

[0123] Where, Q fc,max represents the maximum value of the air volume of the fan coil unit fan.

[0124] Operating status of the variable air volume system u vav,t Constraint:

[0125] u vav,t ∈ {0, 1}, Q vav ≤ u vev,t ·Q vav,max

[0126] Where, Q vav,max represents the maximum value of the air volume of the variable air volume system fan.

[0127] The temperature constraint is:

[0128] T min ≤ T set,t ≤ T max

[0129] Where, T set,t represents the set temperature at time t; T min represents the minimum settable temperature; T max represents the maximum settable temperature.

[0130] The robust load constraint includes:

[0131]

[0132] Ensure that the cooling load meets the air-conditioning load prediction results output by the load prediction model within the allowable deviation through the robust load constraint;

[0133] Among them, δ represents the maximum allowable deviation value. Based on historical load data records, according to the requirements of prediction accuracy (such as daily / hourly / minutely), the fluctuation range of the load at different time scales can be statistically analyzed to determine the deviation value.

[0134] In some embodiments, in the energy efficiency optimization model, considering the uncertainty Δx of the input features and the uncertainty Δθ of the model parameters, robust optimization is performed by maximizing the objective function in the worst-case scenario;

[0135] Introducing the uncertainty of the input features includes:

[0136] x′ i =x i +Δx i , Δx i =||x i -x i0 ||∈Ω

[0137] Introducing the uncertainty of the model parameters includes:

[0138] Δθ = ||θ - θ0||∈Θ

[0139] Among them, x i0 is the original input feature, Ω represents the set of uncertainties of the input features; θ0 is the initial parameter of the model, and Θ represents the set of uncertainties of the model parameters.

[0140] Furthermore, verify the effectiveness and stability of the robust optimization scheme in practical applications.

[0141] In the embodiments of the present disclosure, tools such as Gurobi or the PuLP library of Python are used to solve the optimization model.

[0142] The method of the present disclosure can be used in the real-time operation of an air conditioning system. According to the load predicted by the SVR model and the results of the optimization model, the operating states and parameters of devices such as chillers, pumps, and cooling towers are adjusted in real time to achieve energy-saving operation of the system; by comparing the energy consumption data before and after optimization, the energy-saving effect is evaluated, and further energy-saving strategies are formulated.

[0143] In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0144] Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An energy efficiency optimization method for an air conditioning system based on load prediction, characterized in that, Including: Collecting air-conditioning load-related data, where the air-conditioning load-related data includes outdoor meteorological data, indoor environmental parameters, and historical air-conditioning load data; Constructing a load prediction model, where the load prediction model uses a support vector regression model, and uses the air-conditioning load-related data to train the support vector regression model; in the support vector regression model, a hybrid kernel function of a linear kernel and a Gaussian kernel is used to construct a model function, where the linear kernel is used to capture the air-conditioning load trend, and the Gaussian kernel is used to capture local fluctuations; obtaining an air-conditioning load prediction result according to the load prediction model; Constructing an energy efficiency optimization model, where the energy efficiency optimization model uses a mixed-integer linear programming model, and the mixed-integer linear programming model takes the minimum total energy consumption of the air-conditioning system as the objective function, considers the energy consumption of various devices in the air-conditioning system, and establishes an energy consumption function; the energy consumption function represents the energy consumption under given control variables and the air-conditioning load prediction result.

2. The energy efficiency optimization method for an air conditioning system based on load prediction according to claim 1, wherein The model function of the support vector regression model is: Among them, N represents the number of samples; i represents the sample number; α i represents the Lagrange multiplier; represents the dual form of the Lagrange multiplier; x represents the input data; x i represents the sample of the input training data; b represents the bias term; K mix (x, x i ) represents the mixed kernel function; The expression of the hybrid kernel function is as follows: K mix (x, x i ) = αK lin (x, x i ) + (1 - α)K rbf (x, x i ) K lin (x,x i )=x T x i K rbf (x,x i ) = exp(-γ ∥x - x i ∥ 2 ) Among them, K lin (x, x i ) represents a linear kernel function; K rbf (x, x i ) represents a Gaussian kernel function; α represents the weight coefficient of the mixed kernel function, which is used to balance the contributions of the linear kernel function and the Gaussian kernel function; γ represents the width coefficient of the Gaussian function.

3. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 2, characterized in that The selection of the weight coefficient of the hybrid kernel function is related to the humidity parameter in the historical air-conditioning load data. The higher the humidity control requirement, the smaller the value of the weight coefficient of the hybrid kernel function.

4. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 3, characterized in that In the load prediction model, the defined input data includes outdoor meteorological data, indoor environmental parameters, and air-conditioning equipment power in the air-conditioning load-related data, and the defined output data includes air-conditioning load data in the air-conditioning load-related data; The method for training the load prediction model includes: Dividing the air-conditioning load-related data into a training set and a test set according to a set ratio; Training the load prediction model based on the training set data, where a hyperparameter optimization method is used to traverse all parameter combinations in combination with cross-validation to find the best parameter combination; After training is completed, use the test set data to evaluate the model, and adjust the feature selection and parameters of the load prediction model according to the evaluation results.

5. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 1, wherein The constructing of the energy efficiency optimization model includes: Constructing energy consumption models for various devices in the air-conditioning system, where the various devices in the air-conditioning system at least include a chiller, a circulating water pump, a cooling tower, and a terminal air handling unit; Integrating the energy consumption models of various devices in the air-conditioning system to obtain the total energy consumption function of the air-conditioning system; Integrating the air-conditioning load prediction result output by the load prediction model into the energy efficiency optimization model, taking the minimum total energy consumption of the air-conditioning system as the objective function, defining control variables and setting constraint conditions to complete the modeling of the energy efficiency optimization model; where the nonlinear parts in the constraint conditions and the objective function are linearized.

6. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 5, characterized in that The total energy consumption function is: Among them, represents the total energy consumption function, where represents the predicted result of the air-conditioning load output by the load prediction model; u t represents the control variable vector; E chiller (Q c , T cw , T ww ) represents the energy consumption model of the chiller, Q c represents the refrigeration load, T cw represents the cooling water temperature, T ww represents the chilled water temperature; E pump (Q p , H p ) represents the energy consumption model of the circulating water pump, Q p represents the pump flow rate, H p represents the pump head; E coolingtower (Q f ) represents the energy consumption model of the cooling tower, Q f represents the cooling tower fan air volume; E FCU (Q fc , P fc ) represents the energy consumption model of the fan coil unit, Q fc represents the fan coil unit air volume; P fc represents the pressure difference of the fan coil unit fan; E VAV (Q vav , P vav ) represents the energy consumption model of the variable air volume system, Q vav represents the variable air volume system air volume; P vav represents the pressure difference of the variable air volume system fan.

7. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 6, characterized in that The expression of the energy efficiency optimization model is: Where, T represents the time period; t represents the moment.

8. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 7, wherein, The control variables include: The operating state of the chiller, the operating state of the water pump, the operating state of the cooling tower fan, the operating state of the fan coil, the operating state of the variable air volume system, the set temperature, the cooling load, the water pump flow rate, the cooling tower fan air volume, the fan coil air volume, and the variable air volume system air volume.

9. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 7, wherein The constraint conditions include: equipment operating state constraints, temperature constraints, and robust load constraints; Among them, the robust load constraints include: Ensure that the refrigeration load is within the allowable deviation range to meet the air-conditioning load prediction results output by the load prediction model through robust load constraints; Among them, δ represents the maximum allowable deviation value, which is based on historical load data records.

10. The energy efficiency optimization method of the air conditioning system based on load prediction according to claim 7, wherein In the energy efficiency optimization model, considering the uncertainty Δx of the input features and the uncertainty Δθ of the model parameters, robust optimization is carried out by maximizing the objective function in the worst case; The introduced uncertainty of the input features includes: x′ i = x i + Δx i , Δx i = ||x i - x i0 || ∈ Ω The introduced uncertainty of the model parameters includes: Among them, x i0 is the original input feature, and Ω represents the uncertainty set of the input feature; θ0 is the initial parameter of the model, and Θ represents the uncertainty set of the model parameters.

Images