Operation control method for evaporative condensing unit

Abstract

According to the operation control method for the evaporation condensing unit, an LSTM-STL bimodal prediction model and a dynamic constraint MPC algorithm are creatively fused, and precise matching of cold energy supply and demand and energy consumption collaborative optimization are achieved; the method comprises the following steps: firstly, decomposing complex time sequence characteristics of effective decoupling cooling capacity requirements by using STL, and combining the strong fitting capability of LSTM to a nonlinear trend and accurate modeling of Fourier series to periodicity, so that the prediction precision is remarkably improved; secondly, a cooling capacity margin self-adaptive adjustment mechanism is introduced into the MPC framework, the safety boundary is dynamically relaxed according to the predicted error standard deviation, frequent starting and stopping of a compressor are avoided, and temperature fluctuation of a machine room is reduced; in addition, the prediction model is corrected in real time through Kalman filtering, and control drift caused by environmental disturbance is effectively restrained. According to the method, the refrigeration energy consumption is reduced by 24.7% compared with a traditional strategy while the cooling reliability is guaranteed, and the method can be expanded and applied to various dynamic load scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of data center refrigeration, and in particular to an operation control method for an evaporative condensing unit. Background Art

[0002] With the rapid development of cloud computing and artificial intelligence technologies, data centers continue to expand in size, and their energy consumption is becoming increasingly prominent. Refrigeration systems are the second-largest energy consumer in data centers (accounting for approximately 30-40%). Evaporative condensing units, particularly those suitable for dry climates, are widely used to cool large data centers. However, data center cooling demand is significantly nonlinear and time-varying, influenced by factors such as real-time computing load and ambient temperature and humidity, posing significant challenges to traditional control strategies.

[0003] Existing control methods for evaporative condensing units primarily rely on fixed-threshold start-stop control or PID feedback regulation. For example, some solutions linearly adjust the compressor frequency by monitoring the deviation of return water temperature from a setpoint; others employ empirical rule libraries to match preset operating points. However, these methods suffer from three major drawbacks: First, they suffer from significant response lag. When IT load suddenly changes, the imbalance between cooling supply and demand can cause room temperature fluctuations exceeding ±2°C (research shows that every 1°C temperature fluctuation increases server failure rates by 4%). Second, they suffer from insufficient energy efficiency optimization, leading to low unit efficiency at partial load, with measured COP values ​​15-25% lower than the theoretical optimal value. Third, multi-unit coordination lacks global optimization, resulting in independent operation of each unit, leading to a "cold rush" phenomenon and exacerbating energy waste. Particularly noteworthy is that existing predictive control methods often employ a single ARIMA model, failing to effectively decouple the trend, cyclical, and random components of cooling demand, resulting in prediction errors generally exceeding 20%.

[0004] Furthermore, while patent CN107726683A proposes energy-saving control based on evaporation and condensation parameter matching, it relies on a static matching mechanism based on a historical database, making it incapable of adapting to the dynamic load characteristics of data centers and failing to address the issue of forward-looking forecasting of cooling demand. Industry reports indicate that global data center electricity consumption has exceeded 200 billion kWh annually, with ineffective cooling accounting for 18% due to flawed control strategies.

[0005] Therefore, developing an intelligent control method that integrates high-precision cooling demand prediction and dynamic optimization to achieve efficient and stable operation of evaporative condensing units has become a core technical problem that needs to be solved urgently. Summary of the Invention

[0006] The purpose of the present invention is to provide an operation control method for an evaporative condensing unit to solve the problems existing in the above-mentioned prior art.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] The present invention provides an operation control method for an evaporative condensing unit, comprising the following steps:

[0009] S1. Collect historical operation data of the data center, including cooling demand time series data Q req (t), ambient temperature T amb (t), ambient humidity H amb (t), evaporative condensing unit compressor frequency f c (t), evaporative condensing unit fan speed v f (t), evaporative condensing unit valve opening α v (t) and load factor L IT (t);

[0010] S2. Preprocess the historical data, fill missing values ​​using Lagrange interpolation and perform Z-score standardization;

[0011] S3. Build a cooling demand forecasting model and decompose the original cooling time series into trend items T through STL decomposition. trend 、Seasonal Item S season and the residual term R resid , respectively using LSTM network to predict trend terms, Fourier series to fit seasonal terms, SVR model to predict residual terms, and synthesize the predicted value Q req_pred (t+Δt);

[0012] S4. Establish the cooling output equation of the unit:

[0013] Q out (t) = η·[k1f c (t)+k2v f (t) β +k3α v (t)·ΔT evap ];

[0014] Among them, η is the system efficiency coefficient, k1, k2, k3 are equipment characteristic parameters, β is the fan aerodynamic index, ΔT evap is the evaporator temperature difference;

[0015] S5. Design a model predictive controller with the optimization goals of minimizing cooling tracking error and minimizing energy consumption, and solve the objective function:

[0016]

[0017] Among them, N p is the prediction time domain, w1, w2, w3 are weight coefficients, is the total power consumption model;

[0018] S6. Dynamically adjust the cooling capacity safety margin according to the prediction error standard deviation σ error Setting δQ tol , when σ error When <5%, take 0.1Q req_pred Otherwise, take 0.15Q req_pred ;

[0019] S7. Implement feedback correction and obtain the actual cooling capacity Q every 5 minutes real , calculate the prediction deviation ∈(t) and update the LSTM hidden layer state through Kalman filtering.

[0020] Preferably, in step S1, the ambient temperature and the ambient humidity data are collected through a distributed sensor network.

[0021] Preferably, in step S2, the formula of the Lagrange interpolation method is:

[0022]

[0023] The interpolation window n is adaptively adjusted according to the data missing rate.

[0024] Preferably, in step S3, the formula of the cooling demand prediction model is:

[0025]

[0026] Among them, P is the period, a k is the Fourier coefficient, and Δt is the prediction step size.

[0027] Preferably, in step S4, the equipment characteristic parameters are measured by variable frequency testing, and the fan aerodynamic index is measured by wind tunnel testing.

[0028] Preferably, in step S5, the prediction time domain is 4, and the weight coefficients are 0.6, 0.3, and 0.1 respectively.

[0029] Preferably, in step S7, the update formula of the Kalman filter is:

[0030] x t|t =x t|t1+ K t (∈(t)-Hx t|t-1 );

[0031] Among them, K t is the Kalman gain matrix, and H is the observation matrix.

[0032] Preferably, it also includes:

[0033] S8. Multi-unit coordinated control, solving the load distribution optimization problem using the interior point method:

[0034]

[0035] where a i , b i is the unit efficiency curve coefficient.

[0036] Compared with the prior art, the present invention has achieved the following beneficial technical effects:

[0037] The present invention provides an operation control method for an evaporative condensing unit. This method achieves precise matching of cooling supply and demand and coordinated optimization of energy consumption through the innovative integration of an LSTM-STL dual-modal prediction model and a dynamically constrained MPC algorithm. First, STL decomposition is used to effectively decouple the complex time series characteristics of cooling demand. This, combined with LSTM's strong ability to fit nonlinear trends and Fourier series' precise modeling of periodicity, significantly improves prediction accuracy. Secondly, an adaptive cooling margin adjustment mechanism is introduced into the MPC framework, dynamically relaxing the safety margin based on the standard deviation of the prediction error, thereby avoiding frequent compressor starts and stops and reducing room temperature fluctuations. Furthermore, the prediction model is corrected in real time through Kalman filtering, effectively suppressing control drift caused by environmental disturbances. This method reduces cooling energy consumption by 24.7% compared to traditional strategies while ensuring cooling reliability, and can be extended to various dynamic load scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 This is a flow chart of the operation control method of the evaporative condensing unit provided by the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] The purpose of the present invention is to provide an operation control method for an evaporative condensing unit to solve the problems existing in the prior art.

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] Example 1:

[0044] This embodiment provides an evaporative condensing unit operation control method, which is implemented based on the evaporative condensing unit operation control system. The evaporative condensing unit operation control system deploys the core algorithm in the edge computing module and communicates with the evaporative condensing unit PLC through the Modbus / TCP protocol. Figure 1 As shown, the method includes the following steps:

[0045] S1. Collect historical operation data of the data center, including cooling demand time series data Q req (t), ambient temperature T amb (t), ambient humidity H amb (t), evaporative condensing unit compressor frequency f c (t), evaporative condensing unit fan speed v f (t), evaporative condensing unit valve opening α v (t) and load factor L IT (t); wherein the ambient temperature and the ambient humidity data are collected through a distributed sensor network, and the compressor frequency f of the evaporative condensing unit c (t), evaporative condensing unit fan speed v f (t), evaporative condensing unit valve opening α v (t) and load factor L IT (t) Real-time reporting by the evaporative condensing unit;

[0046] S2. Preprocess the historical data, use the Lagrange interpolation method to fill missing values ​​and perform Z-score standardization; the formula of the Lagrange interpolation method is:

[0047]

[0048] The interpolation window n is adaptively adjusted according to the data missing rate;

[0049] S3. Build a cooling demand forecasting model and decompose the original cooling time series into trend items T through STL decomposition. trend 、Seasonal Item S season and the residual term R resid , respectively using LSTM network to predict trend terms, Fourier series to fit seasonal terms, SVR model to predict residual terms, and synthesize the predicted value Q req_pred (t+Δt);

[0050] The formula of the cooling demand prediction model is:

[0051]

[0052] Among them, P is the period, a k is the Fourier coefficient, Δt is the prediction step length;

[0053] S4. Establish the cooling output equation of the unit:

[0054] Q out (t) = η·[k1f c (t)+k2v f (t) β +k3α v (t)·ΔT evap ];

[0055] Among them, η is the system efficiency coefficient, k1, k2, k3 are equipment characteristic parameters, β is the fan aerodynamic index, ΔT evap is the evaporator temperature difference;

[0056] The equipment characteristic parameters are measured by variable frequency testing, and the fan aerodynamic index is measured by wind tunnel testing;

[0057] S5. Design a model predictive controller with the optimization goals of minimizing cooling tracking error and minimizing energy consumption, and solve the objective function:

[0058]

[0059] Among them, N p is the prediction time domain, w1, w2, w3 are weight coefficients, is the total power consumption model;

[0060] The prediction time domain is 4, and the weight coefficients are 0.6, 0.3, and 0.1 respectively;

[0061] S6. Dynamically adjust the cooling capacity safety margin according to the standard deviation of the prediction error σ error Setting δQ tol , when σ error When <5%, take 0.1Q req_pred Otherwise, take 0.15Q req_pred ;

[0062] S7. Implement feedback correction and obtain the actual cooling capacity Q every 5 minutes real , calculate the prediction deviation ∈(t) and update the LSTM hidden layer state through Kalman filtering;

[0063] The update formula of the Kalman filter is:

[0064] x t|t =x t|t-1 +Kt (∈(t)-Hx t|t-1 );

[0065] Among them, K t is the Kalman gain matrix, H is the observation matrix;

[0066] S8. Multi-unit coordinated control, solving the load distribution optimization problem using the interior point method:

[0067]

[0068] where a i , b i is the unit efficiency curve coefficient.

[0069] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A method for controlling the operation of an evaporative condensing unit, characterized in that: The following steps are involved: S1. Collect historical operation data of the data center, including cooling demand time series data Q req (t), ambient temperature T amb (t), ambient humidity H amb (t), evaporative condensing unit compressor frequency f c (t), evaporative condensing unit fan speed v f (t), evaporative condensing unit valve opening α v (t) and load factor L IT (t); S2. Preprocess the historical data, fill missing values ​​using Lagrange interpolation and perform Z-score standardization; S3. Build a cooling demand forecasting model and decompose the original cooling time series into trend items T through STL decomposition. trend 、Seasonal Item S season and the residual term R resid , respectively using LSTM network to predict trend terms, Fourier series to fit seasonal terms, SVR model to predict residual terms, and synthesize the predicted value Q req_pred (t+Δt); S4. Establish the cooling output equation of the unit: Q out (t)=η·[k1f c (t)+k2v f (t) β +k3α v (t)·ΔT evap ]; Among them, η is the system efficiency coefficient, k1, k2, k3 are equipment characteristic parameters, β is the fan aerodynamic index, ΔT evap is the evaporator temperature difference; S5. Design a model predictive controller with the optimization goals of minimizing cooling tracking error and minimizing energy consumption, and solve the objective function: Among them, N p is the prediction time domain, w1, w2, w3 are weight coefficients, is the total power consumption model; S6. Dynamically adjust the cooling capacity safety margin according to the prediction error standard deviation σ error Setting δQ tol , when σ error When <5%, take 0.1Q req_pred Otherwise, take 0.15Q req_pred ; S7. Implement feedback correction and obtain the actual cooling capacity Q every 5 minutes real , calculate the prediction deviation ∈(t) and update the LSTM hidden layer state through Kalman filtering.

2. The evaporative condensing unit operation control method according to claim 1, characterized in that: In step S1, the ambient temperature and the ambient humidity data are collected through a distributed sensor network.

3. The evaporative condensing unit operation control method according to claim 1, characterized in that: In step S2, the formula of the Lagrange interpolation method is: The interpolation window n is adaptively adjusted according to the data missing rate.

4. The evaporative condensing unit operation control method according to claim 1, characterized in that: In step S3, the formula of the cooling demand prediction model is: Among them, P is the period, a k is the Fourier coefficient, and Δt is the prediction step size.

5. The evaporative condensing unit operation control method according to claim 1, characterized in that: In step S4, the equipment characteristic parameters are measured through variable frequency testing, and the fan aerodynamic index is measured through wind tunnel testing.

6. The evaporative condensing unit operation control method according to claim 1, characterized in that: In step S5, the prediction time domain is 4, and the weight coefficients are 0.6, 0.3, and 0.1 respectively.

7. The evaporative condensing unit operation control method according to claim 1, characterized in that: In step S7, the update formula of the Kalman filter is: x t|t =x t|t-1 +K t (∈(t)-Hx t|t-1 ); Among them, K t is the Kalman gain matrix, and H is the observation matrix.

8. The evaporative condensing unit operation control method according to claim 1, characterized in that: Also includes: S8. Multi-unit coordinated control, solving the load distribution optimization problem using the interior point method: where a i , b i is the unit efficiency curve coefficient.

Images