Energy efficiency monitoring method and system for real-time operation of central air conditioner host system

Abstract

The invention provides an energy efficiency monitoring method and system for real-time operation of a central air conditioner host system, and the method comprises the steps: collecting operation data of an air conditioner system, and fusing the operation data according to a preset data weight to obtain fused data; and performing time alignment and space mapping on the fused data to obtain standardized data, and performing dimension feature extraction on the standardized data to obtain dimension feature data. And performing spatial feature extraction on the dimension feature data through a preset convolutional neural network to obtain spatial feature data. And performing time sequence feature extraction on the spatial feature data through a preset long-short-term memory network to obtain an initial load prediction value. And correcting the initial load prediction value according to a preset time sequence change regularization item to obtain a corrected load prediction value. And the accuracy of predicting the load of the air conditioning system is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of air conditioning load prediction, and particularly relates to an energy efficiency monitoring method and system for real-time operation of a central air conditioning host system. BACKGROUND

[0002] Energy consumption in the field of buildings has attracted widespread attention, especially in the energy efficiency management of air conditioning systems. As one of the largest energy consumption devices in buildings, the energy efficiency management of air conditioning systems plays a crucial role in energy saving and emission reduction. In related technologies, the energy efficiency monitoring method of air conditioning systems mainly relies on simple statistical models or empirical formulas to predict air conditioning load. This leads to inaccurate prediction and makes it impossible to accurately optimize energy efficiency and intelligently adjust the air conditioning system according to the predicted load. Therefore, how to accurately predict the load of the air conditioning system has become a technical problem to be solved. SUMMARY

[0003] The purpose of the present application is to design an energy efficiency monitoring method and system for real-time operation of a central air conditioning host system, which can accurately predict the load of the air conditioning system.

[0004] To achieve the above purpose, in a first aspect of the present application, an energy efficiency monitoring method for real-time operation of a central air conditioning host system is provided, which comprises:

[0005] Collecting operation data of the air conditioning system, fusing the operation data according to a preset data weight to obtain fused data;

[0006] Performing time alignment and space mapping on the fused data to obtain standardized data, performing dimension feature extraction on the standardized data to obtain dimension feature data;

[0007] Performing space feature extraction on the dimension feature data through a preset convolutional neural network to obtain space feature data;

[0008] Performing time sequence feature extraction on the space feature data through a preset long short-term memory network to obtain an initial load prediction value;

[0009] Correcting the initial load prediction value according to a preset time sequence change regularization term to obtain a corrected load prediction value.

[0010] As a preferred embodiment, after the initial load prediction value is corrected according to the preset time sequence change regularization term to obtain the corrected load prediction value, the method further comprises:

[0011] Obtaining a user demand vector;

[0012] correcting the corrected load prediction value according to the user demand vector and a preset weighting factor, to obtain a corrected load prediction value; wherein the weighting factor is used to control the fusion degree of the corrected load prediction value and the user demand vector.

[0013] As preferred, after the step of correcting the corrected load prediction value according to the user demand vector and a preset weighting factor, to obtain a corrected load prediction value, the method further comprises:

[0014] obtaining a microclimate feature at a previous time and a microclimate feature at a current time;

[0015] constructing an adjustment factor according to the corrected load prediction value at the previous time and the microclimate feature at the current time; wherein the adjustment factor represents a nonlinear relationship between microclimate and load adjustment;

[0016] adjusting the corrected load prediction value according to the adjustment factor and a preset adjustment factor weight, to obtain a target load prediction value.

[0017] As preferred, the step of adjusting the corrected load prediction value according to the adjustment factor and a preset adjustment factor weight, to obtain a target load prediction value, comprises:

[0018] obtaining a microclimate feature at a previous time;

[0019] constructing a smooth load regularization term according to the microclimate feature at the previous time and the microclimate feature at the current time;

[0020] adjusting the corrected load prediction value according to the adjustment factor, a preset adjustment factor weight and the smooth load regularization term, to obtain a target load prediction value.

[0021] As preferred, the user demand vector comprises a plurality of user demands, and after the step of adjusting the corrected load prediction value according to the adjustment factor and a preset adjustment factor weight, to obtain a target load prediction value, the method further comprises:

[0022] weighting and synthesizing a plurality of the user demands, to obtain a comprehensive demand index;

[0023] weighting and fusing the comprehensive demand index and the target load prediction value according to a preset proportional weighting coefficient, to obtain a personalized load prediction value.

[0024] As preferred, after the step of weighting and fusing the comprehensive demand index and the target load prediction value according to a preset proportional weighting coefficient, to obtain a personalized load prediction value, the method further comprises:

[0025] obtaining an actual load demand;

[0026] According to the actual load demand and the personalized load prediction value, a difference calculation is performed to obtain a prediction error;

[0027] According to the prediction error and a preset learning rate, the proportional weighting coefficient is adjusted to obtain an adjusted proportional weighting coefficient.

[0028] Preferably, the adjusting the proportional weighting coefficient according to the prediction error and the preset learning rate to obtain the adjusted proportional weighting coefficient comprises:

[0029] An upper time point proportional weighting coefficient is obtained;

[0030] A stable regularization term is constructed according to the upper time point proportional weighting coefficient and the current time point proportional weighting coefficient;

[0031] According to the prediction error, the preset learning rate and the stable regularization term, the proportional weighting coefficient is adjusted to obtain the adjusted proportional weighting coefficient.

[0032] In a second aspect of the present application, an energy efficiency monitoring system for real-time operation of a central air conditioner host system is provided, and the system comprises:

[0033] A collection unit is configured to collect operation data of the air conditioner system, and fuse the operation data according to a preset data weight to obtain fused data;

[0034] A standardization unit is configured to perform time alignment and space mapping on the fused data to obtain standardized data, and perform dimension feature extraction on the standardized data to obtain dimension feature data;

[0035] A space extraction unit is configured to perform space feature extraction on the dimension feature data through a preset convolutional neural network to obtain space feature data;

[0036] A time sequence extraction unit is configured to perform time sequence feature extraction on the space feature data through a preset long short-term memory network to obtain an initial load prediction value;

[0037] A correction unit is configured to correct the initial load prediction value according to a preset time sequence change regularization term to obtain a corrected load prediction value.

[0038] The present application has at least the following beneficial technical effects:

[0039] In view of the above problems, the present application provides an energy efficiency monitoring method and system for real-time operation of a central air conditioner host system, the core of which is to effectively solve the problems existing in the energy efficiency monitoring and optimization method of the existing air conditioner system by means of multi-dimensional data fusion, microclimate modeling, individualized demand regulation and adaptive learning, etc., so as to realize accurate load prediction of the air conditioner system. The energy efficiency and user comfort of the system are significantly improved, and energy waste is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0040] The present application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present application. Other drawings can be obtained by those of ordinary skill in the art without creative labor on the basis of the following drawings.

[0041] Figure 1 is a flowchart of the energy efficiency monitoring method for real-time operation of a central air conditioner host system provided by the present application.

[0042] Figure 2 is a flowchart of the energy efficiency monitoring method for real-time operation of a central air conditioner host system provided by another embodiment of the present application.

[0043] Figure 3 is a flowchart of the energy efficiency monitoring method for real-time operation of a central air conditioner host system provided by a third embodiment of the present application.

[0044] Figure 4 is a flowchart of step S303 in Figure 3 .

[0045] Figure 5 is a flowchart of the energy efficiency monitoring method for real-time operation of a central air conditioner host system provided by a fourth embodiment of the present application.

[0046] Figure 6 is a flowchart of the energy efficiency monitoring method for real-time operation of a central air conditioner host system provided by a fifth embodiment of the present application.

[0047] Figure 7 is a flowchart of step S603 in Figure 6 .

[0048] Figure 8 is a structural schematic diagram of the energy efficiency monitoring system for real-time operation of a central air conditioner host system provided by the present application. DETAILED DESCRIPTION

[0049] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0050] In one or more embodiments, please refer to Figure 1 , Figure 1 This is a flow chart of a method for real-time energy efficiency monitoring of a central air-conditioning host system provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S105.

[0051] Step S101: collecting operating data of the air-conditioning system, and fusing the operating data according to preset data weights to obtain fused data;

[0052] Step S102: performing time alignment and spatial mapping on the fused data to obtain standardized data, and performing dimensional feature extraction on the standardized data to obtain dimensional feature data;

[0053] Step S103, performing spatial feature extraction on the dimensional feature data using a preset convolutional neural network to obtain spatial feature data;

[0054] Step S104, extracting time series features from the spatial feature data using a preset long short-term memory network to obtain an initial load forecast value;

[0055] Step S105 , correcting the initial load forecast value according to a preset time series variation regularization term to obtain a corrected load forecast value.

[0056] In step S101 of some embodiments, the operating data of the air-conditioning system is collected in real time through sensors and monitoring equipment, including but not limited to indoor and outdoor temperature, humidity, air quality (such as PM2.5 concentration), CO2 concentration, activity status of people in the area, user preferences, etc. In order to process these multi-source data, sensor data denoising and outlier detection are used. The specific denoising method adopts a filtering algorithm based on adaptive noise estimation, which dynamically adjusts the filtering parameters according to the sampling frequency, change rate and environmental characteristics of each data type to ensure the accuracy of the collected data. It can effectively remove errors caused by sensor failure or external interference and improve the reliability of the data. As shown in the following formula (1):

[0057] (1);

[0058] in, Indicates the Operation data collected at all times (such as temperature, humidity, CO2 concentration, etc.). is the real data obtained by adaptive filtering. is the noise term, which depends on the estimate of the system noise and dynamically adjusts its size.

[0059] Next, the collected various types of operation data are integrated to form a comprehensive data set using multi-dimensional data fusion technology. To ensure the information integrity in the data fusion process, a weighted collaborative fusion algorithm is used, which assigns data weights to each data type based on the reliability, accuracy, and real-time performance of different data sources. The data weights are dynamically adjusted based on the relevance and stability of historical data to further ensure the accuracy of the fused data.

[0060] In an example, assuming there are types of data sources, denoted as , the operation data of each data source at time is (where ). The data weight of each data source is , and the update of the data weight is based on error analysis in the past period of time, optimized through adaptive learning method. The final weighted fusion formula is shown in the following formula (2):

[0061] (2);

[0062] wherein is the fused data after weighted fusion. is the data weight of the th data source at time . is the operation data of the th data source at time . The weighted collaborative fusion algorithm can effectively eliminate the deviation of certain data sources by dynamically adjusting the weights of different data sources, ensuring the accuracy of the final fused data.

[0063] In step S102 of some embodiments, due to the different time scales and spatial distribution characteristics of various types of operation data. To match these data to a unified time-space framework, time alignment and spatial mapping are performed on different data sources to ensure that various types of data can be synchronized and scale differences are eliminated under a unified time scale.

[0064] Specifically, for the time dimension, a time interpolation algorithm (such as spline interpolation) is used to uniformly process irregularly time-sampled data. For the spatial dimension, a spatial position-based weighting algorithm is used to standardize environmental data in different regions to a unified regional scale. The specific standardization formula is shown in the following formula (3):

[0065] (3);

[0066] wherein, is the standardized data. is the fused data at time . is the mean of . is the standard deviation of . This standardization method can eliminate the scale difference between different data sources, ensuring that the subsequent modeling process can be carried out under uniform standards.

[0067] Further, through the multi-dimensional feature extraction method, the key features related to air conditioner energy efficiency are extracted from the fused data, and the dimensional feature data is obtained. For example, indoor temperature change rate, user activity, air conditioner load fluctuation, etc. These features will be used as the basis for input in the subsequent steps of load prediction and personalized adjustment. Feature selection method based on information gain is used in feature extraction to remove redundant information and retain the most helpful features for energy efficiency prediction. The simplified representation of the dimensional feature extraction process is shown in the following formula (4):

[0068] (4);

[0069] wherein, is the dimensional feature data extracted at time . is the standardized data. represents the feature selection method calculated by information gain, which returns the most relevant features for air conditioner energy efficiency prediction. Through this method, key features can be efficiently extracted from multi-dimensional data, reducing data dimensionality and improving the computational efficiency and prediction accuracy of subsequent models.

[0070] The goal of this step is to provide accurate input data for subsequent air conditioner load prediction and optimization through multi-dimensional data fusion methods. The multi-dimensional data acquisition, weighted collaborative fusion, spatio-temporal standardization and feature extraction methods designed in this step combine the data characteristics of air conditioning systems and provide a solution with high accuracy and operability. Through these technical means, data can be effectively processed and utilized in complex and variable air conditioning operating environments, thereby laying a solid foundation for subsequent modeling, prediction and adjustment.

[0071] In some embodiments, in steps S103 to S104, the dimensional feature data set includes multiple dimensional feature data, which contains environmental parameters, air conditioner operating status, user behavior and other information. Based on these data, a composite model is constructed to predict air conditioner load.

[0072] Specifically, since the air conditioning load is affected by the spatial correlation of multiple factors such as environmental changes and sensor status, spatial features are first extracted from the dimensional feature data. Convolutional neural networks (CNNs) are used to capture these spatial dependencies. Specifically, two-dimensional convolution operations are performed on Processing is performed to extract spatial patterns related to air conditioning load and obtain spatial feature data , contains the characteristics of the spatial distribution of the environment and air conditioning system at each moment. Assume It is an input data matrix containing spatial information. The CNN model can extract corresponding features in different regions and provide more accurate spatial information for subsequent load forecasting.

[0073] Acquiring spatial feature data Finally, we need to consider the time dependency of the load. We use LSTM (Long Short-Term Memory) to capture the long-term dependency of time series. LSTM can remember the historical load pattern and predict the future load demand over time. Assume that at time , the time series of air conditioning load is ,Will As the input of LSTM, the time series dynamics of air conditioning load is learned through LSTM network. The output of LSTM model is the initial load forecast value .

[0074] In step S105 of some embodiments, since the air conditioning load prediction is not only dependent on environmental factors, but also affected by sudden events (such as weather changes, etc.), it is necessary to introduce a regularization term to reduce overfitting and better adapt to time series fluctuations. Therefore, the time series change regularization term is used to suppress excessive fluctuations between prediction values, so that the model can maintain a stable prediction effect when facing drastic changes in the environment. That is, according to the time series change regularization term as part of the loss function, the difference between adjacent prediction values ​​is constrained during the model training process, so that the final prediction result is more stable when facing sudden changes in the environment, and the initial load prediction value is corrected to obtain a corrected load prediction value. The temporal variation regularization term is shown in the following formula (5):

[0075] (5);

[0076] in, is the time series change regularization term, which constrains the smoothness of the model output through the training process. For the moment The initial load forecast value, For the moment The initial load forecast value. is the length of the training cycle, is a regularization coefficient, used to adjust the adaptability of the model to the timing fluctuations.

[0077] The steps S101 to S105 shown in the embodiments of the present application, by collecting the operation data of the air conditioning system, fusing the operation data according to the preset data weight, obtaining the fused data. The fused data is time-aligned and space-mapped to obtain standardized data, and the dimension feature of the standardized data is extracted to obtain the dimension feature data. The dimension feature data is space feature extracted by the preset convolutional neural network to obtain the space feature data. The space feature data is time series feature extracted by the preset long short-term memory network to obtain the initial load prediction value. The initial load prediction value is corrected according to the preset timing change regularization term to obtain the corrected load prediction value. The accuracy of the air conditioning system load prediction is improved.

[0078] Please refer to Figure 2 In some embodiments, after step S105, the energy efficiency monitoring method for the real-time operation of the central air conditioning host system can further include but is not limited to steps S201 to S202:

[0079] Step S201, obtaining a user demand vector;

[0080] Step S202, correcting the corrected load prediction value according to the user demand vector and the preset weighting factor to obtain the corrected load prediction value; wherein the weighting factor is used to control the fusion degree of the corrected load prediction value and the user demand vector.

[0081] In step S201 of some embodiments, in the air conditioning load prediction, the individualized demand of the user (such as temperature setting preference, comfort demand, etc.) is an important factor that cannot be ignored. Therefore, the user demand information is introduced into the load prediction to ensure that the prediction result matches the actual demand of the user. The user demand vector is used to represent the individualized preference of the user at time .

[0082] In step S202 of some embodiments, the corrected load prediction value is corrected according to the user demand vector and the preset weighting factor to obtain the corrected load prediction value. As shown in the following formula (6):

[0083] (6);

[0084] Wherein, represents the corrected load prediction value, represents the weighting factor, which is used to control the fusion degree of the corrected load prediction value and the user demand vector. represents the corrected load prediction value, represents the user demand vector.

[0085] By the above steps S201 to S202, a more accurate corrected load prediction value is further obtained , ensuring that the air conditioning load prediction not only reflects the environment and air conditioning system state, but also meets the user's personalized needs.

[0086] Please refer to Figure 3 In some embodiments, after step S202, the energy efficiency monitoring method running in real time towards the central air conditioning host system can further include but is not limited to steps S301 to S303:

[0087] Step S301, obtaining the corrected load prediction value at the previous time and the microclimate characteristics at the current time;

[0088] Step S302, constructing an adjustment factor according to the corrected load prediction value at the previous time and the microclimate characteristics at the current time; wherein the adjustment factor represents the nonlinear relationship between the microclimate and the load adjustment;

[0089] Step S303, adjusting the corrected load prediction value according to the adjustment factor and the preset adjustment factor weight to obtain the target load prediction value.

[0090] In step S301 of some embodiments, the corrected load prediction value takes into account environmental parameters, air conditioning operating conditions, and user needs and other factors. Next, in this step, the corrected load prediction value will be used in combination with microclimate data to fine-tune the air conditioning load, thereby optimizing regional energy efficiency, so as to obtain the corrected load prediction value at the previous time and the microclimate characteristics at the current time.

[0091] In steps S302 to S303 of some embodiments, an adjustment factor is constructed based on the corrected load prediction value at the previous time and the microclimate characteristics at the current time . The adjustment factor describes the nonlinear relationship between the microclimate and the load adjustment. The corrected load prediction value is adjusted according to the adjustment factor and the adjustment factor weight to obtain the target load prediction value. As shown in the following formula (7):

[0092] (7);

[0093] Wherein, represents the target load prediction value, represents the corrected load prediction value, represents the adjustment factor weight, reflecting the influence degree of the microclimate on the load adjustment. represents the adjustment factor, which is the microclimate characteristics at the current time and the corrected load prediction value at the previous time​ The complex nonlinear relationship between them. Machine learning methods (such as neural networks) can be used to model and improve the accuracy of load forecasting.

[0094] Through the above steps S301 to S303, the adjustment factor It can provide personalized load adjustment for each area under changing climate conditions. In combination with microclimate data, it can fine-tune the air conditioning load and further improve the accuracy of air conditioning system load forecasting.

[0095] Please refer to Figure 4 In some embodiments, step S303 may also include but is not limited to steps S401 to S403:

[0096] Step S401, obtaining microclimate characteristics at the previous moment;

[0097] Step S402: constructing a smooth load regularization term based on the microclimate characteristics of the previous moment and the microclimate characteristics of the current moment;

[0098] Step S403 : adjusting the corrected load prediction value according to the adjustment factor, the preset adjustment factor weight, and the smooth load regularization term to obtain a target load prediction value.

[0099] In some embodiments, in steps S401 to S402, in order to prevent over-regulation from causing severe load fluctuations, a regularization term is introduced to smooth the load adjustment. The regularization term is related to the microclimate change rate, that is, a smooth load regularization term is constructed based on the microclimate characteristics of the previous moment and the microclimate characteristics of the current moment. . It can be expressed by the following formula (8):

[0100] (8);

[0101] in, Indicates the current time In the The microclimate characteristics of each sensor location. Indicates the last moment In the The microclimate characteristics of each sensor location. Indicates the time step, usually 1 hour. Indicates the number of sensors in the area. Represents the regularization coefficient, which is used to adjust the smoothness of load adjustment.

[0102] In step S403 of some embodiments, the correction load prediction value is adjusted according to the adjustment factor, the preset adjustment factor weight, and the smooth load regularization term to obtain a target load prediction value, as shown in the following formula (9):

[0103] (9);

[0104] wherein, represents the target load prediction value, represents the correction load prediction value. represents the adjustment factor weight, reflecting the degree of influence of the microclimate on load adjustment. represents the adjustment factor, which is a complex nonlinear relationship between the microclimate characteristics at the current time and the correction load prediction value at the previous time. represents the smooth load regularization term.

[0105] Through the above steps S401 to S403, the smooth load regularization term is designed to calculate the microclimate characteristic change rate, limit the violent fluctuation of load adjustment, ensure the smoothness of the air conditioning system operation, and further improve the accuracy of the air conditioning system load prediction. Not only can it dynamically adjust the air conditioning load under complex microclimate conditions, but also can ensure the stable operation of the system through the smoothing mechanism. It provides a precise energy efficiency optimization scheme for large-scale building or regional air conditioning systems, effectively reduces energy consumption, and improves comfort.

[0106] Please refer to Figure 5 , in some embodiments, the user demand vector includes multiple user demands. After step S303, the energy efficiency monitoring method for the real-time operation of the central air conditioning host system can further include but is not limited to steps S501 to S502:

[0107] Step S501, the multiple user demands are weighted and synthesized to obtain a comprehensive demand index;

[0108] Step S502, the comprehensive demand index and the target load prediction value are weighted and fused according to the preset proportional weight coefficient to obtain a personalized load prediction value.

[0109] In step S501 of some embodiments, the user demand vector represents different demands (such as temperature, humidity, air quality, etc.) of the user at time point . is the th user demand (for example: temperature preference , humidity requirement ). is the dimension of the user demand.

[0110] Specifically, the weighted average method is used to combine multiple user needs Perform weighted synthesis to obtain a comprehensive demand index , as shown in the following formula (10):

[0111] (10);

[0112] in, Indicates that the user is at time The comprehensive demand index. For the The demand weighting coefficient of the user demand is satisfied . For the User requirements.

[0113] In step S502 of some embodiments, the comprehensive demand index Target load forecast value Perform weighted fusion to obtain personalized load forecast value As shown in the following formula (11):

[0114] (11);

[0115] in, Indicates the personalized load forecast value. Represents the proportional weighting coefficient, which controls the integration ratio of the forecast load and user demand. .

[0116] After obtaining the personalized load prediction value through steps S501 to S502, it is used as input to the air conditioning system for dynamic adjustment. The air conditioning system adjusts its operating mode according to the personalized load prediction value to maximize energy efficiency and ensure user comfort.

[0117] Please refer to Figure 6 In some embodiments, after step S502, the energy efficiency monitoring method for the real-time operation of the central air-conditioning host system may further include but is not limited to steps S601 to S603:

[0118] Step S601, obtaining actual load demand;

[0119] Step S602: performing a difference calculation based on the actual load demand and the personalized load forecast value to obtain a forecast error;

[0120] Step S603 , adjusting the proportional weighting coefficient according to the prediction error and the preset learning rate to obtain an adjusted proportional weighting coefficient.

[0121] In steps S601 and S602 of some embodiments, a feedback mechanism is introduced to dynamically adjust the proportionality weighting coefficient . The actual load demand is obtained, and the personalized predicted load value is calculated according to the actual load demand The difference between the two is calculated to obtain the prediction error . As shown in the following formula (12):

[0122] (12);

[0123] wherein, represents the prediction error, represents the actual load demand, represents the personalized predicted load value.

[0124] In step S603 of some embodiments, the proportionality weighting coefficient is adjusted according to the prediction error and the preset learning rate to obtain the adjusted proportionality weighting coefficient. As shown in the following formula (13):

[0125] (13);

[0126] wherein, represents the adjusted proportionality weighting coefficient, represents the proportionality weighting coefficient before adjustment. represents the learning rate, which controls the adjustment range of . .represents the prediction error.

[0127] Through the above steps S601 to S603, the proportionality weighting coefficient is dynamically adjusted according to the feedback mechanism to achieve more personalized and flexible air conditioning load prediction and adjustment.

[0128] Please refer to Figure 7 In some embodiments, step S603 can further include but is not limited to steps S701 to S703:

[0129] Step S701, obtaining the proportionality weighting coefficient at the previous time;

[0130] Step S702, constructing a stable regularization term according to the proportionality weighting coefficient at the previous time and the proportionality weighting coefficient at the current time;

[0131] Step S703, adjusting the proportionality weighting coefficient according to the prediction error, the preset learning rate and the stable regularization term to obtain the adjusted proportionality weighting coefficient.

[0132] In steps S701 and S702 of some embodiments, the weighting coefficient at the previous time is obtained​ , the proportionality weighting coefficient of the previous time and the proportionality weighting coefficient of the current time A stable regularization term is constructed. As shown in the following formula (14):

[0133] (14);

[0134] wherein, represents the stable regularization term, is a regularization coefficient, which controls the penalty strength of the adjustment amplitude. represents the proportionality weighting coefficient of the current time . represents the proportionality weighting coefficient of the previous time . represents the length of the time series.

[0135] In step S703 of some embodiments, the proportionality weighting coefficient is adjusted according to the prediction error, the learning rate and the stable regularization term, to obtain an adjusted proportionality weighting coefficient. That is, in the training process, the stable regularization term is added to the loss function, and the parameters are automatically updated by the optimization algorithm, so that the difference between adjacent times is constrained, so that a smooth and stable sequence is gradually obtained.

[0136] Through the above steps S701 to S703, the stable regularization term is introduced to suppress the excessive fluctuation of the proportionality weighting coefficient , thereby improving the stability of the prediction.

[0137] In some embodiments, in order to minimize the long-term operating cost of the air conditioning system and at the same time maximize the comfort , considering the trade-off in actual demand, the following optimization objective is set, as shown in the following formula (15):

[0138] (15);

[0139] represents the comprehensive optimization objective function value at time , represents the operating cost of the air conditioning system at time , represents the user's comfort score, is the adjustment coefficient between comfort and cost, .

[0140] The prediction error between the actual load demand and the personalized load prediction value to adjust the optimization direction of the system. The prediction error reflects the accuracy of the prediction and is crucial for long-term optimization.

[0141] Using reinforcement learning algorithms such as Q-learning, the prediction model is optimized through continuous interaction with the environment. Based on the current personalized load prediction value , the system can adaptively adjust the control strategy and obtain feedback rewards. The reward function is shown in the following formula (16):

[0142] (16);

[0143] where, is the reward value at time , represents the error penalty coefficient, is the weight of comfort improvement, represents the prediction error, is the user comfort score. The design of the reward function enables the system to prioritize reducing load prediction error while enhancing user comfort. Through long-term feedback of reinforcement learning, the system can continuously adjust the load prediction strategy and output the final optimization result when the convergence condition is met. When the system converges, the strategy is optimal, and the final load prediction value

[0144] is output, representing the optimal prediction of the system considering long-term cost and comfort. The convergence condition is that when the change of the reward function is less than a preset threshold

[0145] , the system considers that it has reached a stable state. At this time, the final load prediction value of the air conditioning system can be used for decision-making in actual applications. Through this step, it is ensured that the air conditioning load prediction not only reflects the current user demand and environmental changes, but also continuously optimizes itself in the long-term operation, achieving the optimal balance of energy efficiency and comfort. This optimization process effectively avoids short-term overfitting through reinforcement learning and error feedback mechanism, improving long-term stability.

[0146] Please refer to

[0147] , the embodiments of the present application also provide an energy efficiency monitoring system for real-time operation of a central air conditioning host system, which can implement the energy efficiency monitoring method for real-time operation of the central air conditioning host system. The system comprises: Figure 8 a collection unit 801 for collecting operation data of the air conditioning system, fusing the operation data according to a preset data weight to obtain fused data;

[0148]

[0149] ​The standardization unit 802 is configured to perform time alignment and space mapping on the fusion data to obtain standardized data, perform dimension feature extraction on the standardized data, and obtain dimension feature data.

[0150] The space extraction unit 803 is configured to perform space feature extraction on the dimension feature data by using a preset convolutional neural network to obtain space feature data.

[0151] The time sequence extraction unit 804 is configured to perform time sequence feature extraction on the space feature data by using a preset long short-term memory network to obtain an initial load prediction value.

[0152] The correction unit 805 is configured to correct the initial load prediction value according to a preset time sequence change regularization term to obtain a corrected load prediction value.

[0153] The specific implementation of the energy efficiency monitoring system for the real-time operation of the central air conditioner host system is basically the same as that of the above-mentioned specific embodiment of the energy efficiency monitoring method for the real-time operation of the central air conditioner host system, and will not be described here.

[0154] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. An energy efficiency monitoring method for real-time operation of a central air-conditioning host system, characterized in that: The method comprises: Collecting operating data of the air-conditioning system, and fusing the operating data according to preset data weights to obtain fused data; Performing time alignment and spatial mapping on the fused data to obtain standardized data, and performing dimensional feature extraction on the standardized data to obtain dimensional feature data; Performing spatial feature extraction on the dimensional feature data through a preset convolutional neural network to obtain spatial feature data; Extracting time series features from the spatial feature data using a preset long short-term memory network to obtain an initial load forecast value; The initial load forecast value is corrected according to a preset time series change regularization term to obtain a corrected load forecast value.

2. The energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to claim 1 is characterized in that: After the initial load forecast value is corrected according to the preset time series change regularization term to obtain a corrected load forecast value, the method further includes: Obtain user demand vector; The corrected load forecast value is corrected according to the user demand vector and a preset weighting factor to obtain a corrected load forecast value; wherein the weighting factor is used to control the degree of fusion between the corrected load forecast value and the user demand vector.

3. The energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to claim 2 is characterized in that: After correcting the modified load forecast value according to the user demand vector and a preset weighting factor to obtain the corrected load forecast value, the method further includes: Obtain the corrected load forecast value of the previous moment and the microclimate characteristics of the current moment; Constructing an adjustment factor based on the corrected load prediction value at the previous moment and the microclimate characteristics at the current moment; wherein the adjustment factor represents a nonlinear relationship between microclimate and load adjustment; The corrected load prediction value is adjusted according to the adjustment factor and the preset adjustment factor weight to obtain a target load prediction value.

4. The energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to claim 3 is characterized in that: The adjusting the corrected load prediction value according to the adjustment factor and the preset adjustment factor weight to obtain a target load prediction value includes: Get the microclimate characteristics of the previous moment; Constructing a smooth load regularization term according to the microclimate characteristics at the previous moment and the microclimate characteristics at the current moment; The corrected load prediction value is adjusted according to the adjustment factor, the preset adjustment factor weight and the smooth load regularization term to obtain a target load prediction value.

5. The energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to claim 3 is characterized in that: The user demand vector includes multiple user demands. After the corrected load prediction value is adjusted according to the adjustment factor and the preset adjustment factor weight to obtain the target load prediction value, the method further includes: Perform weighted synthesis on the multiple user demands to obtain a comprehensive demand index; The comprehensive demand index and the target load forecast value are weighted and fused according to a preset proportional weighting coefficient to obtain a personalized load forecast value.

6. The energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to claim 5 is characterized in that: After weighting and fusing the comprehensive demand index and the target load forecast value according to a preset weighting coefficient to obtain a personalized load forecast value, the method further includes: Obtain actual load demand; Performing a difference calculation based on the actual load demand and the personalized load forecast value to obtain a forecast error; The proportional weighting coefficient is adjusted according to the prediction error and a preset learning rate to obtain an adjusted proportional weighting coefficient.

7. The energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to claim 6, characterized in that: The adjusting the proportional weighting coefficient according to the prediction error and the preset learning rate to obtain the adjusted proportional weighting coefficient includes: Get the proportional weighting coefficient of the previous moment; Constructing a stable regularization term according to the proportional weighting coefficient at the previous moment and the proportional weighting coefficient at the current moment; The proportional weighting coefficient is adjusted according to the prediction error, the preset learning rate and the stable regularization term to obtain the adjusted proportional weighting coefficient.

8. The energy efficiency monitoring system for the real-time operation of the central air-conditioning host system is characterized by: The system comprises: a collection unit, configured to collect operating data of the air-conditioning system, and fuse the operating data according to preset data weights to obtain fused data; a normalization unit, configured to perform time alignment and spatial mapping on the fused data to obtain normalized data, and perform dimensional feature extraction on the normalized data to obtain dimensional feature data; A spatial extraction unit, configured to extract spatial features from the dimensional feature data using a preset convolutional neural network to obtain spatial feature data; A time series extraction unit is used to extract time series features from the spatial feature data through a preset long short-term memory network to obtain an initial load forecast value; The correction unit is used to correct the initial load prediction value according to a preset time series change regularization term to obtain a corrected load prediction value.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements the energy efficiency monitoring method for real-time operation of a central air-conditioning host system as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the energy efficiency monitoring method for real-time operation of a central air-conditioning host system according to any one of claims 1 to 7 is implemented.

Images