Air conditioner terminal energy consumption prediction method, system, equipment and medium

Abstract

The invention provides an air conditioner terminal energy consumption prediction method, system and device and a medium, and belongs to the technical field of energy management. According to the method, historical operation data of the split air conditioner terminal are collected, energy consumption intensity, operation frequency and energy-saving potential index values are calculated and converted into grade numerical values to form feature vectors, and a K-means clustering algorithm is used for generating a grading result library containing priority labels; configuring energy consumption monitoring modules of corresponding grades for different tail ends based on the priority labels; generating a preprocessed data stream through hierarchical data acquisition and hierarchical filtering rules; performing preprocessing on the data streams according to the priorities, and extracting features of different depths to form a hierarchical feature data set; and finally, respectively training an LSTM-BP fusion model, a simplified LSTM model and a weighted average prediction model based on the hierarchical feature data set, dynamically adjusting model parameters through a deviation between real-time monitoring data and a prediction result, and outputting an energy consumption prediction result.

Description

Technical Field

[0001] This invention belongs to the field of energy management technology, and more specifically relates to a method, system, device and medium for predicting energy consumption at air conditioning terminals. Background Technology

[0002] In the field of building energy conservation, split-type air conditioners are the main energy-consuming devices, and accurate energy consumption prediction is key to achieving effective energy efficiency management. Currently, the mainstream technical approach for energy consumption monitoring and prediction of split-type air conditioners relies on the deployment of independent monitoring devices at the terminal socket level. The core of this approach is to equip each split-type air conditioner terminal with a dedicated energy consumption monitoring module to achieve full data collection of current, voltage, power, and other data, and to build a predictive model based on this data.

[0003] However, this paradigm of full deployment and full data collection has several inherent and insurmountable core flaws. First, at the hardware deployment and system operation and maintenance level, the solution suffers from poor economic efficiency and operability. Because the number of monitoring devices is strictly proportional to the number of air conditioning terminals, the costs of hardware procurement, installation, and subsequent maintenance increase linearly with the scale of the terminals, making it particularly unsuitable for large building complexes or existing building renovation projects. The coexistence of a large number of devices also brings problems such as complex communication protocol adaptation, poor system scalability, and a heavy maintenance burden due to frequent failures during long-term operation, seriously affecting the overall stability and lifespan of the system.

[0004] A more fundamental problem lies in data application and model performance. Existing forecasting methods generally use all collected raw data directly for modeling, lacking the ability to discern and filter the inherent value of the data. This results in massive amounts of redundant information from data from low-energy-consumption, low-frequency operating devices. This invalid data dilutes key features and interferes with the model's ability to uncover high-value patterns, thus becoming a major bottleneck in improving forecast accuracy. Extensive practical experience shows that forecasting models built in this way generally suffer from high error rates, making it difficult to meet the needs of refined energy management. Summary of the Invention

[0005] To address the above problems, the present invention aims to provide a method, system, device, and medium for predicting the energy consumption of air conditioning terminals. By using a priority clustering mechanism to achieve differentiated allocation of monitoring resources, and combining a hierarchical data processing flow with a customized prediction model, the invention effectively improves the prediction accuracy of the energy consumption of air conditioning terminals while significantly reducing hardware costs.

[0006] To achieve the above objectives, the present invention employs the following technical solution: In a first aspect, embodiments of this application provide a method for predicting the energy consumption of an air conditioning terminal, including: Collect historical operation datasets of split air conditioner terminals, calculate energy consumption intensity index, operation frequency index, and energy saving potential index based on the historical operation dataset, convert the three index values ​​into level values ​​to form feature vectors based on preset threshold conversion, and generate a hierarchical result library containing priority labels through K-means clustering algorithm; Based on priority tags, configure corresponding energy consumption monitoring modules for split air conditioning terminals with different priorities, and generate equipment configuration parameter tables. The configured energy consumption monitoring module performs hierarchical data acquisition according to the acquisition parameters in the device configuration parameter table, generates raw monitoring data stream, and applies hierarchical filtering rules to process the raw monitoring data stream to generate preprocessed data stream. Based on priority labels, wavelet transform denoising, moving average denoising, or simple outlier removal are performed on preprocessed data streams of different priorities to extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels. Based on the feature vectors and priority labels in the hierarchical feature dataset, an LSTM-BP fusion model, a simplified LSTM model, and a weighted average prediction model are trained respectively. The error index is calculated by monitoring the deviation between the data and the prediction results in real time, the model parameters are dynamically adjusted, and the energy consumption prediction results are output.

[0007] In an optional implementation, the historical operation dataset of the split air conditioner terminals is collected. Based on the historical operation dataset, energy consumption intensity index, operation frequency index, and energy-saving potential index are calculated. The three index values ​​are converted into graded numerical values ​​to form a feature vector based on a preset threshold conversion. A graded result library containing priority labels is generated using a K-means clustering algorithm, including: Collect historical operation data of split air conditioner terminals. The historical operation data includes cumulative energy consumption, running time, start-stop frequency, rated power, energy efficiency level, and the building space area corresponding to the split air conditioner terminals. The energy intensity index value NH is calculated using the formula NH=SN / S; where SN is the cumulative energy consumption and S is the building space area. The average daily runtime is determined based on runtime and used as the runtime index value PC. The energy-saving potential index value JN is calculated using the formula JN=(SN-ZN) / ZN×100%; where ZN is the theoretical optimal energy consumption data, ZN=T×P0 / L, T is the operating time, P0 is the rated power, and L is the energy efficiency level; The energy intensity index value NH, the operating frequency index value PC, and the energy-saving potential index value JN are compared with the corresponding threshold ranges to determine the energy intensity level value, the operating frequency level value, and the energy-saving potential level value. If NH > 1.2k, the energy intensity level is 3; if 0.5k ≤ NH ≤ 1.2k, the energy intensity level is 2; if NH < 0.5k, the energy intensity level is 1. If PC > 8, the frequency level value is 3; if 3 ≤ PC ≤ 8, the frequency level value is 2; if PC < 3, the frequency level value is 1. If JN > 30%, the energy-saving potential level is 3; if 10% ≤ JN ≤ 30%, the energy-saving potential level is 2; if JN < 10%, the energy-saving potential level is 1. The energy intensity level value, the operation frequency level value, and the energy saving potential level value are combined to form a feature vector, denoted as [energy intensity level value, operation frequency level value, energy saving potential level value]; Set the number of clusters K=3, and use the K-means clustering algorithm to iteratively calculate the Euclidean distance and update the cluster centers to divide the feature vector into three clusters, corresponding to high, medium and low priority labels respectively. Generate a hierarchical result library containing priority labels, where each split air conditioner terminal corresponds to a priority label.

[0008] In an optional implementation, the step of configuring corresponding levels of energy consumption monitoring modules for split-type air conditioning terminals with different priorities based on priority tags, and generating a device configuration parameter table, includes: Read the priority tags in the classification result database and configure the corresponding level of energy consumption monitoring modules for different priority split air conditioner terminals; Configure a high-precision energy consumption monitoring module with an integrated 0.5-level precision sensor for the split air conditioner terminal corresponding to the high priority tag, set the sampling frequency to 1 minute / time, and adopt the real-time transmission mode; Configure a conventional energy consumption monitoring module with an integrated 1.0-level precision sensor for the split air conditioner terminal corresponding to the medium priority label, set the sampling frequency to 5 minutes / time, and adopt the batch transmission mode; For the split air conditioning terminals corresponding to low-priority tags, a shared sampling method is configured, with a sampling frequency set to 2 hours / time. A real-time transmission connection is established between the high-precision energy consumption monitoring module and the data acquisition platform via LoRa / WiFi, and a transmission connection is established between the conventional energy consumption monitoring module and the data acquisition platform via a batch transmission channel. A device configuration parameter table is generated based on the acquisition frequency, transmission mode, sensor accuracy, and communication protocol parameters.

[0009] In an optional implementation, the configured energy consumption monitoring module performs tiered data acquisition according to the acquisition parameters in the device configuration parameter table, generates a raw monitoring data stream, and applies tiered filtering rules to process the raw monitoring data stream to generate a preprocessed data stream, including: The configured energy consumption monitoring module performs hierarchical data acquisition according to the acquisition frequency parameters in the device configuration parameter table, generating a raw monitoring data stream. The raw monitoring data stream includes power, energy consumption, operating status, and environmental parameters. Based on the priority labels corresponding to the original monitoring data streams, the original monitoring data streams are divided into high-priority data streams, medium-priority data streams, and low-priority data streams; The box plot method is used to filter out outliers in high-priority, medium-priority, and low-priority data streams, removing data that exceeds the range of Q1-1.5IQR or Q3+1.5IQR. For medium-priority and low-priority data streams, redundant data removal rules are applied to remove duplicate data that has been collected five times consecutively and has a change rate of less than 1%.

[0010] In an optional implementation, the step of performing wavelet transform denoising, moving average denoising, or simple outlier removal on preprocessed data streams of different priorities according to priority labels, extracting multi-dimensional deep features, core features, or basic features, and integrating them to form a hierarchical feature dataset containing feature vectors and priority labels, including: Based on the preprocessed data stream, it is grouped according to priority labels into high-priority data stream, medium-priority data stream and low-priority data stream; For high-priority data streams, wavelet transform is applied for noise reduction and linear interpolation is used to complete missing values; For medium-priority data streams, moving average noise reduction and nearest neighbor fill-in are applied to complete missing values; For low-priority data streams, outlier removal is performed. Deep features are extracted from high-priority data streams to generate high-priority feature vectors; deep features include peak energy consumption periods, energy consumption-temperature and humidity correlation coefficients, start-stop cycle features, power fluctuation amplitude, and energy consumption trend features; Core features are extracted from the medium-priority data stream to generate a medium-priority feature vector; the core features include cumulative energy consumption, average daily runtime, and peak power. Basic features are extracted from low-priority data streams to generate low-priority feature vectors; the basic features include cumulative energy consumption and operation frequency. A hierarchical feature dataset containing feature vectors and priority labels is generated based on high-priority feature vectors, medium-priority feature vectors, and low-priority feature vectors.

[0011] In an optional implementation, the step of training an LSTM-BP fusion model, a simplified LSTM model, and a weighted average prediction model based on the feature vectors and priority labels in the hierarchical feature dataset, respectively, calculating error indicators by real-time monitoring of the deviation between the data and the prediction results, dynamically adjusting model parameters, and outputting energy consumption prediction results includes: Based on the hierarchical feature dataset, high-priority feature vectors are input into the LSTM-BP fusion model for model training, medium-priority feature vectors are input into the simplified LSTM model for model training, and low-priority feature vectors are input into the weighted average prediction model for model training. During model training, mean squared error is used as the loss function, and the Adam optimizer is used for parameter optimization. The learning rate is set to 0.001, the batch size is set to 32, the maximum number of iterations is 100, and the early stopping mechanism is enabled when the validation set error no longer decreases after 5 consecutive iterations. After training, a graded energy consumption prediction model is generated, and the preprocessed data of the split air conditioner terminals collected in real time is input into the graded energy consumption prediction model to output the energy consumption prediction results. The mean absolute percentage error (MAPE) is calculated every 24 hours based on the deviation between real-time monitoring data and the prediction results of the graded energy consumption prediction model. The calculation formula is as follows:

[0012] in, This represents the actual energy consumption at time point i. This represents the predicted energy consumption value at the i-th time point, and n represents the total number of data points; When MAPE exceeds 10%, model parameter adjustment is triggered, the learning rate of the LSTM model is adjusted to 0.0005 and the weight coefficients are recalculated, and the time decay weight coefficients of the weighted average prediction model are recalculated.

[0013] In an optional implementation, the LSTM layer of the LSTM-BP fusion model is set to 128 units, the input sequence length is set to 24 hours, an attention mechanism is introduced to calculate feature weights, the hidden layer of the BP neural network is set to 64 nodes, and the ReLU activation function is configured. The simplified LSTM model has 64 LSTM layers and an input sequence length of 12 hours.

[0014] Secondly, embodiments of this application also provide an air conditioning terminal energy consumption prediction system, including: The historical data acquisition and classification module is used to collect historical operation datasets of split air conditioner terminals, calculate energy consumption intensity index, operation frequency index, and energy saving potential index based on the historical operation dataset, convert the three index values ​​into grade values ​​to form a feature vector based on a preset threshold conversion, and generate a classification result library containing priority labels through K-means clustering algorithm. The equipment configuration module is used to configure the corresponding level of energy consumption monitoring module for split air conditioning terminals with different priorities based on priority tags, and generate an equipment configuration parameter table. The data acquisition and preprocessing module is used to perform hierarchical data acquisition according to the acquisition parameters in the device configuration parameter table through the configured energy consumption monitoring module, generate raw monitoring data stream, and apply hierarchical filtering rules to process the raw monitoring data stream to generate preprocessed data stream; The feature extraction module is used to perform wavelet transform noise reduction, moving average noise reduction, or simple outlier removal on preprocessed data streams of different priorities according to priority labels, extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels. The model training and prediction module is used to train the LSTM-BP fusion model, the simplified LSTM model, and the weighted average prediction model based on the feature vectors and priority labels in the hierarchical feature dataset. It calculates the error index by monitoring the deviation between the data and the prediction results in real time, dynamically adjusts the model parameters, and outputs the energy consumption prediction results.

[0015] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the air conditioning terminal energy consumption prediction method as described in any of the above.

[0016] Fourthly, embodiments of this application also provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the air conditioning terminal energy consumption prediction method as described in any of the above claims.

[0017] As can be seen from the above technical solutions, the present invention has the following advantages: The air conditioning terminal energy consumption prediction method provided in this application significantly improves the accuracy and efficiency of energy consumption prediction by constructing a hierarchical data processing and prediction framework. First, it prioritizes and clusters air conditioning terminals based on historical data, establishing a data value assessment mechanism. Then, it implements differentiated collection, noise reduction, and feature extraction strategies for data of different priorities, effectively eliminating data redundancy and concentrating computing resources on high-value information processing. Finally, it forms a complete technical closed loop from data cleaning and feature optimization to model adaptation by matching priority-based customized prediction models and adjusting dynamic parameters based on real-time errors. This method significantly improves overall data processing efficiency while ensuring the accuracy of core data prediction, providing accurate and reliable technical support for air conditioning terminal energy consumption management.

[0018] This application establishes a scientific and reasonable priority evaluation system for terminal equipment through multi-indicator comprehensive evaluation based on historical operating data and K-means cluster analysis. This system can accurately identify key terminal equipment with high energy consumption, high operating frequency, and significant energy-saving potential, achieving precise allocation of monitoring resources. By configuring differentiated energy consumption monitoring modules for terminals of different priorities, a high-precision, high-frequency monitoring scheme is used only for a few high-priority terminals, while a low-cost shared sampling method is used for a large number of low-priority terminals. This significantly reduces hardware procurement costs and installation and maintenance expenses, greatly improving the project's economic efficiency, and is particularly suitable for energy consumption monitoring scenarios of large-scale split air conditioning terminals.

[0019] This application implements differentiated processing strategies for data streams of different priorities. High-priority data is processed using advanced methods such as wavelet transform denoising and linear interpolation completion to ensure data quality; medium- and low-priority data are processed using lightweight methods such as moving average denoising and simple outlier removal. Through box plotting and redundant data removal rules, outlier data and duplicate information are effectively identified and filtered, improving data quality from the source. This hierarchical processing mechanism avoids wasting computational resources, ensures that high-value data receives sufficient processing attention, and guarantees overall data processing efficiency, laying a solid foundation for subsequent feature extraction and model training.

[0020] This application designs differentiated feature extraction schemes based on data priority, achieving a precise match between feature engineering and business needs. For high-priority data, multi-dimensional deep features are extracted, including peak energy consumption periods, energy consumption-temperature and humidity correlation coefficients, and start-stop cycle characteristics. For medium-priority data, core features such as cumulative energy consumption and average daily runtime are extracted. For low-priority data, only basic statistical features are extracted. This hierarchical feature extraction strategy ensures feature richness at the critical end while avoiding excessive feature engineering on low-value ends. Simultaneously, LSTM-BP fusion models, simplified LSTM models, and weighted average prediction models are adapted for different priority data, ensuring a reasonable match between model complexity and data value, and improving overall prediction efficiency.

[0021] This application effectively ensures prediction accuracy and long-term system stability by constructing a multi-level prediction model system and a dynamic parameter adjustment mechanism. The LSTM-BP fusion model combines the advantages of time series processing and nonlinear mapping, and introduces an attention mechanism to enhance key feature learning; the simplified LSTM model controls computational overhead while maintaining basic prediction capabilities; the weighted average prediction model provides a lightweight solution for low-priority data. By periodically calculating the MAPE index and setting a 10% adjustment threshold, prediction deviations can be detected in a timely manner and parameter optimization can be automatically triggered, ensuring continuous prediction accuracy.

[0022] The standardized priority assessment framework and modular processing flow provided in this application endow the system with excellent scalability and engineering applicability. Newly added air conditioning terminals can be quickly integrated into the existing management system through unified index calculation and cluster analysis, automatically assigning corresponding monitoring levels and processing strategies, greatly simplifying the system expansion process. The hierarchical configuration scheme is compatible with monitoring equipment of different accuracy levels and supports multiple communication methods such as LoRa / WiFi, enhancing the system's adaptability to different scenarios. The end-to-end hierarchical management, from data acquisition and processing to feature extraction and model prediction, forms standardized operating procedures, providing reliable technical support for large-scale engineering deployment and long-term operation and maintenance, and has good prospects for widespread application. Attached Figure Description

[0023] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart illustrating the air conditioning terminal energy consumption prediction method provided in this application.

[0025] Figure 2This is a schematic diagram of the structure of the air conditioning terminal energy consumption prediction system provided in this application.

[0026] Figure 3 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0027] The various embodiments of this disclosure will be described more fully in the detailed steps of the air conditioning terminal energy consumption prediction method described below. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0028] In the following, the terms “comprising” or “may include”, which may be used in various embodiments of this disclosure, indicate the presence of the disclosed functions, operations, or elements, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in various embodiments of this disclosure, the terms “comprising,” “having,” and their cognates are intended only to indicate a particular feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or the possibility of adding one or more combinations of the foregoing.

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please see Figure 1 The diagram shows a flowchart of a method for predicting the energy consumption of an air conditioning terminal in a specific embodiment. The method includes: S1: Collect historical operation datasets of split air conditioner terminals, calculate energy intensity index, operation frequency index, and energy saving potential index based on the historical operation dataset, convert the three index values ​​into level values ​​to form feature vectors based on preset threshold conversion, and generate a hierarchical result library containing priority labels through K-means clustering algorithm.

[0031] In a specific implementation, firstly, historical operational data sets of the split-type air conditioning terminals are collected using temporary monitoring equipment deployed within the building. The collection period is no less than 30 days to ensure coverage of operational characteristics during weekdays, weekends, and different seasons. The collected data includes equipment parameters and external related parameters. The equipment parameters cover cumulative energy consumption, operating time, start-stop frequency, rated power, and energy efficiency rating. The external related parameters mainly refer to the building space area served by each split-type air conditioning terminal.

[0032] After the data collection is completed, three core indicators are calculated.

[0033] The energy intensity index NH is calculated using the formula NH=SN / S. Where SN is the cumulative energy consumption and S is the building space area.

[0034] The operating frequency index (PC) is directly taken as the average daily operating time. The energy-saving potential index (JN) is calculated using the formula JN=(SN-ZN) / ZN×100%, where the theoretical optimal energy consumption ZN =T×P0 / L, T is the operating time, P0 is the rated power, and L is the energy efficiency level.

[0035] Based on the calculation results, energy intensity levels are classified according to preset thresholds. When NH > 1.2 kWh / m², the energy intensity level is recorded as 3; when 0.5 kWh / m² ≤ NH ≤ 1.2 kWh / m², it is recorded as 2; when NH < 0.5 kWh / m², it is recorded as 1. The operating frequency level is classified according to the operating frequency index value PC: PC > 8 hours is recorded as 3, 3 hours ≤ PC ≤ 8 hours is recorded as 2, and PC < 3 hours is recorded as 1. The energy saving potential level is classified according to the energy saving potential index value JN: JN > 30% is recorded as 3, 10% ≤ JN ≤ 30% is recorded as 2, and JN < 10% is recorded as 1.

[0036] Three levels of numerical values ​​are combined to form a feature vector [energy consumption intensity level value, operation frequency level value, and energy saving potential level value], which is then input into the K-means clustering algorithm. The number of clusters, K=3, is set, and the Euclidean distance from each feature vector to the cluster center is iteratively calculated, continuously updating the cluster center positions until convergence. The final output of three clusters corresponds to high, medium, and low priority labels, forming a standardized hierarchical result library. This result library is stored in the form of a database table, supporting querying priority labels by the split air conditioner terminal device number.

[0037] S2: Based on priority tags, configure corresponding energy consumption monitoring modules for split air conditioning terminals with different priorities, and generate equipment configuration parameter tables.

[0038] In a specific implementation, a differentiated monitoring equipment deployment strategy is implemented based on the priority tags in the hierarchical result database. For high-priority split-type air conditioner terminals, a high-precision energy consumption monitoring module is installed in its power supply circuit. This high-precision energy consumption monitoring module integrates a 0.5-level accuracy current / voltage sensor, with a sampling frequency set to 1 minute / time, and uses a LoRa or WiFi communication module to achieve real-time data transmission.

[0039] For medium-priority split-type air conditioner terminals, a standard precision energy consumption monitoring module is deployed, equipped with a Class 1.0 precision sensor. The data acquisition frequency is set to 5 minutes / time, and a timed batch transmission mode is used, with data being packaged and transmitted every 30 minutes. For low-priority split-type air conditioner terminals, no separate dedicated monitoring equipment is deployed. Instead, shared sampling is achieved through the expansion interface of a nearby energy consumption monitoring module, with the acquisition frequency set to 2 hours / time.

[0040] Simultaneously, a hierarchical communication network architecture is established. The high-precision energy consumption monitoring module establishes a real-time connection with the local gateway via LoRa / WiFi, while the conventional precision energy consumption monitoring module uses a batch transmission mode via the 4G network. Based on the above configuration, a device configuration parameter table is generated. This table records the device number, installation location, acquisition frequency, transmission mode, and communication protocol parameters of each energy consumption monitoring module, providing a configuration basis for subsequent data acquisition.

[0041] S3: Through the configured energy consumption monitoring module, hierarchical data acquisition is performed according to the acquisition parameters in the device configuration parameter table to generate raw monitoring data stream, and hierarchical filtering rules are applied to process the raw monitoring data stream to generate preprocessed data stream.

[0042] In a specific implementation, the hierarchical data acquisition process is first initiated based on the equipment configuration parameter table. The high-precision energy consumption monitoring module continuously collects power, energy consumption, operating status, and environmental parameter data at 1-minute intervals; the conventional precision energy consumption monitoring module collects core energy consumption data at 5-minute intervals; and the shared sampling module collects cumulative energy consumption data at 2-hour intervals. All collected data forms a raw monitoring data stream, including fields such as timestamp, equipment identifier, and energy consumption value.

[0043] Then, a tiered filtering process is implemented on the raw monitoring data stream. First, outlier detection is performed on all data using box plots, removing outliers exceeding the range [Q1-1.5IQR, Q3+1.5IQR]. For medium-priority and low-priority data streams, an additional redundant data removal rule is applied: when the energy consumption change rate of five consecutive collections is less than 1%, only the first and last data points are retained, effectively reducing data storage and transmission pressure.

[0044] Finally, a dynamic adjustment mechanism is established. When a sudden change in the operating status of a split-type air conditioner terminal is detected, the system automatically adjusts the sampling frequency of that terminal. For example, if a low-priority split-type air conditioner terminal experiences a surge in energy consumption, the sampling frequency is temporarily adjusted from once every 2 hours to once every 30 minutes to ensure complete recording of abnormal states.

[0045] S4: Based on the priority labels, perform wavelet transform denoising, moving average denoising, or simple outlier removal on the preprocessed data streams of different priorities to extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels.

[0046] In a specific implementation, the preprocessed data streams are first grouped according to priority labels. For high-priority data streams, wavelet transform is used for noise reduction, high-frequency noise is filtered out through thresholding, and missing data points are filled in using linear interpolation. For medium-priority data streams, moving average noise reduction is used for data smoothing, and missing values ​​are filled in using nearest neighbor values. For low-priority data streams, only outlier removal is performed.

[0047] In the feature extraction stage, multi-dimensional deep features are extracted for high-priority data streams, including: identifying peak energy consumption periods through a sliding window, calculating the correlation coefficient between energy consumption and temperature and humidity, extracting start-stop cycle features, calculating power fluctuation amplitude, and fitting energy consumption trend features through linear regression, ultimately forming a 15-dimensional feature vector. For medium-priority data streams, three core features are extracted: cumulative energy consumption, average daily runtime, and peak power. For low-priority data streams, two basic features are extracted: cumulative energy consumption and operating frequency.

[0048] All features are aligned according to time series, and high-priority feature data is controlled to account for 60%-70%, while medium-priority and low-priority feature data account for 30%-40% through random sampling, forming a hierarchical feature dataset. This hierarchical feature dataset is divided into training, validation, and test sets in a 7:2:1 ratio to prepare for model training.

[0049] S5: Based on the feature vectors and priority labels in the hierarchical feature dataset, train the LSTM-BP fusion model, the simplified LSTM model, and the weighted average prediction model respectively. Calculate the error index by monitoring the deviation between the data and the prediction results in real time, dynamically adjust the model parameters, and output the energy consumption prediction results.

[0050] In a specific implementation, a differentiated prediction model architecture is constructed based on a hierarchical feature dataset. High-priority feature vectors are input into an LSTM-BP fusion model, where the LSTM layer has 128 units, the input sequence length is 24 hours, an attention mechanism is introduced to dynamically weight important features, and the BP neural network has 64 hidden layers. Medium-priority feature vectors are input into a simplified LSTM model, with the LSTM layer having 64 units and the input sequence length being 12 hours. Low-priority feature vectors are used with a weighted average prediction model, based on historical 7-day energy consumption data, calculating predicted values ​​according to time decay weights. The time decay weight for the most recent day is 0.4, the time decay weight for the day before is 0.3, and so on, decreasing sequentially.

[0051] During model training, mean squared error was used as the loss function, and the Adam optimizer was used for parameter optimization. The initial learning rate was set to 0.001, the batch size to 32, and the maximum number of iterations to 100. An early stopping mechanism was employed during training, automatically terminating training when the validation set error no longer decreased after 5 consecutive iterations to prevent overfitting.

[0052] After training, a graded energy consumption prediction model is generated and deployed. The preprocessed data of the split air conditioner terminals collected in real time is input into the graded energy consumption prediction model, and the energy consumption prediction results are output.

[0053] After model deployment, a dynamic calibration mechanism is established. The mean absolute percentage error (MAPE) is calculated every 24 hours using the following formula:

[0054] in, This represents the actual energy consumption at time point i. Let represent the predicted energy consumption value at time point i, and n represent the total number of data points.

[0055] When the mean absolute percentage error (MAPE) exceeds 10%, parameter adjustments are triggered: the learning rate of the LSTM model is reduced to 0.0005, and the weight coefficients are recalculated; the time decay weights of the weighted average prediction model are recalculated. Simultaneously, the model is retrained weekly using the latest hierarchical feature dataset to ensure continuous optimization of prediction accuracy.

[0056] Finally, the graded energy consumption prediction model outputs energy consumption prediction results, including predicted values, 95% confidence intervals, and accuracy indicators. When the predicted value exceeds 20% of the historical energy consumption value for the same period, an energy consumption anomaly warning is automatically triggered, providing decision support for energy management.

[0057] In this embodiment, a complete hierarchical technology system was constructed to optimize the entire process of monitoring and predicting energy consumption at the air conditioning terminal. A scientific hierarchical standard was established based on multi-dimensional feature analysis, enabling accurate classification of monitoring objects. A differentiated monitoring strategy was adopted, significantly reducing system costs while ensuring data quality. Intelligent data governance methods effectively improved data value density. The established hierarchical prediction model architecture ensured the prediction accuracy of key equipment while optimizing the overall allocation of computing resources. The entire solution possesses excellent adaptability, dynamically adjusting system parameters according to equipment operating status, providing a reliable technical foundation for refined building energy consumption management.

[0058] like Figure 2 As shown, the following are embodiments of the air conditioning terminal energy consumption prediction system provided in this disclosure. This system and the air conditioning terminal energy consumption prediction methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the air conditioning terminal energy consumption prediction system, please refer to the embodiments of the above air conditioning terminal energy consumption prediction methods.

[0059] An air conditioning terminal energy consumption prediction system includes: The historical data acquisition and grading module is used to collect historical operation datasets of split air conditioner terminals. Based on the historical operation datasets, it calculates energy consumption intensity index values, operation frequency index values, and energy-saving potential index values. Based on a preset threshold conversion, it converts the three index values ​​into graded numerical values ​​to form a feature vector. The K-means clustering algorithm is used to generate a grading result library containing priority labels.

[0060] The device configuration module is used to configure the corresponding level of energy consumption monitoring module for split air conditioning terminals with different priorities based on priority tags, and generate a device configuration parameter table.

[0061] The data acquisition and preprocessing module is used to perform hierarchical data acquisition according to the acquisition parameters in the device configuration parameter table through the configured energy consumption monitoring module, generate raw monitoring data stream, and apply hierarchical filtering rules to process the raw monitoring data stream to generate preprocessed data stream.

[0062] The feature extraction module is used to perform wavelet transform denoising, moving average denoising, or simple outlier removal on preprocessed data streams of different priorities according to priority labels, extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels.

[0063] The model training and prediction module is used to train the LSTM-BP fusion model, the simplified LSTM model, and the weighted average prediction model based on the feature vectors and priority labels in the hierarchical feature dataset. It calculates the error index by monitoring the deviation between the data and the prediction results in real time, dynamically adjusts the model parameters, and outputs the energy consumption prediction results.

[0064] The air conditioning terminal energy consumption prediction system provided in this embodiment achieves precise allocation of monitoring resources through priority clustering analysis based on historical operating data, effectively reducing hardware costs. By constructing a hierarchical data processing flow, it implements differentiated data cleaning strategies such as wavelet transform noise reduction and moving average processing for data of different priorities, significantly improving data quality. By matching customized prediction algorithms such as LSTM-BP fusion models and simplified LSTM models to data of different priorities, and establishing a dynamic parameter adjustment mechanism based on real-time errors, it achieves optimized allocation of computing resources while ensuring the prediction accuracy of core equipment, ultimately achieving the best balance between energy consumption prediction accuracy and economic benefits.

[0065] Figure 3 A schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.

[0066] The air conditioning terminal energy consumption prediction method provided in this application embodiment can be applied to electronic devices. Those skilled in the art will understand that the electronic device structure involved in the embodiments of this invention does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this invention, the electronic device includes, but is not limited to, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0067] Electronic devices may include processors, external memory interfaces, internal memory, universal serial bus (USB) interfaces, charging management modules, power management modules, batteries, wireless communication modules, audio modules, speakers, microphones, sensor modules, buttons, cameras, displays, and SIM card interfaces, etc.

[0068] A processor may include one or more processing units, such as: a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors.

[0069] The processor can serve as the nerve center and command center of an electronic device. The controller can generate operation control signals based on the instruction opcode and timing signals to control the fetching and execution of instructions.

[0070] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from this memory. This avoids repeated accesses, reduces processor latency, and thus improves system efficiency.

[0071] An external storage interface (ESI) can be used to connect external memory cards, such as microSD cards, to expand the storage capacity of electronic devices. The external memory card communicates with the processor through the ESI to perform data storage functions, such as saving music and video files on the external memory card.

[0072] Internal memory can be used to store computer executable program code, which includes instructions. The processor executes various functional applications and data processing of electronic devices by running the instructions stored in internal memory. Internal memory can include a program storage area and a data storage area. Internal memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one disk storage device, flash memory device, universal flash storage (UFS), etc.

[0073] Wireless communication functionality in electronic devices can be achieved through antennas, wireless communication modules, modem processors, and baseband processors.

[0074] Wireless communication modules can provide solutions for wireless communication applications in electronic devices, including wireless local area networks (WLANs) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), and infrared (IR) technologies.

[0075] Electronic devices can implement audio functions through audio modules, speakers, receivers, microphones, headphone jacks, and application processors.

[0076] Electronic devices can achieve shooting functions through ISPs, cameras, video codecs, GPUs, displays, and application processors.

[0077] Electronic devices can achieve display functions through GPUs, displays, and application processors.

[0078] A GPU is a microprocessor for image processing, connected to the display screen and application processor. GPUs are used to perform mathematical and geometric calculations for graphics rendering. A processor may include one or more GPUs, which execute program instructions to generate or modify display information.

[0079] A display screen is used to display images, videos, etc. A display screen includes a display panel.

[0080] The aforementioned electronic device realizes the energy consumption prediction method for air conditioning terminals in this application, which achieves differentiated resource allocation based on priority clustering, establishes a hierarchical data processing flow, and adopts a customized prediction model, thus achieving the beneficial effect of significantly reducing monitoring costs while effectively improving the accuracy of energy consumption prediction.

[0081] The storage medium provided in this application stores a program product capable of implementing an air conditioning terminal energy consumption prediction method.

[0082] Methods for predicting energy consumption at air conditioning terminals include: Collect historical operation datasets of split air conditioner terminals, calculate energy consumption intensity index, operation frequency index, and energy saving potential index based on the historical operation dataset, convert the three index values ​​into level values ​​to form feature vectors based on preset threshold conversion, and generate a hierarchical result library containing priority labels through K-means clustering algorithm; Based on priority tags, configure corresponding energy consumption monitoring modules for split air conditioning terminals with different priorities, and generate equipment configuration parameter tables. The configured energy consumption monitoring module performs hierarchical data acquisition according to the acquisition parameters in the device configuration parameter table, generates raw monitoring data stream, and applies hierarchical filtering rules to process the raw monitoring data stream to generate preprocessed data stream. Based on priority labels, wavelet transform denoising, moving average denoising, or simple outlier removal are performed on preprocessed data streams of different priorities to extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels. Based on the feature vectors and priority labels in the hierarchical feature dataset, an LSTM-BP fusion model, a simplified LSTM model, and a weighted average prediction model are trained respectively. The error index is calculated by monitoring the deviation between the data and the prediction results in real time, the model parameters are dynamically adjusted, and the energy consumption prediction results are output.

[0083] In some possible implementations, the air conditioning terminal energy consumption prediction method of this disclosure can be implemented as a program product that includes program code, which, when the program product is run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of this disclosure.

[0084] The storage medium disclosed herein can take the form of any combination of one or more readable media. A readable medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0085] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting energy consumption at air conditioning terminals, characterized in that, include: Collect historical operation datasets of split air conditioner terminals, calculate energy consumption intensity index, operation frequency index, and energy saving potential index based on the historical operation dataset, convert the three index values ​​into level values ​​to form feature vectors based on preset threshold conversion, and generate a hierarchical result library containing priority labels through K-means clustering algorithm; Based on priority tags, configure corresponding energy consumption monitoring modules for split air conditioning terminals with different priorities, and generate equipment configuration parameter tables. The configured energy consumption monitoring module performs hierarchical data acquisition according to the acquisition parameters in the device configuration parameter table, generates raw monitoring data stream, and applies hierarchical filtering rules to process the raw monitoring data stream to generate preprocessed data stream. Based on priority labels, wavelet transform denoising, moving average denoising, or simple outlier removal are performed on preprocessed data streams of different priorities to extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels. Based on the feature vectors and priority labels in the hierarchical feature dataset, an LSTM-BP fusion model, a simplified LSTM model, and a weighted average prediction model are trained respectively. The error index is calculated by monitoring the deviation between the data and the prediction results in real time, the model parameters are dynamically adjusted, and the energy consumption prediction results are output.

2. The method for predicting energy consumption at air conditioning terminals according to claim 1, characterized in that, The system collects historical operation datasets from split-type air conditioning terminals. Based on these datasets, it calculates energy consumption intensity, operation frequency, and energy-saving potential indicators. Then, using a preset threshold, it converts these three indicators into numerical levels to form a feature vector. A hierarchical result library containing priority labels is generated using the K-means clustering algorithm, including: Collect historical operation data of split air conditioner terminals. The historical operation data includes cumulative energy consumption, running time, start-stop frequency, rated power, energy efficiency level, and the building space area corresponding to the split air conditioner terminals. The energy intensity index value NH is calculated using the formula NH=SN / S; where SN is the cumulative energy consumption and S is the building space area. The average daily runtime is determined based on runtime and used as the runtime index value PC. The energy-saving potential index value JN is calculated using the formula JN=(SN-ZN) / ZN×100%; where ZN is the theoretical optimal energy consumption data, ZN=T×P0 / L, T is the operating time, P0 is the rated power, and L is the energy efficiency level; The energy intensity index value NH, the operating frequency index value PC, and the energy-saving potential index value JN are compared with the corresponding threshold ranges to determine the energy intensity level value, the operating frequency level value, and the energy-saving potential level value. If NH > 1.2k, the energy intensity level is 3; if 0.5k ≤ NH ≤ 1.2k, the energy intensity level is 2; if NH < 0.5k, the energy intensity level is 1. If PC > 8, the frequency level value is 3; if 3 ≤ PC ≤ 8, the frequency level value is 2; if PC < 3, the frequency level value is 1. If JN > 30%, the energy-saving potential level is 3; if 10% ≤ JN ≤ 30%, the energy-saving potential level is 2; if JN < 10%, the energy-saving potential level is 1. The energy intensity level value, the operation frequency level value, and the energy saving potential level value are combined to form a feature vector, denoted as [energy intensity level value, operation frequency level value, energy saving potential level value]; Set the number of clusters K=3, and use the K-means clustering algorithm to iteratively calculate the Euclidean distance and update the cluster centers to divide the feature vector into three clusters, corresponding to high, medium and low priority labels respectively. Generate a hierarchical result library containing priority labels, where each split air conditioner terminal corresponds to a priority label.

3. The method for predicting energy consumption at air conditioning terminals according to claim 2, characterized in that, The method involves configuring corresponding energy consumption monitoring modules for split-type air conditioning terminals with different priorities based on priority tags, and generating a device configuration parameter table, including: Read the priority tags in the classification result database and configure the corresponding level of energy consumption monitoring modules for different priority split air conditioner terminals; Configure a high-precision energy consumption monitoring module with an integrated 0.5-level precision sensor for the split air conditioner terminal corresponding to the high priority tag, set the sampling frequency to 1 minute / time, and adopt the real-time transmission mode; Configure a conventional energy consumption monitoring module with an integrated 1.0-level precision sensor for the split air conditioner terminal corresponding to the medium priority label, set the sampling frequency to 5 minutes / time, and adopt the batch transmission mode; For the split air conditioning terminals corresponding to low-priority tags, a shared sampling method is configured, with a sampling frequency set to 2 hours / time. A real-time transmission connection is established between the high-precision energy consumption monitoring module and the data acquisition platform via LoRa / WiFi, and a transmission connection is established between the conventional energy consumption monitoring module and the data acquisition platform via a batch transmission channel. A device configuration parameter table is generated based on the acquisition frequency, transmission mode, sensor accuracy, and communication protocol parameters.

4. The air conditioning terminal energy consumption prediction method according to claim 3, characterized in that, The configured energy consumption monitoring module performs tiered data acquisition according to the acquisition parameters in the device configuration parameter table, generates a raw monitoring data stream, and applies tiered filtering rules to process the raw monitoring data stream to generate a preprocessed data stream, including: The configured energy consumption monitoring module performs hierarchical data acquisition according to the acquisition frequency parameters in the device configuration parameter table, generating a raw monitoring data stream. The raw monitoring data stream includes power, energy consumption, operating status, and environmental parameters. Based on the priority labels corresponding to the original monitoring data streams, the original monitoring data streams are divided into high-priority data streams, medium-priority data streams, and low-priority data streams; The box plot method is used to filter out outliers in high-priority, medium-priority, and low-priority data streams, removing data that exceeds the range of Q1-1.5IQR or Q3+1.5IQR. For medium-priority and low-priority data streams, redundant data removal rules are applied to remove duplicate data that has been collected five times consecutively and has a change rate of less than 1%.

5. The method for predicting energy consumption at air conditioning terminals according to claim 4, characterized in that, The process involves performing wavelet transform denoising, moving average denoising, or simple outlier removal on preprocessed data streams of different priorities based on priority labels, extracting multi-dimensional deep features, core features, or basic features, and integrating them to form a hierarchical feature dataset containing feature vectors and priority labels, including: Based on the preprocessed data stream, it is grouped according to priority labels into high-priority data stream, medium-priority data stream and low-priority data stream; For high-priority data streams, wavelet transform is applied for noise reduction and linear interpolation is used to complete missing values; For medium-priority data streams, moving average noise reduction and nearest neighbor fill-in are applied to complete missing values; For low-priority data streams, outlier removal is performed. Deep features are extracted from high-priority data streams to generate high-priority feature vectors; deep features include peak energy consumption periods, energy consumption-temperature and humidity correlation coefficients, start-stop cycle features, power fluctuation amplitude, and energy consumption trend features; Core features are extracted from the medium-priority data stream to generate a medium-priority feature vector; the core features include cumulative energy consumption, average daily runtime, and peak power. Basic features are extracted from low-priority data streams to generate low-priority feature vectors; the basic features include cumulative energy consumption and operation frequency. A hierarchical feature dataset containing feature vectors and priority labels is generated based on high-priority feature vectors, medium-priority feature vectors, and low-priority feature vectors.

6. The method for predicting the energy consumption of air conditioning terminals according to claim 5, characterized in that, Based on the feature vectors and priority labels in the hierarchical feature dataset, an LSTM-BP fusion model, a simplified LSTM model, and a weighted average prediction model are trained respectively. Error indices are calculated by real-time monitoring of the deviation between the data and the prediction results, model parameters are dynamically adjusted, and energy consumption prediction results are output, including: Based on the hierarchical feature dataset, high-priority feature vectors are input into the LSTM-BP fusion model for model training, medium-priority feature vectors are input into the simplified LSTM model for model training, and low-priority feature vectors are input into the weighted average prediction model for model training. During model training, mean squared error is used as the loss function, and the Adam optimizer is used for parameter optimization. The learning rate is set to 0.001, the batch size is set to 32, the maximum number of iterations is 100, and the early stopping mechanism is enabled when the validation set error no longer decreases after 5 consecutive iterations. After training, a graded energy consumption prediction model is generated, and the preprocessed data of the split air conditioner terminals collected in real time is input into the graded energy consumption prediction model to output the energy consumption prediction results. The mean absolute percentage error (MAPE) is calculated every 24 hours based on the deviation between real-time monitoring data and the prediction results of the graded energy consumption prediction model. The calculation formula is as follows: in, This represents the actual energy consumption at time point i. This represents the predicted energy consumption value at the i-th time point, and n represents the total number of data points; When MAPE exceeds 10%, model parameter adjustment is triggered, the learning rate of the LSTM model is adjusted to 0.0005 and the weight coefficients are recalculated, and the time decay weight coefficients of the weighted average prediction model are recalculated.

7. The method for predicting energy consumption at air conditioning terminals according to claim 6, characterized in that, The LSTM-BP fusion model has 128 units in its LSTM layer, 24 hours in its input sequence, and an attention mechanism to calculate feature weights. The BP neural network has 64 hidden layers and is configured with the ReLU activation function. The simplified LSTM model has 64 LSTM layers and an input sequence length of 12 hours.

8. An air conditioning terminal energy consumption prediction system, characterized in that, The system employs the air conditioning terminal energy consumption prediction method as described in any one of claims 1 to 7; The system includes: The historical data acquisition and classification module is used to collect historical operation datasets of split air conditioner terminals, calculate energy consumption intensity index, operation frequency index, and energy saving potential index based on the historical operation dataset, convert the three index values ​​into grade values ​​to form a feature vector based on a preset threshold conversion, and generate a classification result library containing priority labels through K-means clustering algorithm. The equipment configuration module is used to configure the corresponding level of energy consumption monitoring module for split air conditioning terminals with different priorities based on priority tags, and generate an equipment configuration parameter table. The data acquisition and preprocessing module is used to perform hierarchical data acquisition according to the acquisition parameters in the device configuration parameter table through the configured energy consumption monitoring module, generate raw monitoring data stream, and apply hierarchical filtering rules to process the raw monitoring data stream to generate preprocessed data stream; The feature extraction module is used to perform wavelet transform noise reduction, moving average noise reduction, or simple outlier removal on preprocessed data streams of different priorities according to priority labels, extract multi-dimensional deep features, core features, or basic features, and integrate them to form a hierarchical feature dataset containing feature vectors and priority labels. The model training and prediction module is used to train the LSTM-BP fusion model, the simplified LSTM model, and the weighted average prediction model based on the feature vectors and priority labels in the hierarchical feature dataset. It calculates the error index by monitoring the deviation between the data and the prediction results in real time, dynamically adjusts the model parameters, and outputs the energy consumption prediction results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the air conditioning terminal energy consumption prediction method as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the air conditioning terminal energy consumption prediction method as described in any one of claims 1 to 7.

Images