Energy storage method and system for base station air conditioner energy consumption management

Abstract

The application discloses a base station air conditioner energy consumption management energy storage method and system, and the method comprises the following steps: collecting multi-source data in real time according to the temperature sensor, humidity sensor, air conditioner running state monitoring equipment and communication equipment load data in the base station machine room, and generating a multi-source fusion data set; predicting the dynamic refrigeration demand in the future time window for the multi-source fusion data set, and constructing a dynamic optimization model of air conditioner energy consumption through the energy consumption efficiency curve; dynamically adjusting the running mode, air speed and temperature setting value of the air conditioner according to the prediction result of the dynamic refrigeration demand and the energy consumption optimization model, optimizing the energy storage scheduling strategy, and generating a joint optimization scheme of air conditioner running and energy storage scheduling; and dynamically adjusting the joint optimization scheme to generate the final base station air conditioner energy consumption management scheme by monitoring the air conditioner running state in real time. According to the embodiment of the application, the air conditioner energy consumption management has real-time and accuracy, and more efficient energy consumption control and energy storage scheduling optimization can be realized.

Description

Technical Field

[0001] This invention belongs to the field of energy storage technology, and in particular to a method and system for energy storage management of base station air conditioning. Background Technology

[0002] With the rapid development of wireless communication technology, the surge in the number of base stations has led to a corresponding increase in the demand for electricity. The cooling systems required for base station operation, especially air conditioning, account for a significant portion of the total electricity consumption. The high energy consumption of base station air conditioning not only increases operating costs but may also cause base stations to malfunction due to overheating, thereby affecting the stability and reliability of communication services. Therefore, how to effectively manage and optimize the energy consumption of base station air conditioning has become a pressing technical problem for the telecommunications industry.

[0003] Traditional base station energy management typically relies on a single data source, such as real-time data from air conditioning temperature and humidity sensors. This method cannot comprehensively reflect the operational status within the base station. Furthermore, real-time load changes in base station communication equipment and external environmental temperature fluctuations are not adequately considered, leading to inaccurate energy consumption predictions and difficulty in effectively adjusting air conditioning operation strategies. Existing energy management methods largely depend on static models, lacking dynamic response capabilities. This results in the air conditioning's cooling demands not being met in a timely manner under high load or extreme weather conditions, thereby increasing excessive air conditioning operation and energy consumption. Summary of the Invention

[0004] The purpose of this invention is to provide a base station air conditioning energy consumption management and energy storage method and system to overcome the shortcomings of the prior art, so that air conditioning energy consumption management can not only be real-time and accurate, but also achieve more efficient energy consumption control and energy storage scheduling optimization through intelligent means.

[0005] One embodiment of this application provides a base station air conditioning energy consumption management and energy storage method, the method comprising:

[0006] Based on the temperature sensor, humidity sensor, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room, a data acquisition and preprocessing method based on edge computing is adopted to collect multi-source data in real time. Through adaptive filtering algorithm, environmental noise and electromagnetic interference are eliminated to generate multi-source fusion dataset.

[0007] For multi-source fusion datasets, a prediction model based on lightweight spatiotemporal neural networks is adopted. Combining the real-time load changes of base station communication equipment and external environmental temperature fluctuations, the dynamic cooling demand within the future time window is predicted. And through the energy consumption efficiency curve, a dynamic optimization model of air conditioning energy consumption is constructed.

[0008] Based on the prediction results of dynamic cooling demand and the energy consumption optimization model, a reinforcement learning-based air conditioning operation strategy optimization algorithm is adopted to dynamically adjust the air conditioning operation mode, fan speed and temperature setpoint. At the same time, combined with the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

[0009] The system monitors the air conditioning operation status in real time, uses a fault early warning model based on anomaly detection algorithm to identify potential air conditioning faults, and dynamically adjusts the joint optimization scheme through an adaptive control mechanism to ensure the stability and energy efficiency of air conditioning operation, thus generating the final base station air conditioning energy consumption management scheme.

[0010] Optionally, based on temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data within the base station equipment room, a data acquisition and preprocessing method based on edge computing is used to collect multi-source data in real time. An adaptive filtering algorithm is then used to eliminate environmental noise and electromagnetic interference, generating a multi-source fusion dataset, including:

[0011] Based on the temperature sensor, humidity sensor, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room, an edge computing-based data acquisition framework is adopted. Multi-source data is acquired in real time through distributed data acquisition nodes. Each acquisition node is equipped with a lightweight data caching mechanism to ensure the real-time and continuous nature of data acquisition.

[0012] For the collected multi-source data, an adaptive filtering algorithm based on wavelet transform is used to separate environmental noise and electromagnetic interference in the signal. Through a dynamic threshold adjustment mechanism, differentiated filtering is performed on the noise characteristics of different sensor data to obtain a preliminary denoised multi-source dataset.

[0013] For the denoised multi-source dataset, a time alignment algorithm based on dynamic time warping is used to eliminate timestamp differences between different sensor data; at the same time, a missing value imputation method based on spatiotemporal correlation is used to interpolate and fill in the missing parts of the data to generate a time-synchronized complete dataset.

[0014] For a complete dataset that is synchronized in time, a feature fusion method based on a multi-head attention mechanism is used to extract multi-dimensional features of temperature, humidity, air conditioning operation status and communication equipment load data. Through an adaptive feature weight allocation mechanism, a high-precision multi-source fusion dataset is generated.

[0015] Optionally, for the multi-source fusion dataset, a prediction model based on a lightweight spatiotemporal neural network is used, combining real-time load changes of base station communication equipment and external environmental temperature fluctuations, to predict dynamic cooling demand within a future time window, and a dynamic optimization model for air conditioning energy consumption is constructed through energy efficiency curves, including:

[0016] For the multi-source fusion dataset, a prediction model based on a lightweight spatiotemporal neural network is adopted. The temporal features of the load change of base station communication equipment are extracted through the temporal convolutional layer, and the local spatial features of the temperature distribution in the equipment room are captured through the spatial graph convolutional layer to generate a spatiotemporal feature representation.

[0017] By combining external environmental temperature fluctuation data, an attention-based external environment fusion technique is adopted to weightedly fuse external temperature change characteristics with spatiotemporal feature representations, thereby enhancing the predictive model's ability to predict dynamic cooling demand and generating an environmentally enhanced spatiotemporal feature representation.

[0018] The spatiotemporal features of the enhanced environment are input into the output layer of the prediction model. A time series decoder based on multi-step prediction is used to predict the dynamic cooling demand of base station equipment rooms within future time windows and generate cooling demand prediction results.

[0019] Based on historical operating data of air conditioners, an energy efficiency curve fitting method based on piecewise linear regression is used to construct a relationship model between air conditioner energy consumption, cooling capacity, and operating mode. Combined with the cooling demand prediction results, a dynamic optimization model for air conditioner energy consumption is generated.

[0020] Optionally, based on the predicted results of dynamic cooling demand and the energy consumption optimization model, a reinforcement learning-based air conditioning operation strategy optimization algorithm is used to dynamically adjust the air conditioning operation mode, fan speed, and temperature setpoint. Simultaneously, combined with the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling, including:

[0021] Based on the dynamic cooling demand prediction results and energy consumption optimization model, an air conditioning operation strategy optimization algorithm based on deep reinforcement learning is adopted to construct a reward function with the goal of minimizing energy consumption and satisfying cooling demand. Through a dual Q network structure, the air conditioning operation mode, fan speed and temperature setpoint are dynamically adjusted to generate a preliminary air conditioning operation optimization strategy.

[0022] By combining the current state of charge of the energy storage unit, the charge and discharge efficiency curves, and the grid electricity price fluctuation data, a multi-dimensional constraint model of the energy storage state is constructed. Through energy conservation constraints, charge and discharge rate constraints, and grid load balance constraints, the feasible solution space for energy storage scheduling is defined.

[0023] For the energy storage state constraint model, an energy storage scheduling method based on distributed constraint optimization algorithm is adopted. Combined with the energy consumption demand in the air conditioning operation optimization strategy, the charging and discharging plan of the energy storage unit is dynamically adjusted. Through time-segmented electricity price sensitivity analysis, the energy storage scheduling strategy is optimized and a preliminary energy storage scheduling scheme is generated.

[0024] The air conditioning operation optimization strategy and energy storage scheduling scheme are optimized in a coordinated manner. A joint optimization framework based on game theory is adopted. Through the dynamic game between the air conditioning and energy storage systems, the energy consumption cost and the cooling demand satisfaction are balanced to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

[0025] Optionally, the real-time monitoring of the air conditioner's operating status, employing a fault early warning model based on anomaly detection algorithms to identify potential air conditioner faults, and dynamically adjusting the joint optimization scheme through an adaptive control mechanism to ensure the air conditioner's operational stability and energy efficiency, generates the final base station air conditioner energy consumption management scheme, including:

[0026] Real-time monitoring of air conditioner operation status; collection of air conditioner operation parameters and energy consumption data; processing of collected data in real time through edge computing nodes; generation of time series dataset of air conditioner operation status.

[0027] For time series datasets, an anomaly detection algorithm based on a combination of isolated forest and long short-term memory network is used to identify potential air conditioner faults and generate fault warning signals through a dynamic threshold adjustment mechanism.

[0028] Based on the fault warning signal, an adaptive control mechanism based on fuzzy logic is adopted to dynamically adjust the joint optimization scheme. Through a multi-objective optimization model, the fault recovery cost and energy efficiency are balanced to generate an adaptive control strategy.

[0029] The adaptive control strategy is integrated with the joint optimization scheme, and a dynamic adjustment method based on feedback correction mechanism is adopted to update the air conditioning operating parameters and energy storage scheduling plan in real time, thereby generating the final base station air conditioning energy consumption management scheme.

[0030] Another embodiment of this application provides a base station air conditioning energy consumption management and energy storage system, the system comprising:

[0031] The data acquisition module is used to collect multi-source data in real time based on temperature sensors, humidity sensors, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room. It adopts edge computing-based data acquisition and preprocessing methods, eliminates environmental noise and electromagnetic interference through adaptive filtering algorithms, and generates multi-source fusion datasets.

[0032] The prediction module is used to predict the dynamic cooling demand within a future time window by using a prediction model based on a lightweight spatiotemporal neural network on a multi-source fusion dataset, combined with the real-time load changes of base station communication equipment and external environmental temperature fluctuations. It also constructs a dynamic optimization model for air conditioning energy consumption through energy efficiency curves.

[0033] The adjustment module is used to dynamically adjust the air conditioner's operating mode, fan speed, and temperature setpoint based on the predicted results of dynamic cooling demand and the energy consumption optimization model, using a reinforcement learning-based air conditioner operation strategy optimization algorithm. At the same time, it optimizes the energy storage scheduling strategy by combining the charging and discharging status of the energy storage unit, and generates a joint optimization scheme for air conditioner operation and energy storage scheduling.

[0034] The management module is used to monitor the air conditioner's operating status in real time. It adopts a fault early warning model based on anomaly detection algorithm to identify potential air conditioner faults. Through an adaptive control mechanism, it dynamically adjusts the joint optimization scheme to ensure the stability and energy efficiency of the air conditioner's operation and generate the final base station air conditioner energy consumption management scheme.

[0035] Another embodiment of this application provides a storage medium storing a computer program, wherein the computer program is configured to execute the method described in any of the preceding claims when running.

[0036] Another embodiment of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the method described in any of the preceding claims.

[0037] Compared with existing technologies, the present invention provides a base station air conditioning energy consumption management and energy storage method. This method collects multi-source data in real time from temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data within the base station equipment room, generating a multi-source fusion dataset. Based on this multi-source fusion dataset, it predicts dynamic cooling demand within future time windows and constructs a dynamic optimization model for air conditioning energy consumption using energy efficiency curves. According to the predicted dynamic cooling demand and the energy consumption optimization model, it dynamically adjusts the air conditioning operation mode, fan speed, and temperature setpoints, optimizes energy storage scheduling strategies, and generates a joint optimization scheme for air conditioning operation and energy storage scheduling. It monitors the air conditioning operation status in real time, dynamically adjusts the joint optimization scheme, and generates the final base station air conditioning energy consumption management scheme. This enables air conditioning energy consumption management to not only be real-time and accurate but also achieve more efficient energy consumption control and energy storage scheduling optimization through intelligent means. Attached Figure Description

[0038] Figure 1 A hardware structure block diagram of a computer terminal for a base station air conditioning energy consumption management and energy storage method provided in an embodiment of the present invention;

[0039] Figure 2 A flowchart illustrating a base station air conditioning energy consumption management and energy storage method provided in an embodiment of the present invention;

[0040] Figure 3 This is a schematic diagram of a base station air conditioning energy consumption management and energy storage system provided in an embodiment of the present invention. Detailed Implementation

[0041] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0042] This invention first provides a base station air conditioning energy consumption management and energy storage method, which can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0043] The following detailed explanation uses a computer terminal as an example. Figure 1 This is a hardware structure block diagram of a computer terminal for a base station air conditioning energy consumption management and energy storage method provided in an embodiment of the present invention. (See diagram below.) Figure 1 As shown, the computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0044] Non-volatile storage media can store operating systems and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any base station air conditioning energy management and energy storage method.

[0045] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0046] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any base station air conditioning energy management and energy storage method.

[0047] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0048] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0049] See Figure 2 The present invention provides a base station air conditioning energy storage management method, which may include the following steps:

[0050] S201: Based on the temperature sensor, humidity sensor, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room, the data acquisition and preprocessing method based on edge computing is adopted to collect multi-source data in real time, and through adaptive filtering algorithm, environmental noise and electromagnetic interference are eliminated to generate multi-source fusion dataset;

[0051] This method utilizes edge computing-based data acquisition and preprocessing technology to collect data in real time from various sensors within the base station equipment room. These sensors include temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and load data from communication equipment. Adaptive filtering algorithms effectively eliminate environmental noise and electromagnetic interference, ensuring the accuracy and reliability of the collected data. The resulting multi-source fusion dataset provides a solid foundation for subsequent dynamic cooling demand prediction and energy consumption optimization, guaranteeing the overall system's operational efficiency.

[0052] This combination of data acquisition and preprocessing ensures data quality and validity while monitoring the environment and equipment operating status within the base station equipment room in real time. This not only improves the accuracy of energy management and reduces error risks, but also provides solid data support for intelligent management decisions. By integrating multiple data sources, a more comprehensive understanding of the base station's operating status can be achieved, enabling effective responses to various emergencies, optimization of air conditioning energy management and energy storage scheduling strategies, and ultimately achieving multiple goals such as energy conservation, consumption reduction, and improved base station operational stability.

[0053] Specifically, based on the temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room, an edge computing-based data acquisition framework can be adopted to acquire multi-source data in real time through distributed data acquisition nodes. Each acquisition node is equipped with a lightweight data caching mechanism to ensure the real-time and continuous nature of data acquisition.

[0054] This design ensures that the data acquisition process is not only efficient and real-time, but also guarantees data continuity and integrity. The distributed data acquisition architecture effectively reduces the burden on the central processing node while improving the overall system's flexibility and adaptability to changing data collection needs. Real-time acquisition of multi-source data provides a timely and accurate foundation for subsequent data processing and analysis, contributing to the development of more scientific air conditioning energy consumption management strategies.

[0055] In practical implementation, multiple distributed data acquisition nodes can be set up within the base station equipment room. Each node is responsible for monitoring sensor data within a specific area; for example, one node might focus on the environment around air conditioning equipment, while another might be located next to communication equipment. Each node is equipped with a microprocessor and memory module, enabling it to collect real-time data streams from sensors, such as temperature, humidity, air conditioning operating status, and communication equipment load data. By implementing edge computing on the nodes, data can be preliminarily processed locally, reducing latency in data transmission to the central server and improving overall response speed.

[0056] The data acquisition node design also includes a lightweight data caching mechanism. This mechanism ensures that nodes can continue to operate and cache the latest data even when the network is unstable or the central server fails. For example, when a node cannot connect to the server, it temporarily stores the data in memory and sends it to the central system as soon as the network is restored. This mechanism not only improves the reliability of data acquisition but also ensures seamless monitoring over long periods, enabling continuous and consistent recording of environmental data from the base station equipment room.

[0057] In this way, base stations can not only acquire various monitoring data in real time, but also ensure the effectiveness of data collection under any conditions. This flexible, edge-computing-based data acquisition architecture provides a solid foundation for subsequent data processing and analysis, and lays extremely important support for air conditioning energy consumption management and energy storage scheduling.

[0058] For the collected multi-source data, an adaptive filtering algorithm based on wavelet transform is used to separate environmental noise and electromagnetic interference in the signal. Through a dynamic threshold adjustment mechanism, differentiated filtering is performed on the noise characteristics of different sensor data to obtain a preliminary denoised multi-source dataset.

[0059] By using an adaptive filtering algorithm based on wavelet transform, noise caused by environmental influences can be effectively removed, improving the signal-to-noise ratio of the data. This is a crucial step in ensuring the accuracy of subsequent analysis results. Differentiated filtering processes can make data from different types of sensors more reliable, enhancing the quality of the entire multi-source fusion dataset and thus providing accurate basis for subsequent data analysis and decision-making.

[0060] In implementing this step, data from different sensors must first be input into the wavelet transform algorithm. Wavelet transform is an effective signal processing tool that decomposes a signal into multiple frequency levels, allowing us to clearly distinguish between noise and useful signals. For example, temperature sensors may be affected by airflow or electromagnetic interference, while the load data of communication equipment may fluctuate due to interference from other electrical devices. Through wavelet transform, the system can extract the true data features from these frequency components.

[0061] Next, a dynamic threshold adjustment mechanism was introduced to differentiate processing based on the characteristics of different sensors. Traditional fixed thresholds may not be suitable for noise levels under different environmental conditions. Therefore, the system monitors the ambient noise in real time and dynamically adjusts the filter threshold based on the statistical information of the collected signals. For example, in a high-noise environment, the system may increase the threshold so that small fluctuations will not affect the final result, while in a quieter environment, the threshold can be decreased to capture more subtle signal changes.

[0062] Finally, the filtered multi-source dataset yields a denoised signal set, which can be used for subsequent data analysis and dynamic prediction. The denoising not only improves data quality but also provides clear and reliable information for subsequent decision-making. This process ensures accurate reflection of the actual environmental conditions within the base station equipment room, creating conditions for precise implementation of air conditioning energy consumption optimization and energy storage scheduling.

[0063] For the denoised multi-source dataset, a time alignment algorithm based on dynamic time warping is used to eliminate timestamp differences between different sensor data; at the same time, a missing value imputation method based on spatiotemporal correlation is used to interpolate and fill in the missing parts of the data to generate a time-synchronized complete dataset.

[0064] By eliminating timestamp differences, the system can achieve synchronous processing of data from different sensors, making subsequent data analysis more accurate. Missing data is a common phenomenon in sensor data; therefore, spatiotemporal correlation is used to impute missing values, ensuring the integrity and coherence of the dataset, which helps improve model training performance and prediction accuracy.

[0065] In this step, the Dynamic Time Warping (DTW) algorithm is first applied to the denoised dataset to eliminate timestamp differences between different sensors. Considering that different sensors may collect data at different time frequencies—for example, a temperature sensor updates every 5 seconds while a humidity sensor updates every 10 seconds—the DTW algorithm can reasonably match these time differences, thereby creating consistency for the data from all sensors within a standard time frame. For instance, the DTW algorithm will construct an optimal data trajectory for each time point, ensuring that readings from all sensors are accurately linked within the same time period.

[0066] Next, since sensor data may be missing during actual operation, a missing value imputation method based on spatiotemporal correlation will be used to handle these missing parts. This method not only considers historical time-series data but also introduces spatial correlation, i.e., data from other adjacent sensors within the base station. For example, if the temperature sensor data is missing at a certain point in time, the system will analyze relevant data such as humidity and air flow at the same point in time and use interpolation methods to predict the missing temperature value based on this information, thereby imputing the missing data.

[0067] Through the above steps, the resulting time-synchronized complete dataset greatly facilitates subsequent data analysis. This dataset not only possesses multidimensional features but also eliminates temporal interference, providing high-quality input data for training models of dynamic cooling demand forecasting and energy consumption optimization, ensuring the accuracy and reliability of subsequent analysis.

[0068] For a complete dataset that is synchronized in time, a feature fusion method based on a multi-head attention mechanism is used to extract multi-dimensional features of temperature, humidity, air conditioning operation status and communication equipment load data. Through an adaptive feature weight allocation mechanism, a high-precision multi-source fusion dataset is generated.

[0069] Feature fusion can comprehensively consider the influence of multiple sensors, extracting more distinctive features and improving the accuracy and depth of data analysis. The introduction of a multi-head attention mechanism enables the system to focus on more relevant information, enhancing the expressive power of the data and generating a high-precision multi-source fusion dataset, providing an accurate foundation for subsequent dynamic cooling demand prediction.

[0070] In this implementation step, the system extracts features from the complete time-synchronized dataset and fuses them using an algorithm based on a multi-head attention mechanism. The advantage of the multi-head attention mechanism lies in its ability to process multiple information sources in parallel, allowing for better capture of the relationships between different features. For example, the system can set up several attention heads, focusing on temperature, humidity, and air conditioning status respectively, to conduct in-depth analysis of these features. Each attention head independently processes the input data and generates a weighted representation of different signals in the output.

[0071] Next, the adaptive feature weight allocation mechanism dynamically adjusts the weights based on the contribution of each feature to the actual prediction. Initially, the system scores the features and then updates the weights in real time based on the model's performance during training. If temperature significantly impacts cooling demand within a certain timeframe, the weight of the temperature feature will increase; conversely, it may decrease. This flexible weight allocation mechanism ensures that the model always focuses on the most important features, maximizing prediction accuracy.

[0072] The resulting high-precision multi-source fusion dataset will be used for subsequent dynamic cooling demand prediction and energy consumption optimization models. This dataset not only comprehensively reflects the environmental conditions and equipment operating status within the base station equipment room, but also provides clearer data support for decision-making. Optimization in this process makes subsequent air conditioning energy management and energy storage scheduling strategies more intelligent and effective, contributing to improved overall energy efficiency and operational efficiency of the base station.

[0073] S202, for multi-source fusion datasets, adopts a prediction model based on lightweight spatiotemporal neural networks, combines real-time load changes of base station communication equipment and external environmental temperature fluctuations to predict dynamic cooling demand within future time windows, and constructs a dynamic optimization model for air conditioning energy consumption through energy efficiency curves;

[0074] This step aims to utilize a prediction model based on a lightweight spatiotemporal neural network to process a multi-source fusion dataset to predict the dynamic cooling demand of base station equipment rooms within future time windows. This process combines real-time load changes of base station communication equipment with external environmental temperature fluctuations. The model extracts temporal features of load changes through temporal convolutional layers and captures local spatial features of temperature distribution within the equipment room using spatial graph convolutional layers, ultimately generating a spatiotemporal feature representation. Furthermore, a dynamic optimization model for air conditioning energy consumption is constructed using energy efficiency curves, providing data support for energy-saving operation of the air conditioning system.

[0075] Applying lightweight spatiotemporal neural networks to predict dynamic cooling demand can significantly improve the response speed and accuracy of energy management systems. By capturing and analyzing real-time load changes in base station communication equipment and the impact of the external environment, this method helps to predict air conditioning demand in advance, thereby allowing for targeted adjustments to air conditioning operation strategies and reducing energy consumption. Furthermore, the establishment of energy efficiency curves provides a basis for developing more scientific energy management schemes, enabling base stations to achieve optimal energy efficiency and operational efficiency during operation, thus improving overall service quality and stability.

[0076] Specifically, for multi-source fusion datasets, a prediction model based on lightweight spatiotemporal neural networks can be used to extract the temporal features of base station communication equipment load changes through temporal convolutional layers and capture the local spatial features of temperature distribution in the equipment room through spatial graph convolutional layers to generate spatiotemporal feature representations.

[0077] In practical implementation, the first step is to import the multi-source fusion dataset into a lightweight spatiotemporal neural network model. This model consists of two key parts: a temporal convolutional layer and a spatial graph convolutional layer. The temporal convolutional layer works by using a sliding window technique to perform convolution operations on the load data of base station communication equipment, thereby capturing the temporal characteristics of load trends and periodic changes. For example, if communication traffic suddenly increases within a certain period, this layer can automatically learn this change pattern through the convolutional kernel and extract the time point of load increase, forming an effective feature representation.

[0078] Next, the spatial graph convolutional layer analyzes the data from various temperature sensors within the computer room. By constructing a graph structure between the sensors, the model can capture the spatial relationships of temperatures in different areas. For example, assuming the temperature on the left side of the computer room suddenly spikes while the right side remains normal, the spatial graph convolutional layer will learn this spatially uneven distribution characteristic through the connections between nodes, enabling the model to understand the impact of temperature changes in local areas on the overall cooling demand.

[0079] Finally, through joint processing by temporal and spatial graph convolutional layers, the generated spatiotemporal feature representation fully reflects the interaction between load changes and temperature distribution. This feature representation not only provides fundamental information for subsequent demand forecasting but also provides necessary data support for policy optimization in reinforcement learning algorithms.

[0080] By combining external environmental temperature fluctuation data, an attention-based external environment fusion technique is adopted to weightedly fuse external temperature change characteristics with spatiotemporal feature representations, thereby enhancing the predictive model's ability to predict dynamic cooling demand and generating an environmentally enhanced spatiotemporal feature representation.

[0081] In this step, the system first needs to acquire real-time temperature fluctuation data of the external environment. This external data includes weather forecasts and real-time monitoring data from meteorological stations, which have a significant impact on the cooling requirements of the computer room. Using an attention-based external environment fusion technique, the external temperature change characteristics can be effectively combined with the previously generated spatiotemporal feature representation. During this process, the attention mechanism assigns different weights to each feature to highlight the more important factors.

[0082] For example, assuming that the external temperature rises rapidly due to climate change during a certain period, the model will increase the weight of the external temperature feature through an attention mechanism, making the prediction model more sensitive to this change when judging cooling demand. This dynamic weight allocation ensures that the model can respond to changes in the external environment in real time and makes the prediction more accurate, thereby effectively improving the energy efficiency of air conditioning.

[0083] Through this weighted fusion process, the generated enhanced spatiotemporal feature representation will accurately reflect environmental changes inside and outside the data center. This enhancement not only improves the model's sensitivity to changes in cooling demand but also provides richer data support for subsequent demand forecasting and strategy optimization, ensuring that the system can maintain efficient operation under different weather conditions.

[0084] The spatiotemporal features of the enhanced environment are input into the output layer of the prediction model. A time series decoder based on multi-step prediction is used to predict the dynamic cooling demand of base station equipment rooms within future time windows and generate cooling demand prediction results.

[0085] In implementing this step, the environmentally enhanced spatiotemporal feature representation is fed as input into the output layer of the prediction model. The decoder used here is based on a multi-step prediction time series model, whose structure allows the model to jointly predict multiple future time points. This design enables the system not only to generate predictions for the next time point but also to simultaneously assess changes in cooling demand across multiple time points.

[0086] The system automatically generates future cooling demand forecasts based on the learned spatiotemporal feature representations. For example, during the forecasting process, if the model identifies that both load and temperature were rising in the previous time period, it will infer that cooling demand is likely to continue to increase in the next few time periods based on this trend. This capability allows base stations to prepare in advance, thereby effectively adjusting the air conditioning operation mode.

[0087] Ultimately, the cooling demand forecasts generated by the time-series decoder are fed back into the energy management system. These results not only guide the air conditioning operation strategy but also help to develop more reasonable energy management measures, ensuring the stability and energy efficiency of base station operation and reducing unnecessary energy waste.

[0088] Based on historical operating data of air conditioners, an energy efficiency curve fitting method based on piecewise linear regression is used to construct a relationship model between air conditioner energy consumption, cooling capacity, and operating mode. Combined with the cooling demand prediction results, a dynamic optimization model for air conditioner energy consumption is generated.

[0089] In this step, the system first needs to accumulate historical operating data of the air conditioner, including operating modes (such as cooling, dehumidification, etc.), cooling capacity, and corresponding energy consumption. This data will be used for fitting energy efficiency curves based on piecewise linear regression. By analyzing energy consumption data under different operating modes, the system can identify energy consumption patterns under each mode, thus providing a basis for subsequent optimization decisions.

[0090] For example, suppose that an air conditioner consumes significantly different amounts of energy under normal cooling and high-load cooling conditions. The system will construct different linear regression models for these two modes, identify their energy efficiency ratios by analyzing historical data, and incorporate external influencing factors such as weather changes into the models. Such analysis helps the system clarify the complex relationship between energy consumption and cooling demand.

[0091] Finally, by combining the cooling demand forecast results, the system will generate a dynamic optimization model for air conditioning energy consumption. This model will be updated in real time, automatically adjusting the air conditioning's operating mode and energy output based on the predicted cooling demand. This ensures that cooling needs are met while minimizing energy consumption, achieving efficient energy management. This dynamic optimization model provides a scientific and effective solution for the energy-saving operation of base stations.

[0092] S203, based on the prediction results of dynamic cooling demand and energy consumption optimization model, adopts a reinforcement learning-based air conditioning operation strategy optimization algorithm to dynamically adjust the air conditioning operation mode, wind speed and temperature setpoint. At the same time, combined with the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

[0093] Based on the predicted dynamic cooling demand and energy consumption optimization model, this step utilizes a reinforcement learning-based air conditioning operation strategy optimization algorithm to dynamically adjust the air conditioner's operating mode, fan speed, and temperature setpoint. Simultaneously, it optimizes the energy storage scheduling strategy by considering the charging and discharging status of the energy storage unit, generating a joint optimization scheme for air conditioning operation and energy storage scheduling. The reinforcement learning algorithm continuously learns and adjusts the strategy through environmental feedback, enabling the air conditioner to achieve the optimal operating mode in the face of real-time changes in load demand and external environmental conditions. This optimization process not only considers the air conditioner's own operating efficiency but also, based on the state of the energy storage unit, allows the system to find the optimal balance between energy consumption and cooling effect.

[0094] Employing reinforcement learning-based optimization algorithms can significantly improve the intelligence level of air conditioning systems, enabling them to dynamically adjust their operating strategies based on real-time data. The core of this method lies in using historical data and real-time status for learning and decision-making, which not only improves the energy efficiency of the air conditioning system but also enhances its responsiveness to sudden changes in demand. This flexible intelligent adjustment allows the base station to reduce energy consumption while ensuring cooling needs are met, improving economic efficiency, and being environmentally friendly, thus contributing to the achievement of sustainable development goals.

[0095] Specifically, based on the dynamic cooling demand prediction results and energy consumption optimization model, an air conditioning operation strategy optimization algorithm based on deep reinforcement learning can be adopted to construct a reward function with the goal of minimizing energy consumption and satisfying cooling demand. Through a dual Q network structure, the air conditioning operation mode, fan speed and temperature setpoint can be dynamically adjusted to generate a preliminary air conditioning operation optimization strategy.

[0096] In this step, the system first collects historical and real-time monitoring data and uses machine learning techniques to predict dynamic cooling demand and energy consumption information. Next, a deep reinforcement learning-based air conditioning operation strategy optimization algorithm is constructed. The core of this algorithm lies in building a reward function that simultaneously considers both energy minimization and meeting cooling demand. By using a dual Q-network structure, the system can quickly learn the optimal action strategy under different operating conditions, i.e., how to reasonably adjust the air conditioner's operating mode, fan speed, and temperature setpoint to minimize energy consumption while meeting cooling demand.

[0097] This optimization process not only makes the air conditioning system more intelligent but also enables it to dynamically adjust based on real-time data, thereby effectively improving energy efficiency. Through deep reinforcement learning, the system can respond more flexibly to environmental changes and load fluctuations than traditional control methods, thus reducing energy costs while ensuring indoor comfort. This intelligent operation strategy lays the foundation for subsequent energy storage scheduling and comprehensive energy management.

[0098] In this step, the system will first establish a dynamic cooling demand prediction model. This model, based on historical temperature, humidity, and air conditioning usage habits, integrates machine learning algorithms to achieve real-time prediction of cooling demand. When the ambient temperature is predicted to be above 30°C, the model will predict an increase in users' demand for air conditioning cooling, while also considering users' lifestyle habits, such as changes in comfort needs at different times of day and night. This prediction result will serve as input to the deep reinforcement learning algorithm.

[0099] Next, the system will construct a reward function using a deep reinforcement learning-based approach. The reward function will be designed to simultaneously minimize energy consumption and meet cooling needs, guiding the model to continuously optimize the air conditioner's operating strategy during the learning process. For example, if the system successfully reduces energy consumption and meets the user's cooling needs, it will receive a higher reward; conversely, if energy consumption is too high or cooling needs are not met, the reward will decrease. Through continuous iteration, the model will gradually learn the optimal strategy, thereby adjusting the air conditioner's operating status under different climatic conditions.

[0100] Finally, a dual-Q network structure will be used to dynamically adjust the air conditioner's operating mode, fan speed, and temperature setpoint. This structure, through alternating learning using two Q networks, effectively reduces variance during strategy updates, improving learning stability and effectiveness. For example, when the indoor temperature exceeds the setpoint, the system can intelligently increase the fan speed and decrease the temperature setpoint, optimizing the air conditioner's operating mode and generating a preliminary air conditioner operation optimization strategy. This strategy will provide the necessary data foundation for subsequent energy storage scheduling steps.

[0101] By combining the current state of charge of the energy storage unit, the charge and discharge efficiency curves, and the grid electricity price fluctuation data, a multi-dimensional constraint model of the energy storage state is constructed. Through energy conservation constraints, charge and discharge rate constraints, and grid load balance constraints, the feasible solution space for energy storage scheduling is defined.

[0102] At this stage, the main task of the system is to establish a multi-dimensional constraint model based on the actual situation of the energy storage unit. This model will consider the current state of charge of the energy storage unit, the charge and discharge efficiency curve, and the fluctuation of the grid electricity price. Specifically, the energy conservation constraint ensures that the charging and discharging process of the energy storage unit conforms to the energy transfer law, that is, the relationship between the charging amount and the discharging amount; the charge and discharge rate constraint limits the rate of the energy storage device during the charging and discharging process, ensuring that it operates within a safe and stable range; and the grid load balance constraint takes into account the load changes of the grid throughout the entire dispatch cycle to balance power supply and demand.

[0103] By constructing such a constraint model, the system can effectively define the feasible solution space for energy storage dispatch, providing a computational basis for subsequent dispatch strategy formulation. This constraint model, which considers multiple dimensions, ensures the scientific and rational nature of energy storage dispatch, enabling the energy storage system to achieve its maximum efficiency under different electricity prices and climate conditions, thus guaranteeing the stability and economy of power supply.

[0104] In this stage, the system will integrate the state of charge (SBC) and charge / discharge efficiency curves of the energy storage units to establish a multi-dimensional constraint model of the energy storage state. The current SBC can be monitored in real time via sensors, ensuring the system always accurately grasps the energy storage unit's capacity. For example, if the current SBC of the energy storage unit is 80%, the maximum charging and discharging limits must be considered when formulating the charging and discharging plan. Simultaneously, the charge / discharge efficiency curves will be used to evaluate the energy conversion losses under specific charging and discharging conditions, ensuring the optimized plan is practically feasible.

[0105] Next, the energy conservation constraint requires that the total energy input and output must remain equal during the charging and discharging process of the energy storage unit. This means that if the energy storage unit discharges 100kWh of energy at a certain moment, then at a later moment, the difference between the charging and discharging amounts must be equal to the initial total amount of energy. Furthermore, the charge / discharge rate constraint ensures that the energy storage system does not exceed its designed maximum capacity during charging and discharging; for example, it is not allowed to discharge more than 50kWh in a short period of time to ensure the stability and safety of the system.

[0106] Finally, grid load balance constraints ensure that the grid load is reasonably transferred within a given time period, meaning that the grid's carrying capacity is not exceeded during high-load periods. For example, during peak electricity consumption periods, the system should reasonably control the discharge volume to avoid putting excessive pressure on the grid. Through the construction of these comprehensive constraints, the system defines the feasible solution space for energy storage dispatch, providing the necessary theoretical foundation for subsequent optimization schemes of energy storage dispatch.

[0107] For the energy storage state constraint model, an energy storage scheduling method based on distributed constraint optimization algorithm is adopted. Combined with the energy consumption demand in the air conditioning operation optimization strategy, the charging and discharging plan of the energy storage unit is dynamically adjusted. Through time-segmented electricity price sensitivity analysis, the energy storage scheduling strategy is optimized and a preliminary energy storage scheduling scheme is generated.

[0108] In this step, the system uses a distributed constraint optimization algorithm to schedule energy storage based on the energy storage state constraint model established in the previous step. This method effectively integrates energy consumption demands from the air conditioning operation optimization strategy, dynamically adjusting the charging and discharging plans of energy storage units according to electricity price changes and load demand. The system analyzes electricity price signals according to time periods, making them an important reference in the decision-making process. Through electricity price sensitivity analysis, the system can identify the optimal timing for charging and discharging in different time periods, thereby formulating appropriate scheduling schemes.

[0109] By dynamically adjusting the charging and discharging schedules of energy storage units, the system can better cope with electricity price fluctuations and load changes, improving the efficiency of the energy storage system. This not only reduces electricity costs during periods of high electricity prices but also ensures stable system operation during power shortages. Furthermore, optimizing energy storage dispatch strategies enhances the flexibility and resilience of overall power management, ensuring both sustainability and economic benefits.

[0110] In this stage, the system will employ a distributed constraint optimization algorithm to schedule the energy storage state constraint model built in the previous step. The advantage of this algorithm lies in its ability to run in parallel across multiple processing nodes, improving computational efficiency. The system will collect electricity demand data and electricity price information for different time periods, and by transforming this data into model inputs, it will optimize the charging and discharging plans of the energy storage units in real time. For example, if the system detects that the electricity price is at its peak during a certain period, the distributed optimization algorithm will prioritize the discharging of energy storage units to reduce overall electricity costs.

[0111] In conjunction with the energy consumption requirements in the air conditioning operation optimization strategy, the energy storage dispatch method will dynamically adjust according to the expected energy consumption of the air conditioning system at different times. For example, when the energy consumption demand of the air conditioning is high during the daytime peak period, the system may arrange for energy storage units to discharge to meet the air conditioning operation requirements, while avoiding overloading of the power grid. At this time, the system will continuously evaluate the charging and discharging strategy to ensure that the maximum energy can be effectively utilized when needed.

[0112] Furthermore, the system will perform time-segmented electricity price sensitivity analysis to optimize energy storage dispatch strategies. By analyzing electricity price changes in each time period, the system can identify when charging costs are lowest and when discharging returns are highest, thereby formulating an optimal charging and discharging plan. For example, if electricity prices are low late at night, the system may schedule energy storage units to charge; while during peak daytime electricity prices, it will adjust to discharging mode. Through the comprehensive application of these strategies, the system ultimately generates a preliminary energy storage dispatch scheme, laying the foundation for subsequent joint optimization.

[0113] The air conditioning operation optimization strategy and energy storage scheduling scheme are optimized in a coordinated manner. A joint optimization framework based on game theory is adopted. Through the dynamic game between the air conditioning and energy storage systems, the energy consumption cost and the cooling demand satisfaction are balanced to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

[0114] In this step, the system integrates the air conditioning operation optimization strategy with the energy storage scheduling scheme, employing a game theory framework for collaborative optimization. Game theory defines the strategic interactions between different participants (the air conditioning system and the energy storage system), enabling each system to dynamically adjust its decisions based on the strategies of others. In this way, the system can achieve an optimal balance between energy costs and cooling demand satisfaction. For example, the air conditioning system may need to reduce load during periods of high electricity prices, while the energy storage system will appropriately discharge according to load balancing requirements to accommodate the needs of both systems.

[0115] This joint optimization scheme helps improve the overall economic efficiency and resource utilization efficiency of the system. Through dynamic game theory, both parties make optimization decisions based on mutual influence, ultimately achieving effective reduction in energy costs and sufficient satisfaction of cooling needs. The advantage of this strategy is that the system can not only cope with instantaneously changing electricity prices, but also adapt to the operational requirements under different load conditions, thereby improving the overall reliability and stability of operation.

[0116] In this step, the system establishes a dynamic game model using game theory by co-optimizing the air conditioning operation optimization strategy and the energy storage scheduling scheme. This game model treats the air conditioning system and the energy storage system as two "players," each optimizing their strategies to minimize energy consumption costs and maximize the satisfaction of cooling demands. Their decisions influence each other; the air conditioning's operating state depends on the energy storage's charging and discharging strategy, while the energy storage scheduling also needs to consider changes in the air conditioning's electricity demand.

[0117] The system will dynamically adjust its strategies through multiple rounds of game theory. For example, if electricity prices surge during a certain period, the air conditioning system can choose to reduce its load to decrease electricity costs, while the energy storage system may discharge at this time to maximize its repayment to the grid. This process is repeated, and after each round of game theory, the system updates the strategies of each participant based on the game results, thereby improving overall operational efficiency and economic benefits.

[0118] Ultimately, the system will integrate the game information from all parties to generate a joint optimization scheme, ensuring that the operation of air conditioning and the scheduling of energy storage are coordinated. For example, when electricity prices remain high, the system may choose to reduce the frequency of energy storage discharge by decreasing the air conditioning load, thereby extending the lifespan of the energy storage units; while during periods of lower electricity prices, the system can moderately increase air conditioning cooling to improve user comfort. This flexible joint scheduling strategy achieves a good balance between energy consumption costs and cooling demand, ensuring the sustainability of electricity use.

[0119] S204 monitors the air conditioner's operating status in real time, uses a fault early warning model based on anomaly detection algorithms to identify potential air conditioner faults, and dynamically adjusts the joint optimization scheme through an adaptive control mechanism to ensure the air conditioner's operational stability and energy efficiency, generating the final base station air conditioner energy consumption management scheme.

[0120] Real-time monitoring of air conditioning operation is crucial for ensuring the efficient and stable operation of base station air conditioning systems. By integrating data from various sensors (such as temperature, humidity, energy consumption, and operating status), the system can track the air conditioning's operation in real time. Employing a fault warning model based on anomaly detection algorithms, the system can effectively identify potential faults. For example, when an unusual increase in air conditioning energy consumption is detected within a short period, or when readings from certain sensors deviate from normal ranges, the system immediately generates a fault warning signal. Based on this, the system dynamically adjusts the joint optimization scheme through an adaptive control mechanism. Specifically, the system automatically optimizes the air conditioning's operating parameters (such as fan speed, temperature setpoint, and operating mode) based on real-time monitoring data and historical performance analysis to ensure that the air conditioning maintains stability and energy efficiency even when a fault occurs. Ultimately, this process generates a comprehensive base station air conditioning energy consumption management scheme, ensuring that the system maintains optimal performance under various operating conditions.

[0121] Implementing a real-time monitoring and fault early warning mechanism significantly enhances the reliability and responsiveness of the air conditioning system. When a potential fault occurs, the system can respond quickly, maintaining normal operation by dynamically adjusting the joint optimization scheme. This mechanism effectively reduces the risk of equipment damage and maintenance costs, avoids operational interruptions due to faults, and thus improves the continuity and reliability of overall service. Furthermore, with the help of real-time data and intelligent adjustment, the air conditioning system can maintain good energy efficiency under conditions of rapid load changes and adverse environments, such as high temperature or high humidity, reducing energy waste. For example, it can automatically increase cooling capacity during high load periods and reduce power consumption when the load decreases, thereby ensuring user comfort and saving operating costs. Ultimately, this dynamically adjusted joint optimization scheme not only improves the energy efficiency of the base station but also lays a solid foundation for the long-term stable operation of the equipment.

[0122] Specifically, it can monitor the air conditioner's operating status in real time, collect the air conditioner's operating parameters and energy consumption data, and process the collected data in real time through edge computing nodes to generate a time series dataset of the air conditioner's operating status.

[0123] In this phase, the system aggregates data from various sensors through edge computing nodes, collecting multiple operating parameters of the air conditioner in real time, such as cooling load, energy consumption, current, temperature, and humidity. Each sensor continuously monitors specific indicators and uploads data to the edge computing nodes at fixed time intervals. The edge nodes are not only responsible for data collection but also for preliminary data processing, such as data cleaning and aggregation, to ensure the accuracy of the information. The generated time-series dataset will provide the foundational data for subsequent fault detection and energy efficiency analysis.

[0124] This real-time monitoring mechanism ensures that the operating status of the air conditioner is accurately reflected, thus providing crucial data support for fault early warning, maintenance decisions, and energy consumption optimization. More timely data acquisition and processing enable faster response to operational anomalies. For example, if a sensor detects a sudden increase in air conditioner energy consumption, the system will immediately record and analyze possible causes to prevent energy waste and equipment damage. This mechanism helps improve the overall operating efficiency of the air conditioning system and provides data support for subsequent decision-making.

[0125] In this phase, the first step is to deploy various sensors within the base station equipment room, including temperature sensors, humidity sensors, wind speed sensors, and energy consumption monitors. Each sensor processes its data centrally via edge computing nodes to ensure real-time performance and efficiency. For example, a temperature sensor can collect temperature changes within the equipment room every second, while the energy consumption monitor records the power consumption of the air conditioning system. This sensor data is transmitted in real-time to the edge computing nodes via wireless or wired networks, ensuring data continuity and real-time performance.

[0126] In processing data, edge computing nodes first preprocess the collected multi-source data, including data cleaning and noise reduction. To improve data quality, edge nodes apply adaptive filtering algorithms to eliminate electromagnetic interference and environmental noise. For example, temperature sensors may produce inaccurate readings due to external noise in certain situations. Adaptive filtering algorithms can identify and correct these abnormal data, thereby generating accurate multi-source datasets. This process is crucial because high-quality data forms the basis for subsequent analysis and decision-making.

[0127] Ultimately, the processed data will be compiled into a time-series dataset containing precise records of air conditioner operating status and energy consumption data. Based on this time-series data, the system will generate real-time monitoring reports to help maintenance personnel assess the air conditioner's efficiency and energy consumption. This data can be used to identify operating trends and potential problems; for example, a significant increase in energy consumption during a certain period may indicate a potential risk of decreased efficiency or malfunction in the air conditioner. In this way, the system can ensure that the air conditioner maintains optimal performance during continuous operation and provides crucial information for subsequent fault warnings and optimized control.

[0128] For time series datasets, an anomaly detection algorithm based on a combination of isolated forest and long short-term memory network is used to identify potential air conditioner faults and generate fault warning signals through a dynamic threshold adjustment mechanism.

[0129] In this step, the system analyzes the previously generated time-series dataset, applying anomaly detection techniques that combine the Isolation Forest and Long Short-Term Memory (LSTM) algorithms. Isolation Forest, as an unsupervised learning algorithm, effectively identifies outliers in the data, while the LSTM network handles the long-term dependencies of the time-series data and evaluates its temporal characteristics. By combining these two algorithms, the system can more accurately detect potential fault signs; for example, when air conditioner energy consumption data suddenly deviates from the normal range, the system will issue a timely warning.

[0130] The implementation of this fault early warning mechanism has significantly improved the reliability and maintenance efficiency of the air conditioning system. By promptly identifying potential faults, the risk of serious equipment failure can be effectively reduced, thereby lowering maintenance costs and extending service life. For example, if the system detects a sudden surge in air conditioning energy consumption within a short period, it may indicate a problem with the equipment, allowing maintenance personnel to intervene in advance and prevent further damage. This mechanism not only enhances the system's safety and stability but also provides better service, enabling more efficient use of the equipment.

[0131] In this step, the system first inputs the time-series dataset into an anomaly detection model based on Isolation Forest. The Isolation Forest algorithm effectively identifies outliers that are significantly different from most data points, especially in the case of high-dimensional data. For example, if energy consumption data suddenly increases sharply within a specific time period, the Isolation Forest algorithm will mark this as a potential anomaly, thus triggering an alarm. This algorithm is applicable to various anomalies in air conditioning operation, such as equipment malfunctions and sensor reading errors, ensuring that potential problems are detected at an early stage.

[0132] Next, by incorporating a Long Short-Term Memory (LSTM) network, the system will utilize the LSTM's memory units to analyze trends and periodic changes in time-series data. LSTM is particularly well-suited for processing continuous time-series data and can capture long-term dependencies. For example, by modeling temperature and energy consumption changes over the past few hours, LSTM can predict normal operating levels under the current environment. If actual data deviates from the range predicted by the LSTM model, the system will trigger a fault warning signal to alert maintenance personnel.

[0133] To improve detection accuracy, the system implements a dynamic threshold adjustment mechanism. Based on real-time monitoring data and changes in environmental conditions, the system automatically updates the alarm thresholds. For example, in the extremely hot summer months, the energy consumption of the air conditioning system naturally increases. Using a fixed threshold for fault detection in such conditions could lead to false alarms. Therefore, the dynamic threshold adjustment mechanism allows the system to automatically adjust warning standards according to changes in the operating environment, ensuring the accuracy and sensitivity of fault detection.

[0134] Based on the fault warning signal, an adaptive control mechanism based on fuzzy logic is adopted to dynamically adjust the joint optimization scheme. Through a multi-objective optimization model, the fault recovery cost and energy efficiency are balanced to generate an adaptive control strategy.

[0135] In this step, the system dynamically adjusts the air conditioner's operating parameters using a fuzzy logic control mechanism based on the fault warning signal generated in the previous step. The fuzzy logic controller can handle uncertainty and fuzziness, adjusting the air conditioner's operating state in real time by defining a series of fuzzy rules. For example, if the fault warning signal indicates an abnormally high energy consumption, the fuzzy logic controller can adaptively adjust by setting the rule "reduce fan speed if energy consumption is high and the temperature target is not achieved." This intelligent control mechanism can quickly take corresponding measures when a fault occurs, such as reducing the air conditioner's cooling capacity, adjusting the temperature setpoint, or switching to a backup system when necessary, to ensure stable operation and energy efficiency optimization. The system can continuously update and optimize control rules based on real-time data and historical performance to adapt to different operating environments and load requirements, ensuring optimal performance under various conditions.

[0136] By implementing a fuzzy logic-based adaptive control mechanism, the system not only improves the air conditioner's responsiveness to potential faults but also significantly enhances operational flexibility and energy efficiency. When the air conditioner is under heavy load or experiences drastic environmental changes, the system can quickly adjust to ensure the equipment is not damaged by overload. For example, on extremely hot days, the air conditioner needs to maintain high cooling efficiency. In this case, the system can automatically optimize operating conditions, allowing the air conditioner to maintain reasonable energy consumption even under high load. This not only reduces energy waste but also helps ensure the long-term stable use of the equipment.

[0137] In this stage, the system will activate a fuzzy logic-based adaptive control mechanism based on the fault warning signal generated in the previous step. Once a potential fault is detected, the system will analyze the type and severity of the fault through the fuzzy logic controller and adjust the air conditioner's operating mode in real time. For example, if a sensor reports insufficient cooling capacity, the fuzzy logic controller may suggest immediately increasing the temperature setpoint or changing the fan speed to reduce the load on the equipment and prevent damage.

[0138] Simultaneously, the system will incorporate joint optimization schemes, involving dynamic adjustments to the air conditioner's operating mode, fan speed, and temperature settings. These adjustments are based not only on real-time fault warning signals but also on energy efficiency. For example, when the cost of restoring the air conditioner from a fault exceeds the energy savings, the system may choose to continue using the current operating mode rather than forcibly repairing the fault. Through such decision-making, the system achieves a trade-off between fault recovery costs and energy efficiency, ensuring that the air conditioner can maintain a relatively stable operating state even in the event of a fault.

[0139] When implementing the adaptive control strategy, the system employs a multi-objective optimization model, comprehensively considering various constraints and performance indicators. While ensuring the normal operation of the air conditioner, the system strives to reduce energy consumption and extend equipment lifespan. For example, if the air conditioner requires maintenance, the system will suggest charging and operation during periods of low electricity prices to minimize maintenance costs and energy consumption. Simultaneously, this control strategy will be continuously updated to adapt to future changes in the operating environment and equipment status, forming a self-optimizing dynamic process.

[0140] The adaptive control strategy is integrated with the joint optimization scheme, and a dynamic adjustment method based on feedback correction mechanism is adopted to update the air conditioning operating parameters and energy storage scheduling plan in real time, thereby generating the final base station air conditioning energy consumption management scheme.

[0141] In this phase, the system will integrate adaptive control strategies with previously developed joint optimization schemes to form a more intelligent and flexible management system. By introducing a feedback correction mechanism, the system can track the actual operating performance of the air conditioner and the charging and discharging performance of the energy storage unit, acquiring performance data in real time and comparing it with preset targets. When actual energy consumption or cooling effect deviates from the ideal state, the system will dynamically adjust operating parameters and energy storage scheduling plans based on feedback information. For example, if the system detects that the air conditioner's energy consumption exceeds the expected target during peak hours, it can adjust the operating mode or lower the set temperature to optimize energy efficiency. Ultimately, the integrated scheme will form a comprehensive base station air conditioning energy consumption management solution, ensuring that the system maintains optimal energy efficiency and stability under different operating conditions and loads.

[0142] This integration step is crucial for improving the system's adaptability. It enables rapid response and adjustment under rapidly changing environmental and load conditions, effectively avoiding the risks of equipment overload or excessive energy consumption. For example, when the external ambient temperature rises sharply, the system can promptly adjust the air conditioning operation strategy to cope with the increased load, while ensuring the rational use of energy storage units and avoiding overcharging and discharging. Through a feedback correction mechanism, the system continuously learns and optimizes, forming a self-improving management process. This not only reduces the complexity of operation and maintenance but also improves the overall operating efficiency of the base station, ensuring minimal energy consumption and maximum comfort, providing strong support for the economic efficiency and sustainability of base station operation.

[0143] In this phase, the system will integrate adaptive control strategies and joint optimization schemes to form a comprehensive air conditioning energy management solution. The system will implement a feedback correction mechanism, continuously monitoring the actual operating status of the air conditioner and comparing it with the output of the predictive model to evaluate the effectiveness of the adaptive control strategy in real time. For example, if the actual energy consumption is significantly higher than expected, the system will automatically adjust the control strategy to match the actual conditions.

[0144] The integrated solution will update the air conditioner's operating parameters, such as temperature settings and fan speed, based on real-time data. Simultaneously, the energy storage scheduling plan will be optimized during this process. By monitoring the charging and discharging status of the energy storage units in real time, the system can make informed decisions to adjust energy utilization, such as charging during periods of lower electricity prices and releasing energy from the storage units to power the air conditioner during peak hours. This strategy ensures that cooling needs are met while reducing overall energy costs.

[0145] The final solution not only meets immediate operational needs but also provides data support for future energy efficiency improvements and equipment maintenance. The system can continuously optimize control strategies based on historical operational data and real-time feedback, forming a closed-loop management system. This approach ensures the operational stability of the air conditioning system while improving overall energy efficiency, thus providing a continuous and stable energy management solution for the base station and ensuring efficient and reliable equipment operation.

[0146] As can be seen, multi-source data is collected in real time from temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room to generate a multi-source fusion dataset. Based on this dataset, dynamic cooling demand within future time windows is predicted, and a dynamic optimization model for air conditioning energy consumption is constructed using energy efficiency curves. According to the predicted dynamic cooling demand and the energy consumption optimization model, the air conditioning operation mode, fan speed, and temperature setpoints are dynamically adjusted, and energy storage scheduling strategies are optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling. Real-time monitoring of the air conditioning operation status and dynamic adjustment of the joint optimization scheme generate the final base station air conditioning energy consumption management scheme. This enables air conditioning energy consumption management to not only be real-time and accurate but also to achieve more efficient energy consumption control and energy storage scheduling optimization through intelligent means.

[0147] Another embodiment of the present invention provides a base station air conditioning energy consumption management and energy storage system, see [link to relevant documentation]. Figure 3 The system may include:

[0148] The acquisition module 301 is used to acquire multi-source data in real time based on the temperature sensor, humidity sensor, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room, and to generate a multi-source fusion dataset by using an edge computing-based data acquisition and preprocessing method, and to eliminate environmental noise and electromagnetic interference through an adaptive filtering algorithm.

[0149] The prediction module 302 is used to predict the dynamic cooling demand within a future time window by using a prediction model based on a lightweight spatiotemporal neural network on a multi-source fusion dataset, combined with the real-time load changes of base station communication equipment and external environmental temperature fluctuations, and to construct a dynamic optimization model for air conditioning energy consumption through energy efficiency curves.

[0150] The adjustment module 303 is used to dynamically adjust the air conditioner's operating mode, fan speed, and temperature setpoint based on the predicted results of dynamic cooling demand and the energy consumption optimization model, using a reinforcement learning-based air conditioner operation strategy optimization algorithm. At the same time, it optimizes the energy storage scheduling strategy by combining the charging and discharging status of the energy storage unit, and generates a joint optimization scheme for air conditioner operation and energy storage scheduling.

[0151] The management module 304 is used to monitor the air conditioner's operating status in real time. It adopts a fault early warning model based on anomaly detection algorithm to identify potential air conditioner faults. Through an adaptive control mechanism, it dynamically adjusts the joint optimization scheme to ensure the stability and energy efficiency of the air conditioner's operation and generate the final base station air conditioner energy consumption management scheme.

[0152] As can be seen, multi-source data is collected in real time from temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room to generate a multi-source fusion dataset. Based on this dataset, dynamic cooling demand within future time windows is predicted, and a dynamic optimization model for air conditioning energy consumption is constructed using energy efficiency curves. According to the predicted dynamic cooling demand and the energy consumption optimization model, the air conditioning operation mode, fan speed, and temperature setpoints are dynamically adjusted, and energy storage scheduling strategies are optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling. Real-time monitoring of the air conditioning operation status and dynamic adjustment of the joint optimization scheme generate the final base station air conditioning energy consumption management scheme. This enables air conditioning energy consumption management to not only be real-time and accurate but also to achieve more efficient energy consumption control and energy storage scheduling optimization through intelligent means.

[0153] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0154] Specifically, in this embodiment, the storage medium can be configured to store a computer program for performing the following steps:

[0155] S201: Based on the temperature sensor, humidity sensor, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room, the data acquisition and preprocessing method based on edge computing is adopted to collect multi-source data in real time, and through adaptive filtering algorithm, environmental noise and electromagnetic interference are eliminated to generate multi-source fusion dataset;

[0156] S202, for multi-source fusion datasets, adopts a prediction model based on lightweight spatiotemporal neural networks, combines real-time load changes of base station communication equipment and external environmental temperature fluctuations to predict dynamic cooling demand within future time windows, and constructs a dynamic optimization model for air conditioning energy consumption through energy efficiency curves;

[0157] S203, based on the prediction results of dynamic cooling demand and energy consumption optimization model, adopts a reinforcement learning-based air conditioning operation strategy optimization algorithm to dynamically adjust the air conditioning operation mode, wind speed and temperature setpoint. At the same time, combined with the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

[0158] S204 monitors the air conditioner's operating status in real time, uses a fault early warning model based on anomaly detection algorithms to identify potential air conditioner faults, and dynamically adjusts the joint optimization scheme through an adaptive control mechanism to ensure the air conditioner's operational stability and energy efficiency, generating the final base station air conditioner energy consumption management scheme.

[0159] As can be seen, multi-source data is collected in real time from temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room to generate a multi-source fusion dataset. Based on this dataset, dynamic cooling demand within future time windows is predicted, and a dynamic optimization model for air conditioning energy consumption is constructed using energy efficiency curves. According to the predicted dynamic cooling demand and the energy consumption optimization model, the air conditioning operation mode, fan speed, and temperature setpoints are dynamically adjusted, and energy storage scheduling strategies are optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling. Real-time monitoring of the air conditioning operation status and dynamic adjustment of the joint optimization scheme generate the final base station air conditioning energy consumption management scheme. This enables air conditioning energy consumption management to not only be real-time and accurate but also to achieve more efficient energy consumption control and energy storage scheduling optimization through intelligent means.

[0160] This invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0161] Specifically, the aforementioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the aforementioned processor, and the input / output device is connected to the aforementioned processor.

[0162] Specifically, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0163] S201: Based on the temperature sensor, humidity sensor, air conditioning operation status monitoring equipment and communication equipment load data in the base station equipment room, the data acquisition and preprocessing method based on edge computing is adopted to collect multi-source data in real time, and through adaptive filtering algorithm, environmental noise and electromagnetic interference are eliminated to generate multi-source fusion dataset;

[0164] S202, for multi-source fusion datasets, adopts a prediction model based on lightweight spatiotemporal neural networks, combines real-time load changes of base station communication equipment and external environmental temperature fluctuations to predict dynamic cooling demand within future time windows, and constructs a dynamic optimization model for air conditioning energy consumption through energy efficiency curves;

[0165] S203, based on the prediction results of dynamic cooling demand and energy consumption optimization model, adopts a reinforcement learning-based air conditioning operation strategy optimization algorithm to dynamically adjust the air conditioning operation mode, wind speed and temperature setpoint. At the same time, combined with the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

[0166] S204 monitors the air conditioner's operating status in real time, uses a fault early warning model based on anomaly detection algorithms to identify potential air conditioner faults, and dynamically adjusts the joint optimization scheme through an adaptive control mechanism to ensure the air conditioner's operational stability and energy efficiency, generating the final base station air conditioner energy consumption management scheme.

[0167] As can be seen, multi-source data is collected in real time from temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room to generate a multi-source fusion dataset. Based on this dataset, dynamic cooling demand within future time windows is predicted, and a dynamic optimization model for air conditioning energy consumption is constructed using energy efficiency curves. According to the predicted dynamic cooling demand and the energy consumption optimization model, the air conditioning operation mode, fan speed, and temperature setpoints are dynamically adjusted, and energy storage scheduling strategies are optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling. Real-time monitoring of the air conditioning operation status and dynamic adjustment of the joint optimization scheme generate the final base station air conditioning energy consumption management scheme. This enables air conditioning energy consumption management to not only be real-time and accurate but also to achieve more efficient energy consumption control and energy storage scheduling optimization through intelligent means.

[0168] The above description, based on the embodiments shown in the figures, details the structure, features, and effects of the present invention. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for energy storage management of base station air conditioning, characterized in that, The method includes: Based on data from temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data within the base station equipment room, an edge computing-based data acquisition and preprocessing method is employed to collect multi-source data in real time. An adaptive filtering algorithm is used to eliminate environmental noise and electromagnetic interference, generating a multi-source fusion dataset. Specifically, an edge computing-based data acquisition framework is used, employing distributed data acquisition nodes to acquire multi-source data in real time. Each acquisition node is equipped with a lightweight data caching mechanism to ensure the real-time and continuous nature of data acquisition. For the acquired multi-source data, an adaptive filtering algorithm based on wavelet transform is used to separate environmental noise and electromagnetic interference from the signal. A dynamic threshold adjustment mechanism is used to perform differentiated filtering processing based on the noise characteristics of different sensor data, resulting in a pre-denoised multi-source dataset. For the denoised multi-source dataset, a time alignment algorithm based on dynamic time warping is used to eliminate timestamp differences between different sensor data. At the same time, a missing value imputation method based on spatiotemporal correlation is used to interpolate and fill in the missing parts of the data to generate a time-synchronized complete dataset. For the time-synchronized complete dataset, a feature fusion method based on multi-head attention mechanism is used to extract multi-dimensional features of temperature, humidity, air conditioning operation status and communication equipment load data. Through an adaptive feature weight allocation mechanism, a high-precision multi-source fusion dataset is generated. For multi-source fusion datasets, a prediction model based on lightweight spatiotemporal neural networks is adopted. Combining the real-time load changes of base station communication equipment and external environmental temperature fluctuations, the dynamic cooling demand within the future time window is predicted. And through the energy consumption efficiency curve, a dynamic optimization model of air conditioning energy consumption is constructed. Based on the prediction results of dynamic cooling demand and the energy consumption optimization model, a reinforcement learning-based air conditioning operation strategy optimization algorithm is adopted to dynamically adjust the air conditioning operation mode, fan speed and temperature setpoint. At the same time, combined with the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling. The system monitors the air conditioning operation status in real time, uses a fault early warning model based on anomaly detection algorithm to identify potential air conditioning faults, and dynamically adjusts the joint optimization scheme through an adaptive control mechanism to ensure the stability and energy efficiency of air conditioning operation, thus generating the final base station air conditioning energy consumption management scheme.

2. The method according to claim 1, characterized in that, The multi-source fusion dataset employs a prediction model based on a lightweight spatiotemporal neural network, combining real-time load changes of base station communication equipment and external environmental temperature fluctuations to predict dynamic cooling demand within future time windows. Furthermore, a dynamic optimization model for air conditioning energy consumption is constructed using energy efficiency curves, including: For the multi-source fusion dataset, a prediction model based on a lightweight spatiotemporal neural network is adopted. The temporal features of the load change of base station communication equipment are extracted through the temporal convolutional layer, and the local spatial features of the temperature distribution in the equipment room are captured through the spatial graph convolutional layer to generate a spatiotemporal feature representation. By combining external environmental temperature fluctuation data, an attention-based external environment fusion technique is adopted to weightedly fuse external temperature change characteristics with spatiotemporal feature representations, thereby enhancing the predictive model's ability to predict dynamic cooling demand and generating an environmentally enhanced spatiotemporal feature representation. The spatiotemporal features of the enhanced environment are input into the output layer of the prediction model. A time series decoder based on multi-step prediction is used to predict the dynamic cooling demand of base station equipment rooms within future time windows and generate cooling demand prediction results. Based on historical operating data of air conditioners, an energy efficiency curve fitting method based on piecewise linear regression is used to construct a relationship model between air conditioner energy consumption, cooling capacity, and operating mode. Combined with the cooling demand prediction results, a dynamic optimization model for air conditioner energy consumption is generated.

3. The method according to claim 2, characterized in that, Based on the predicted dynamic cooling demand and energy consumption optimization model, a reinforcement learning-based air conditioning operation strategy optimization algorithm is used to dynamically adjust the air conditioning operation mode, fan speed, and temperature setpoint. Simultaneously, considering the charging and discharging status of the energy storage unit, the energy storage scheduling strategy is optimized to generate a joint optimization scheme for air conditioning operation and energy storage scheduling, including: Based on the dynamic cooling demand prediction results and energy consumption optimization model, an air conditioning operation strategy optimization algorithm based on deep reinforcement learning is adopted to construct a reward function with the goal of minimizing energy consumption and satisfying cooling demand. Through a dual Q network structure, the air conditioning operation mode, fan speed and temperature setpoint are dynamically adjusted to generate a preliminary air conditioning operation optimization strategy. By combining the current state of charge of the energy storage unit, the charge and discharge efficiency curves, and the grid electricity price fluctuation data, a multi-dimensional constraint model of the energy storage state is constructed. Through energy conservation constraints, charge and discharge rate constraints, and grid load balance constraints, the feasible solution space for energy storage scheduling is defined. For the energy storage state constraint model, an energy storage scheduling method based on distributed constraint optimization algorithm is adopted. Combined with the energy consumption demand in the air conditioning operation optimization strategy, the charging and discharging plan of the energy storage unit is dynamically adjusted. Through time-segmented electricity price sensitivity analysis, the energy storage scheduling strategy is optimized and a preliminary energy storage scheduling scheme is generated. The air conditioning operation optimization strategy and energy storage scheduling scheme are optimized in a coordinated manner. A joint optimization framework based on game theory is adopted. Through the dynamic game between the air conditioning and energy storage systems, the energy consumption cost and the cooling demand satisfaction are balanced to generate a joint optimization scheme for air conditioning operation and energy storage scheduling.

4. The method according to claim 3, characterized in that, The system performs real-time monitoring of the air conditioning's operating status, employs a fault early warning model based on anomaly detection algorithms to identify potential air conditioning faults, and dynamically adjusts the joint optimization scheme through an adaptive control mechanism to ensure the air conditioning's operational stability and energy efficiency, generating the final base station air conditioning energy consumption management scheme, including: Real-time monitoring of air conditioner operation status; collection of air conditioner operation parameters and energy consumption data; processing of collected data in real time through edge computing nodes; generation of time series dataset of air conditioner operation status. For time series datasets, an anomaly detection algorithm based on a combination of isolated forest and long short-term memory network is used to identify potential air conditioner faults and generate fault warning signals through a dynamic threshold adjustment mechanism. Based on the fault warning signal, an adaptive control mechanism based on fuzzy logic is adopted to dynamically adjust the joint optimization scheme. Through a multi-objective optimization model, the fault recovery cost and energy efficiency are balanced to generate an adaptive control strategy. The adaptive control strategy is integrated with the joint optimization scheme, and a dynamic adjustment method based on feedback correction mechanism is adopted to update the air conditioning operating parameters and energy storage scheduling plan in real time, thereby generating the final base station air conditioning energy consumption management scheme.

5. A base station air conditioning energy consumption management and energy storage system, characterized in that, The system includes: The data acquisition module is used to collect multi-source data in real time based on temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room. It employs an edge computing-based data acquisition and preprocessing method, and uses an adaptive filtering algorithm to eliminate environmental noise and electromagnetic interference, generating a multi-source fusion dataset. Specifically, based on the temperature sensors, humidity sensors, air conditioning operation status monitoring equipment, and communication equipment load data in the base station equipment room, an edge computing-based data acquisition framework is used. Distributed data acquisition nodes acquire multi-source data in real time, with each node equipped with a lightweight data caching mechanism to ensure the real-time and continuous nature of data acquisition. For the acquired multi-source data, an adaptive filtering algorithm based on wavelet transform is used to separate environmental noise and electromagnetic interference from the signal. A dynamic threshold adjustment mechanism is used to perform differentiated filtering processing based on the noise characteristics of different sensor data, resulting in a pre-denoised multi-source dataset. For the denoised multi-source dataset, a time alignment algorithm based on dynamic time warping is used to eliminate timestamp differences between different sensor data. At the same time, a missing value imputation method based on spatiotemporal correlation is used to interpolate and fill in the missing parts of the data to generate a time-synchronized complete dataset. For the time-synchronized complete dataset, a feature fusion method based on multi-head attention mechanism is used to extract multi-dimensional features of temperature, humidity, air conditioning operation status and communication equipment load data. Through an adaptive feature weight allocation mechanism, a high-precision multi-source fusion dataset is generated. The prediction module is used to predict the dynamic cooling demand within a future time window by using a prediction model based on a lightweight spatiotemporal neural network on a multi-source fusion dataset, combined with the real-time load changes of base station communication equipment and external environmental temperature fluctuations. It also constructs a dynamic optimization model for air conditioning energy consumption through energy efficiency curves. The adjustment module is used to dynamically adjust the air conditioner's operating mode, fan speed, and temperature setpoint based on the predicted results of dynamic cooling demand and the energy consumption optimization model, using a reinforcement learning-based air conditioner operation strategy optimization algorithm. At the same time, it optimizes the energy storage scheduling strategy by combining the charging and discharging status of the energy storage unit, and generates a joint optimization scheme for air conditioner operation and energy storage scheduling. The management module is used to monitor the air conditioner's operating status in real time. It adopts a fault early warning model based on anomaly detection algorithm to identify potential air conditioner faults. Through an adaptive control mechanism, it dynamically adjusts the joint optimization scheme to ensure the stability and energy efficiency of the air conditioner's operation and generate the final base station air conditioner energy consumption management scheme.

6. The system according to claim 5, characterized in that, The prediction module is specifically used for: For the multi-source fusion dataset, a prediction model based on a lightweight spatiotemporal neural network is adopted. The temporal features of the load change of base station communication equipment are extracted through the temporal convolutional layer, and the local spatial features of the temperature distribution in the equipment room are captured through the spatial graph convolutional layer to generate a spatiotemporal feature representation. By combining external environmental temperature fluctuation data, an attention-based external environment fusion technique is adopted to weightedly fuse external temperature change characteristics with spatiotemporal feature representations, thereby enhancing the predictive model's ability to predict dynamic cooling demand and generating an environmentally enhanced spatiotemporal feature representation. The spatiotemporal features of the enhanced environment are input into the output layer of the prediction model. A time series decoder based on multi-step prediction is used to predict the dynamic cooling demand of base station equipment rooms within future time windows and generate cooling demand prediction results. Based on historical operating data of air conditioners, an energy efficiency curve fitting method based on piecewise linear regression is used to construct a relationship model between air conditioner energy consumption, cooling capacity, and operating mode. Combined with the cooling demand prediction results, a dynamic optimization model for air conditioner energy consumption is generated.

7. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method of any one of claims 1-4 when it is run.

8. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method of any one of claims 1-4.

Images