Intelligent agent intelligent control method, system and equipment for unattended air conditioner and storage medium

Abstract

The invention provides an intelligent agent intelligent control method, system and equipment for an unattended air conditioner and a storage medium, and relates to the technical field of intelligent air conditioner control. The method comprises the steps that operation data of all target air conditioners are obtained, and the operation data comprise environment parameters, energy consumption data and equipment state data; performing edge calculation processing on the operation data to obtain time sequence characteristics, and uploading the time sequence characteristics to a cloud platform; constructing an air conditioner control agent and an air conditioner working condition agent based on the time sequence characteristics; an air conditioner control strategy is generated according to the air conditioner control agent, and all the target air conditioners are controlled to operate according to the air conditioner control strategy; and the health states of the target air conditioners are analyzed based on the air conditioner working condition intelligent agents, and operation reports of all the target air conditioners are generated and visually displayed. By implementing the technical scheme provided by the invention, efficient management of the air conditioning system can be realized in an unattended scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of air conditioners, and in particular to an intelligent agent intelligent control method, system, device and storage medium for unattended air conditioners. BACKGROUND

[0002] With the acceleration of urbanization, the number of elevator machine rooms and outdoor unattended machine rooms is growing rapidly. These places are usually equipped with split air conditioners or semiconductor air conditioners to maintain the environmental temperature required for equipment operation. Due to the particularity of the place, these air conditioning systems often adopt an all-weather operation mode, i.e., they are turned on continuously for 24 hours in the high-temperature season and cannot be flexibly adjusted according to the actual temperature demand.

[0003] Currently, for air conditioner control in unattended scenarios, the industry mainly uses smart home products for management. For example, fixed on-off times are set through a timed socket, or simple constant temperature control is performed using a smart socket with a temperature sensor. Some places also use 4G smart sockets that support remote operation to realize remote switching and temperature adjustment of air conditioners.

[0004] However, in actual applications, especially in scenarios with a large management scale (such as multiple elevator machine rooms in residential complexes), the existing control schemes have obvious limitations. Since each air conditioner needs to be independently configured and managed, the operation and maintenance work is tedious and inefficient. At the same time, this decentralized management method makes it difficult to collect and analyze air conditioner operation data uniformly, and it is difficult to discover and prevent possible abnormalities in equipment in a timely manner, affecting the service life and operating efficiency of air conditioning equipment.

[0005] Therefore, how to realize efficient management of air conditioning systems in unattended scenarios has become a technical problem that needs to be solved at present. SUMMARY

[0006] The present application provides an intelligent agent intelligent control method, system, device and storage medium for unattended air conditioners, which can realize efficient management of air conditioning systems in unattended scenarios.

[0007] In a first aspect of the present application, an intelligent agent intelligent control method for unattended air conditioners is provided, which specifically comprises:

[0008] Obtaining operation data of each target air conditioner, the operation data including environmental parameters, energy consumption data and equipment state data;

[0009] Performing edge computing processing on the operation data to obtain time series features, and uploading the time series features to a cloud platform;

[0010] Constructing an air conditioner control intelligent agent and an air conditioner working condition intelligent agent based on the time series features;

[0011] generate an air conditioner control strategy according to the air conditioner control agent, and control each of the target air conditioners to operate according to the air conditioner control strategy;

[0012] analyze the health status of each of the target air conditioners based on the air conditioner working condition agent, generate an operation report of each of the target air conditioners, and visually display the operation report.

[0013] By adopting the technical solution, the application can automatically acquire operation data of each target air conditioner and perform edge computing processing, and upload the processed time sequence features to a cloud platform, thereby realizing unified collection and preprocessing of air conditioner operation data. The air conditioner control agent constructed based on the time sequence features can adaptively generate a control strategy according to actual environmental requirements, thereby avoiding energy waste caused by continuous operation of the air conditioner in the traditional scheme. Meanwhile, the air conditioner working condition agent can analyze the health status of the target air conditioner and generate an operation report, so that an operation and maintenance personnel can timely master equipment abnormal conditions and effectively prevent faults. The scheme breaks through the limitation of the traditional decentralized management mode, realizes centralized management and intelligent control of unattended air conditioners, significantly improves the operation efficiency and equipment reliability of the air conditioner system, and realizes efficient management of the air conditioner system.

[0014] In a second aspect of the application, an agent intelligent control system of an unattended air conditioner is provided, which specifically comprises:

[0015] a data acquisition module configured to acquire operation data of each target air conditioner, the operation data including environmental parameters, energy consumption data, and equipment state data;

[0016] a data processing module configured to perform edge computing processing on the operation data to obtain time sequence features, and upload the time sequence features to a cloud platform;

[0017] an agent construction module configured to construct an air conditioner control agent and an air conditioner working condition agent based on the time sequence features;

[0018] a strategy generation module configured to generate an air conditioner control strategy according to the air conditioner control agent, and control each of the target air conditioners to operate according to the air conditioner control strategy;

[0019] a report generation module configured to analyze the health status of each of the target air conditioners based on the air conditioner working condition agent, generate an operation report of each of the target air conditioners, and visually display the operation report.

[0020] In a third aspect of the application, an electronic device is provided, which includes a processor, a memory, a user interface, and a network interface. The memory is configured to store instructions, the user interface and the network interface are both configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method described in any one of the above aspects.

[0021] In a fourth aspect of the present application, a computer-readable storage medium is provided, the computer-readable storage medium storing instructions which, when executed, perform the method of any one of the above. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is a system architecture schematic diagram of an intelligent agent intelligent control system of an unattended air conditioner provided by an embodiment of the present application;

[0023] Figure 2 is a flow schematic diagram of an intelligent agent intelligent control method of an unattended air conditioner provided by an embodiment of the present application;

[0024] Figure 3 is a structural schematic diagram of an intelligent agent intelligent control system of an unattended air conditioner provided by an embodiment of the present application;

[0025] Figure 4 is a structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0026] The reference signs are explained as follows: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0027] In order for those skilled in the art to better understand the technical solutions in the specification, the technical solutions in the specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the specification. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0028] In the description of the embodiments of the present application, the words such as "for example" or "for instance" are used to represent an example, illustration or description. Any embodiment or design scheme described as "for example" or "for instance" in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words such as "for example" or "for instance" are intended to present the relevant concept in a specific manner.

[0029] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined with "first", "second" can be explicitly or implicitly included one or more of the features. The terms "include", "contain", "have" and their variants mean "include but not limited to", unless otherwise specifically emphasized.

[0030] Figure 1 A system architecture diagram of an intelligent agent control system of an unattended air conditioner is shown.

[0031] As Figure 1 shown, the system takes "air conditioner control all-in-one machine" as the core hardware carrier, integrates cutting-edge technologies such as Internet of Things, big data, artificial intelligence, and builds a complete architecture system from bottom-layer device access to top-layer intelligent agent control. The architecture mainly includes three levels of edge layer, cloud platform layer and application layer.

[0032] Edge layer: Deploy a dedicated "air conditioner control all-in-one machine" hardware on each unattended target air conditioner, which contains a temperature and humidity sensor to accurately measure the real-time temperature and humidity of the air conditioner operating environment and the refrigeration and heating area; a current sensor to monitor the air conditioner operating current to evaluate the equipment load and energy consumption state; a noise sensor to monitor the environmental noise for cloud-based analysis of working condition information; an infrared emitter head to control the air conditioner on-off, set the operating mode, set the wind speed and temperature, etc.; a 4G communication module to communicate with the "cloud platform layer" stably and efficiently through the Internet of Things communication protocol; an edge computing module to undertake key tasks such as data aggregation, edge computing and instruction forwarding. On the one hand, it quickly receives massive data from sensors, and performs preliminary cleaning, screening and analysis on the data, eliminates abnormal data, and integrates valid information; on the other hand, based on the locally deployed lightweight data analysis algorithm, it quickly analyzes the real-time data, such as energy consumption trend calculation, operating state rationality verification, etc.

[0033] Cloud platform layer: Relying on strong cloud computing and big data storage capabilities, a data storage center, an intelligent agent analysis engine, and a centralized monitoring platform are built. In terms of data storage, a distributed time series database technology is used to safely and reliably store all air conditioning equipment historical operation data, energy consumption data, environmental data, and system logs, etc., providing data support for in-depth analysis; a distributed relational database and a real-time database are used to establish a shadow model for the equipment, store the latest attributes and control interfaces of the equipment, and facilitate centralized monitoring platform to call for control. The intelligent agent analysis engine integrates intelligent agents based on deep learning algorithms and solver technology, such as "air conditioning control intelligent agent" and "air conditioning working condition intelligent agent", which deeply mine and analyze the uploaded data, accurately identify energy consumption abnormal equipment, predict potential fault risks, and generate optimized intelligent control strategies. The centralized monitoring platform provides a unified visual operation interface for users, supporting users to remotely monitor the running status and energy consumption of all air conditioning equipment through various terminal devices such as PC and mobile terminal, and flexibly perform manual control operations to view detailed equipment operation reports and analysis results.

[0034] Application layer: Different application programs are provided for different user roles and use scenarios. For operation and maintenance personnel, a professional operation and maintenance APP is provided, which supports functions such as real-time reception of device alarm information, viewing of device detailed parameters and historical operation records, remote initiation of device maintenance tasks, etc., greatly improving operation and maintenance efficiency; for energy consumption management personnel, an energy consumption analysis and optimization platform is provided, which displays the air conditioning energy consumption distribution of each region and each period through visual charts, deeply analyzes the energy consumption composition and change trend, and assists in formulating targeted energy saving and consumption reduction strategies; for ordinary users (if there are relevant permissions), a simple and easy-to-use user-side APP is provided, which can realize real-time viewing of environmental temperature, autonomous setting of air conditioning temperature range, and other basic operations, improving user experience.

[0035] This layered architecture design realizes closed-loop management from data collection, intelligent analysis to control execution, fully utilizes the advantages of edge computing and cloud computing, and provides reliable technical support for intelligent management of unattended air conditioners.

[0036] The present application provides an intelligent agent intelligent control method for an unattended air conditioner, referring to Figure 2 , Figure 2 is a flowchart of an intelligent agent intelligent control method for an unattended air conditioner provided by an embodiment of the present application, comprising steps 10 to 50, and the above steps are as follows:

[0037] Step 10: Obtain the running data of each target air conditioner, including environmental parameters, energy consumption data, and device state data.

[0038] Among them, in the embodiments of the present application, the target air conditioner refers to the air conditioning equipment installed in the unattended scenes such as elevator machine room, building machine room and base station machine room, which can include split air conditioner and semiconductor air conditioner, etc.

[0039] The operation data refers to various types of data collected by the "air conditioner intelligent control all-in-one machine", which specifically includes environmental parameters: real-time temperature and humidity data of the air conditioning operating environment and the refrigeration and heating area collected by the temperature and humidity sensor;

[0040] Energy consumption data: air conditioning operating current, power and other power information monitored by the current sensor; equipment state data: including the on-off state, operating mode, set temperature, environmental noise and other data reflecting the working state of the equipment of the air conditioner.

[0041] Specifically, since the target air conditioner is deployed in the unattended scenes such as elevator machine room, building machine room and base station machine room, it is necessary to fully grasp the running state of the air conditioner to realize intelligent management and control, therefore the operation data is collected by the air conditioner intelligent control all-in-one machine. The air conditioner intelligent control all-in-one machine collects environmental parameters through a temperature and humidity sensor, including real-time temperature and humidity data of the air conditioning operating environment and the refrigeration and heating area, with a sampling frequency of 1-5 minutes / time to ensure the real-time nature of the environmental data; collects energy consumption data through a current sensor, including air conditioning operating current and real-time power, with a sampling frequency of 1 minute / time for evaluating the power consumption condition of the equipment; collects equipment state data through a noise sensor combined with the on-off state, operating mode, set temperature and other data of the air conditioner, realizing comprehensive monitoring of the working state of the air conditioner. The obtained operation data is preprocessed by the edge computing module of the air conditioner intelligent control all-in-one machine, including data cleaning, outlier filtering and format conversion and other operations to ensure the accuracy and effectiveness of the data. At the same time, a data active acquisition mechanism is set: when the temperature and humidity change exceeds 1 degree or the power change exceeds 2%, a new round of data acquisition is triggered immediately. Through the above operation data acquisition scheme, the system can fully grasp the real-time running state of each target air conditioner, providing accurate data basis for subsequent data analysis and control strategy generation.

[0042] Step 20: performing edge computing processing on the operation data to obtain time series features, and uploading the time series features to the cloud platform.

[0043] Among them, in the embodiments of the present application, the time series features refer to the characteristic values obtained by statistical analysis of the air conditioner operation data according to the time sequence. It can include but not limited to: the fluctuation amplitude of the air conditioner operating temperature curve, the peak-valley change trend of the energy consumption curve, the periodic change law of the environmental temperature and humidity, the time distribution characteristics of the equipment state data, etc. These time series features reflect the running state and performance of the air conditioner at different time periods, and are important basis for the health degree evaluation and fault early warning of the air conditioner working condition intelligent agent. By analyzing the change trend of the time series features, the system can timely find the air conditioner running abnormities and predict potential fault risks

[0044] Specifically, in order to reduce the computing load of the cloud platform and improve the data transmission efficiency, the edge computing module of the air conditioner intelligent control integrated machine is used to preprocess the obtained operation data and extract the time sequence features. The edge computing module performs statistical analysis on the temperature curve to obtain the temperature fluctuation amplitude feature, calculates the energy consumption curve to obtain the energy consumption peak-valley change feature, processes the environmental temperature and humidity data to obtain the periodic change feature, and analyzes the equipment state data to obtain the time distribution feature. The processed time sequence features are encrypted and transmitted to the cloud platform through the 4G communication module using the MQTT protocol. By extracting the time sequence features on the edge side, on the one hand, the amount of transmitted data is reduced, and on the other hand, the cloud platform provides preprocessed structured data for intelligent analysis, improving the overall operation efficiency of the system.

[0045] As an optional embodiment, the step of performing edge computing processing on the operation data to obtain the time sequence features and uploading the time sequence features to the cloud platform can further include steps 101-105:

[0046] Step 101: Data cleaning is performed on the environmental parameters, energy consumption data and equipment state data in the operation data, including removing outliers, supplementing missing values and data smoothing.

[0047] Specifically, in order to ensure the data quality for subsequent analysis, first, the operation data is subjected to data cleaning. The system detects outliers in the temperature and humidity data in the environmental parameters, marks and deletes data that exceeds the normal range (temperature -20℃ to 60℃, humidity 0% to 100%) as outliers; filters data that appears in the energy consumption data zero value or exceeds 150% of the rated power of the air conditioner; baseline calibration is performed on the noise data in the equipment state data to eliminate environmental noise interference. For time points with missing data, the system uses linear interpolation to supplement, i.e. based on the effective data before and after the missing point to calculate the interpolation. Finally, a sliding average algorithm is used to smooth all data sequences, with a window size of 5 sampling points to eliminate the influence of short-term random fluctuations. Through data cleaning, the data reliability for subsequent analysis is significantly improved.

[0048] Step 102: Normalization processing is performed on the cleaned environmental parameters to obtain the environmental parameter change curve.

[0049] Specifically, to make the environmental parameters of different dimensions comparable, the cleaned environmental parameters are normalized. The system applies the maximum-minimum value normalization method to the temperature and humidity data, mapping the data to the [0, 1] interval. The calculation formula is: normalized value = (original value - minimum value) / (maximum value - minimum value), where the maximum value and the minimum value are obtained based on the data statistics of the past 24 hours. For the normalized temperature and humidity data, the system reconstructs them in chronological order to generate the environmental parameter change curve. This curve retains the timing characteristics of temperature and humidity changes while eliminating the influence of dimensional differences, facilitating subsequent feature extraction and pattern recognition.

[0050] Step 103: Perform sliding window analysis on the cleaned energy consumption data to extract energy consumption fluctuation features and obtain the energy consumption change curve.

[0051] Specifically, to extract the dynamic features of energy consumption data, the system uses the sliding window analysis method. The sliding window size is set to 30 minutes, and the step size is 5 minutes. The cleaned energy consumption data is processed window by window. In each window, statistical features such as average power, maximum power, and power variance are calculated; at the same time, the energy consumption change rate between adjacent windows is calculated to identify energy consumption mutation points. The system organizes these feature values in chronological order to construct the energy consumption change curve. This curve not only reflects the overall trend of air conditioner energy consumption but also effectively captures abnormal fluctuations in energy consumption, providing a basis for subsequent energy efficiency analysis and anomaly detection.

[0052] Step 104: Perform Fourier transform on the cleaned equipment state data to extract noise spectrum features and obtain the state feature curve.

[0053] Specifically, to analyze the equipment running state in depth, the system performs frequency domain analysis on the cleaned equipment state data. Fast Fourier Transform (FFT) is applied to the noise data with a sampling frequency of 1 kHz, converting the time domain signal into frequency domain features. In the frequency domain, the system extracts the energy distribution of the main frequency components (0-200 Hz), calculates feature parameters such as spectral centroid and frequency band energy ratio. At the same time, combined with discrete state data such as the on-off state and operating mode of the air conditioner, the state feature curve is constructed. This curve comprehensively reflects the stability and potential abnormalities of equipment operation, enabling early detection of mechanical failure signs.

[0054] Step 105: Align and combine the environmental parameter change curve, the energy consumption change curve, and the state feature curve according to the time stamp to form the time series features.

[0055] Specifically, the system synchronizes the environmental parameter change curve, the energy consumption change curve and the state feature curve based on a unified timestamp (accurate to seconds). For data with inconsistent sampling frequencies, the system aligns them to the same time point through an interpolation method. Then, the system organizes the aligned multi-dimensional features according to a predefined data structure to generate time sequence features in a standard format. The time sequence features contain all-round information of the air conditioner operation and provide structured input data for the analysis and decision of the air conditioner working condition agent.

[0056] Step 30: constructing an air conditioner control agent and an air conditioner working condition agent based on the time sequence features.

[0057] In the embodiments of the present application, the air conditioner control agent refers to a core algorithm model for realizing intelligent management and control of an unattended air conditioner group. The agent is integrated in a cloud platform, adopts a deep reinforcement learning framework, and generates an optimal control strategy in real time based on historical operation data, working condition features and user demands through modules such as air conditioner action space construction, time sequence reward calculation and air conditioner policy network. The air conditioner control agent automatically adjusts the start and stop time, operation mode and target temperature of the air conditioner according to the time sequence features, ensures that the machine room environment indicators meet the equipment operation requirements, and realizes the goal of energy saving and consumption reduction.

[0058] The air conditioner working condition agent refers to an intelligent algorithm model for evaluating and predicting the health status of the air conditioner. The agent is deployed on a cloud platform and analyzes the time sequence features in the air conditioner operation process through functional modules such as air conditioner anomaly detection, health degree evaluation and fault early warning. Based on the air conditioner energy efficiency attenuation curve, component wear feature and fault mode library, the air conditioner working condition agent can calculate the equipment health degree index in real time and provide early warning for potential fault risks, thereby providing data support for equipment operation and maintenance decisions.

[0059] As an optional embodiment, the step of constructing an air conditioner control agent based on the time sequence features can further include steps 201-202:

[0060] Step 201: obtaining time-aligned training parameters based on the time sequence features, the training parameters including a state space, an action space and a reward function, wherein the state space includes a current temperature, a target temperature, a power load and an outdoor temperature, the action space includes an air conditioner switch state, a temperature gear and a running time length, and the reward function is determined in combination with a temperature comfort degree and an energy consumption index.

[0061] Specifically, to enable the air conditioning control agent to break through the limitations of traditional indoor closed-loop control, the system first builds a multi-source data fusion data acquisition system. At the environmental perception level, the air conditioning intelligent control all-in-one machine collects basic data such as indoor temperature and humidity (1 minute / time), equipment operating status (5 minutes / time), and energy consumption data (15 minutes / time); At the same time, access to multiple authoritative weather APIs (such as the Meteorological Bureau), obtain fine weather forecasts for the next 24-72 hours, including hourly temperature prediction, maximum / minimum temperature occurrence period, precipitation probability, cloud cover, solar radiation intensity, gale warning, etc. The system performs multiple preprocessing on the collected data: abnormal value detection (temperature -20℃ to 60℃, humidity 0-100%) and smoothing processing (30-minute moving average) on temperature and humidity data; Weather forecast data confidence assessment (based on historical forecast accuracy) and multi-source cross-validation, when the difference between different sources exceeds the set threshold (such as temperature difference > 3℃), trigger manual review. In the state space construction, the system designs four dimensions of features: 1) Environmental state dimension, including indoor temperature and humidity, set temperature, outdoor temperature and its 24-hour prediction value, weather conditions, etc.; 2) Equipment state dimension, including operating mode, wind speed gear, real-time power ratio, cumulative energy consumption, equipment health index (such as compressor current stability), etc.; 3) Time dimension, including hour number (sine cosine encoding), date type, season information, historical same period data, etc.; 4) Scene feature dimension, including scene type, load prediction, passenger flow density (indirectly estimated through access control, elevator, etc.), special event marking, etc. The system maps each dimension of features to the [0, 1] interval through min-max normalization, and maintains a 24-hour time window. In the action space design, fine-grained control strategies are realized, including temperature fine-tuning (±0.5℃ / 1.0℃), operating mode switching (considering scene characteristics for intelligent switching), wind speed control (based on temperature difference for dynamic adjustment), and time sequence control (considering weather forecast for pre-control decision), while introducing multiple constraints: hard constraints include temperature safety boundaries (computer room 15-30℃), minimum start-stop protection time (10 minutes), etc.; Soft constraints include energy consumption limits, comfort requirements, etc. In the reward function design, a multi-objective weighted combination is adopted: 1) Temperature control reward Rt, based on the deviation between the current temperature and the target temperature, the penalty weight of the computer room scene for temperature over-limit (0.8) is greater than that for under-limit (0.4); 2) Energy consumption control reward Re, considering power ratio, peak-valley electricity price and predicted weather conditions, additional rewards are obtained by actively reducing energy consumption before high temperature periods are predicted; 3) Behavior smoothing reward Rs, which punishes frequent mode switching, wind speed and start-stop; 4) Predictive control reward Rp, which evaluates the effect of advance control and gives positive incentives to pre-control behaviors such as "advance cooling to create buffer". The weights of each reward are dynamically adjusted according to the scene type and time period, such as in the high temperature warning period of the computer room scene, the temperature control weight is increased to 0.7.

[0062] Step 202: training the preset initial air conditioner control agent based on the training parameters until the training meets a preset stop condition, to obtain a trained air conditioner control agent, and the preset stop condition is that the calculation result of the reward function after continuous multiple training iterations is greater than a preset threshold.

[0063] Specifically, in order to enable the air conditioner control agent to have adaptability to complex environments, the system adopts a double-track training mechanism for agent training. In the pre-training phase, the system first constructs a four-layer neural network structure based on a deep Q network (DQN), including an input layer matching the dimension of the state vector, two hidden layers (containing 256 and 128 neurons respectively, using ReLU activation function) and an output layer corresponding to the dimension of the action space. In order to improve the generalization ability of the model, a Dropout layer (dropout rate 0.2) and a Batch Normalization layer are added after the hidden layer. The system adopts a priority-based experience replay mechanism, sets an experience pool with a capacity of 10000, stores five-tuple data (state, action, reward, next state, termination flag), and dynamically adjusts the sampling weight according to the TD error size. On the exploration strategy, a scene-adaptive ε-greedy method is adopted: the initial exploration rate is 0.9, and the exploration rate is adjusted according to ε = max(0.1, 0.9 * 0.995 nThe decay strategy of the reinforcement learning algorithm is gradually reduced to 0.1, and the exploration rate is dynamically adjusted in different scenarios and time periods, such as reducing the exploration rate to 0.05 during the high-temperature early warning period in the machine room to ensure control safety. The system introduces a constrained reinforcement learning mechanism, which directly encodes temperature safety boundaries, device protection time, and other hard constraints into the action mask, and imposes high penalties on behaviors that violate soft constraints (such as frequent start-stop). To cope with non-stationary environments, the system implements concept drift detection, which monitors changes in temperature distribution and load characteristics to identify seasonal transition points and trigger rapid relearning or switch to pre-trained seasonal sub-strategies. In terms of training protection mechanisms, the system implements multi-level fusing: when sensor data is abnormal, communication is interrupted, or prediction confidence is below a threshold (0.85), it automatically switches to a conservative strategy; a complete device protection mechanism is set, including a minimum start-stop interval (10 minutes) and compressor protection logic. During the online fine-tuning phase, the system uses a small learning rate (0.0001) for incremental learning and ensures the safety of policy updates through grayscale release (first run in low-risk periods). During training, performance is evaluated every 100 steps, with 20 episodes (24 hours per episode) running, and when the average reward value of the last 20 evaluations exceeds 0.85 or the reward improvement amplitude of the last 50 evaluations is less than 0.01, training is stopped. The system uses visualization tools to display various indicators in real time during the training process, including temperature control accuracy, energy optimization effect, and policy stability, and saves complete decision logs for auditing and optimization. The trained air conditioning control agent can adaptively adjust the control strategy based on weather forecasts and scene characteristics, such as pre-cooling to build a buffer before a high-temperature period and increasing the set temperature slightly during rainy days to reduce energy consumption, achieving the control goal of "preparing in advance and avoiding overdoing."

[0064] As an optional embodiment, the step of constructing an air conditioner working condition intelligent agent based on time sequence characteristics can further include steps 301-305:

[0065] Step 301: Extract key features from time sequence characteristics as independent variables, and construct a regression model with air conditioner health index and energy consumption index as dependent variables.

[0066] Specifically, the system builds a complete data collection and feature extraction framework. First, a multi-source data collection system is established: real-time collection of temperature curves (15-minute sampling), energy consumption curves, environmental temperature and humidity (±0.5℃ / ±3%RH accuracy), equipment noise, etc. from the equipment perception layer; synchronizing device archives, operation history, maintenance records, fault mode library, etc. from the cloud platform historical database; obtaining real-time weather data and extreme weather warnings through API interface; retrieving technical parameters, maintenance standards and diagnostic knowledge from the knowledge base. In the feature engineering link, the system extracts multi-dimensional key features as independent variables: temperature control features (including temperature fluctuation amplitude, temperature change rate ℃ / min, temperature stability index), energy consumption features (energy consumption peak-valley ratio, 24h trend coefficient, unit refrigeration energy consumption kWh / cooling ton), environmental correlation features (temperature and humidity influence coefficient, weather operation time length proportion, load characteristic ratio). At the same time, a complete dependent variable system is designed: health degree indicators (failure rate 0-1, component wear 0-100, stability 0-1, maintenance demand 0-100) and energy consumption indicators (rationality 0-100, over-standard risk 0-1, optimization potential 0-100, energy saving priority 1-5). A deep learning multiple regression model is built based on these variables: 3-4 layers of hidden layers (64-32-16 neuron configuration), using ReLU activation function and Dropout (0.2) to control overfitting, and L2 regularization (0.01) to constrain the weight size.

[0067] Step 302: Train the regression model based on the historical data of each target air conditioner, establish the corresponding relationship between the independent variables and the dependent variables, and obtain the health degree evaluation model.

[0068] Specifically, the system trains the regression model based on historical data to establish the quantitative relationship between features and health degree indicators. In the data preparation stage, a scientific data division strategy (training set 70%, validation set 15%, test set 15%) is adopted, and all features are standardized. In the model training process, the device failure rate prediction is based on survival analysis method, considering device usage time, environmental conditions, maintenance history, etc.; the component wear degree is quantitatively evaluated by analyzing the running parameter changes of compressor, fan, heat exchanger and other key components; the system stability is comprehensively evaluated by the stability of temperature, pressure, vibration and other parameters; the maintenance demand degree prediction considers the device health status, usage intensity, environmental conditions and other factors. The mean square error loss function and Adam optimizer are used, the initial learning rate is set to 0.001, and the learning rate decay strategy is used to improve the convergence accuracy. The hyperparameters are optimized through grid search, and the early stopping mechanism is implemented to prevent overfitting, and the training is automatically stopped when the validation set loss does not decrease for multiple epochs in a row, and the health degree evaluation model is obtained.

[0069] Step 303: Using clustering analysis method to classify target air conditioners with similar environmental parameter change curves and energy consumption change curves.

[0070] Specifically, the system implements a multi-dimensional clustering analysis method. In the data preprocessing stage, wavelet transform is used to remove high-frequency noise and achieve time series data standardization. In terms of similarity calculation: the DTW algorithm is used to process the time series similarity of temperature and energy consumption curves of different lengths; the Euclidean distance is used to calculate the geometric distance of environmental parameters (room temperature, humidity, air pressure); the cosine similarity is used to evaluate the similarity of operating modes (load characteristics, start-stop characteristics). The similarity fusion adopts a dynamic weight method based on the importance of each index. The clustering algorithm selects DBSCAN, and the key parameters include the minimum sample size (minPts=5) and the neighborhood radius (determined by K-distance plot). The clustering effect is evaluated by multiple indicators: the silhouette coefficient and Calinski-Harabasz index are used to verify the effect; the Bootstrap method is used to analyze the consistency of the results; the intra-class and inter-class distances are calculated to evaluate the clustering quality. In this way, accurate classification of air conditioners with similar operating characteristics is achieved.

[0071] Step 304: Train clustering analysis model based on health data and energy consumption data of target air conditioners in different categories.

[0072] Specifically, the system constructs an anomaly detection model for each clustering category to identify abnormal operating conditions in the same type of equipment. For each category of air conditioner group, the system establishes a baseline model of normal operating conditions based on its health data (including refrigeration efficiency, compressor current, component vibration, and other operating parameters) and energy consumption data (including daily average energy consumption, peak power, energy efficiency ratio, and other indicators). The Isolation Forest algorithm is used for anomaly detection, which identifies outliers by calculating the isolation degree of samples. The model sets the anomaly proportion threshold to 5%, i.e., the 5% samples with the lowest isolation degree score are marked as potential anomalies. For example, when the energy consumption data of a certain air conditioner deviates significantly from the normal distribution range of the same type of equipment (e.g., more than 2 standard deviations above the average), the system will identify it as an abnormal device. At the same time, the model also analyzes the persistence of the anomaly based on the time series characteristics, and if the abnormal state lasts more than a preset time (e.g., continuously for 3 days), an anomaly alarm is triggered. In this way, the system can timely discover air conditioners with performance anomalies or energy consumption anomalies in the same type of equipment, providing decision basis for preventive maintenance.

[0073] Step 305: Combine the health degree evaluation model and the clustering analysis model to form a composite air conditioner operating condition intelligent agent.

[0074] Specifically, the system fuses the regression model and the clustering model to form a composite air conditioner working condition intelligent agent. In terms of model fusion strategy, a weighted fusion method is adopted to assign weights according to the prediction accuracy of each model, a weighted voting mechanism based on model credibility is implemented, and a Bayesian fusion method is adopted to consider the uncertainty and correlation of the prediction. In terms of uncertainty quantification, the prediction interval is calculated by the Bootstrap method and the Bayesian method, a multi-dimensional confidence evaluation model is established, and the prediction error risk is quantified. The system finally outputs health quantitative evaluation (90 points or more is good, 70-89 points needs attention, and 70 points or less is at risk), energy consumption anomaly identification (a multi-level early warning mechanism is set), and automatically generates cleaning and maintenance suggestions, including maintenance period, method and specific scheme, which are pushed to relevant personnel in real time through the operation and maintenance APP, realizing intelligent management of the whole process.

[0075] Among them, the specific generation process of the maintenance suggestion is: according to the health evaluation and anomaly identification results, combined with the maintenance standards and experience data in the air conditioner knowledge base, the system automatically generates targeted air conditioner cleaning and maintenance suggestions. For equipment with reduced heat exchange efficiency and increased energy consumption due to dust accumulation in the condenser and evaporator, cleaning and maintenance are recommended in a timely manner, and the cleaning period and cleaning method are specified. For equipment with abnormal operating parameters and decreased health of key components such as compressors, it is recommended to arrange professional technicians for comprehensive maintenance, and specific schemes for replacing parts or adjusting maintenance are given according to the part wear condition.

[0076] Regarding the intelligent task management part: the system pushes the cleaning and maintenance suggestions to relevant operation and maintenance personnel in real time through the operation and maintenance APP, reminding them to arrange maintenance work in a timely manner. Functions such as maintenance task automatic generation, personnel allocation and time cost estimation are established to support the whole process tracking and effect evaluation of the maintenance process. Through methods such as pre-maintenance comparison and analysis, effect quantitative evaluation and cost-benefit analysis, the maintenance strategy is continuously optimized to ensure that the equipment is always in good operating condition. Through the core analysis capability of the "air conditioner working condition intelligent agent", the system realizes the whole process intelligent management from multi-source data fusion to intelligent maintenance suggestion, and provides a complete technical solution for the scientific operation and maintenance of air conditioning equipment.

[0077] Step 40: generating an air conditioner control strategy according to the air conditioner control intelligent agent, and controlling each target air conditioner to operate according to the air conditioner control strategy.

[0078] Among them, in the embodiments of the present application, the air conditioner control strategy refers to the control decision combination made based on the air conditioner control intelligent agent, specifically including but not limited to: the on-off state of the air conditioner (determining when to turn on and off the equipment), the temperature setting (selecting the appropriate target temperature interval among high, medium and low gears), and the running time arrangement (determining the duration of each operation), etc.

[0079] Specifically, the system generates an air conditioner control strategy based on the air conditioner control agent and executes the corresponding control instructions. The air conditioner control agent first obtains the real-time state information of the target air conditioner, including the current temperature (e.g., 25°C), target temperature (e.g., 23°C), power load (e.g., load rate 80%), and outdoor temperature (e.g., 30°C) and other environmental parameters. Based on these state information, the control agent calculates the value function of different control actions through the trained deep Q network, and selects the action combination with the highest value as the control strategy.

[0080] For example, when detecting that the indoor temperature is 25°C, slightly higher than the target temperature of 23°C, and the outdoor temperature is 30°C in a certain scenario, the control strategy generated by the control agent may include: turning on the air conditioner, setting it to medium temperature (23-25°C), and planning the running time for 30 minutes. The system then converts this control strategy into specific control instructions and sends them to the target air conditioner for execution through the communication interface. During the execution of the strategy, the system continuously monitors the temperature changes and energy consumption data, and when it observes that the indoor temperature has dropped to 24°C, it may adjust the strategy to reduce the running power to save energy.

[0081] Through this real-time perception-decision-execution closed-loop control mechanism, the system can dynamically adjust the control strategy according to environmental changes and user needs, achieving intelligent management of air conditioner operation. With the accumulation of control experience, the air conditioner control agent can generate more accurate control strategies, such as using different control parameter combinations at different times or adjusting the running plan in advance when predicting a peak electricity consumption period, thereby ensuring temperature comfort while optimizing energy consumption.

[0082] In another possible embodiment, the step of generating an air conditioner control strategy based on an air conditioner control agent can also be optimized based on temperature curve analysis and weather forecast data. Specifically, the system extracts the temperature change data of the target air conditioner in the past month from the historical database and uses LSTM (Long Short-Term Memory Network) to build a temperature prediction model. The model takes historical temperature data, time features (such as time period, day of the week), and weather forecast data (temperature and weather conditions for the next 24 hours) as input and outputs a temperature change prediction curve for the next 4 hours.

[0083] Based on the prediction results, the generation process of the control strategy also considers the following optimization scheme: 15 minutes before starting the air conditioner in the morning (such as 7:45), the system performs pre-cooling strategy, starts the air conditioner in advance and sets appropriate temperature (such as 26℃), to avoid rapid indoor temperature rise during working hours; for the predicted daily high temperature period (such as 13:00-15:00), the system adjusts the operating parameters 30 minutes in advance, reduces the temperature setting value to 25℃ and selects medium-high wind speed operation, to establish a comfortable indoor environment in advance; when the temperature starts to drop in the evening, the system gradually increases the set temperature (such as to 27℃), and reduces the operating power; 30 minutes before the scheduled shutdown time, the system starts the shutdown plan, gradually increases the set temperature to 28℃, and realizes smooth transition.

[0084] At the same time, the generation of the control strategy will also be dynamically adjusted in combination with weather forecast data. For example, when it is forecasted that high temperature weather (above 35℃) will occur the next day, the system implements pre-cooling energy storage strategy in the evening, adjusts the indoor temperature to a lower level (such as 25℃) in advance, to reduce the cooling load the next day; when it is forecasted that it will be rainy, the system correspondingly increases the temperature setting value, to avoid excessive cooling causing energy waste. This active control mechanism based on prediction can make the control strategy more forward-looking, to cope with temperature changes in advance, avoid energy waste caused by passive adjustment, and thus realize more intelligent and energy-saving air conditioner operation control.

[0085] Step 50: Based on the air conditioner working condition intelligent agent, the health status of the target air conditioner is analyzed, the operation report of each target air conditioner is generated, and visualized display is performed.

[0086] Specifically, to timely grasp the running state of the target air conditioner and effectively manage it, the system analyzes the health state of the target air conditioner based on the air conditioner working condition agent, generates a running report and visually displays it. The air conditioner working condition agent first calculates the health degree index of the target air conditioner (including key indicators such as equipment failure rate and component wear degree) based on a health degree evaluation model, and simultaneously compares and analyzes the health data and energy consumption data of the target air conditioner with those of similar equipment using a clustering analysis model. For example, when the health degree index of a certain target air conditioner is 75 points and its energy consumption index significantly deviates from the normal distribution range of similar equipment (such as 30% higher than the average value), the air conditioner working condition agent will mark this equipment as a key focus object. Based on the analysis results, the system automatically generates a running report of the target air conditioner. The running report contains equipment basic information (equipment number, installation location, etc.), health state evaluation results (health degree index and its change trend), energy consumption analysis results (comparison of energy consumption index with similar equipment), and targeted operation suggestions (such as checking refrigerant, cleaning heat exchanger, etc.). The system uses multi-dimensional visualization to display the running report: it uses a dashboard to display the health degree index in real time, uses different colors (red for abnormal, yellow for warning, and green for normal) to intuitively reflect the equipment state; uses a line chart to show the historical change trend of the health degree index and the energy consumption index; uses a radar chart to show the performance level of the target air conditioner in multiple dimensions such as temperature control effect and energy efficiency; and uses a column chart to compare and display the index differences between the target air conditioner and similar equipment. This analysis and visualization method based on the air conditioner working condition agent can help managers intuitively understand the running state of each target air conditioner and timely discover potential problems.

[0087] As an optional embodiment, the step of analyzing the health state of the target air conditioner based on the air conditioner working condition agent, generating a running report of each target air conditioner, and visually displaying it can further include steps 401-402:

[0088] Step 401: input the time series features of each target air conditioner into the air conditioner working condition agent, obtain the health degree and energy consumption degree of each target air conditioner through the health degree evaluation model, and classify and identify each target air conditioner through the clustering analysis model to obtain abnormal category air conditioners.

[0089] Specifically, to accurately evaluate the running state of the target air conditioner, the system first obtains the time series feature data of each target air conditioner. These time series features include monitoring data during air conditioner operation, which can include but are not limited to: compressor current value, refrigerant pressure, condenser temperature, evaporator temperature, indoor and outdoor temperature difference, equipment running time, etc. The system collects and stores these time series feature data according to a pre-set time interval (such as every 5 minutes) to form a time series data matrix.

[0090] After inputting the temporal features into the air conditioning operating condition agent, the system analyzes them using a pre-trained health assessment model. This health assessment model employs an LSTM (Long Short-Term Memory) network structure, calculating the health index of the target air conditioner by analyzing the changing trends of the temporal features. The calculation of the health index considers multiple dimensions: equipment failure rate (calculated based on indicators such as compressor current anomalies and refrigerant pressure fluctuations), component wear degree (assessed based on operating time and load conditions), and temperature control effect (assessed based on indoor and outdoor temperature difference and adjustment response time). Simultaneously, the model also calculates energy consumption indicators, including energy consumption per unit cooling capacity and peak power utilization rate.

[0091] After obtaining the health status and energy consumption of each target air conditioner, the system uses a clustering analysis model to classify and identify them. This clustering model employs a modified K-means algorithm, using health status and energy consumption as feature vectors to divide the target air conditioners into multiple categories. By setting reasonable clustering thresholds, the system can identify air conditioners in abnormal categories. For example, when an air conditioner's health status is below 70 points and its energy consumption exceeds the average of similar equipment by more than 30%, that air conditioner will be classified as abnormal. This time-series feature-based analysis method can promptly detect potential equipment anomalies, providing a basis for subsequent maintenance decisions.

[0092] Step 402: Generate an operation report based on the health status, energy consumption, and abnormal air conditioner categories of each target air conditioner, and display it visually.

[0093] Specifically, based on the above analysis results, the system generates an operation report that includes health status, energy consumption, and anomaly category information, and intuitively displays various indicators through multi-dimensional visualization methods (including dashboards, line graphs, radar charts, etc.) to help managers promptly identify and handle equipment anomalies.

[0094] As an optional embodiment, the step of generating and visually displaying an operation report based on the health status, energy consumption, and abnormal air conditioner categories of each target air conditioner may further include steps 501-504:

[0095] Step 501: Quantitatively score each target air conditioner based on its health status, and determine the air conditioners with poor health based on the scores.

[0096] Specifically, to accurately assess the operating conditions of target air conditioners, the system quantitatively scores each target air conditioner based on health degree. The system adopts a percentage scoring mechanism and performs weighted calculation according to multiple dimensions of the health degree index: compressor performance index (weight 0.3, including current stability, start-stop frequency, etc.), refrigeration system index (weight 0.3, including refrigerant pressure, heat exchange efficiency, etc.), temperature control effect index (weight 0.2, including temperature adjustment response time, temperature fluctuation range, etc.), and equipment life index (weight 0.2, including cumulative running time, maintenance records, etc.). The system determines that a target air conditioner with a score below a preset threshold (e.g., 75 points) is a substandard health air conditioner. For example, when the compressor performance score of an air conditioner is 65 points, the refrigeration system score is 70 points, the temperature control effect score is 80 points, and the equipment life score is 75 points, the weighted total score is 71.5 points, and the air conditioner will be marked as a substandard health air conditioner by the system.

[0097] Step 502: Mark the target air conditioner whose energy consumption exceeds the average energy consumption of air conditioners of the same type by a preset proportion as an energy consumption abnormal air conditioner.

[0098] Specifically, the system identifies target air conditioners with energy consumption abnormalities through comparative analysis. First, the system classifies target air conditioners according to their models, use environments, and other characteristics, and calculates the average energy consumption value of each type of air conditioner. Then, the system sets a preset proportion threshold for energy consumption abnormality determination (e.g., exceeding the average value by 25%). For each target air conditioner, the system calculates its recent (e.g., the last 7 days) average energy consumption value and compares it with the average energy consumption of the category to which it belongs. When the energy consumption value of a target air conditioner exceeds the average energy consumption by a preset proportion, the system marks it as an energy consumption abnormal air conditioner. For example, if the average daily energy consumption of a certain type of air conditioner is 100 kWh, and the daily average energy consumption of a certain air conditioner reaches 130 kWh, which is 30% higher than the average, then the air conditioner is marked as an energy consumption abnormal air conditioner.

[0099] Step 503: Determine the abnormal air conditioner set based on substandard health air conditioners, energy consumption abnormal air conditioners, and abnormal category air conditioners.

[0100] Specifically, the system determines the final abnormal air conditioner set through set operations. The system first establishes three subsets: the substandard health air conditioner set A (the scoring substandard air conditioners identified in step 501), the energy consumption abnormal air conditioner set B (the energy consumption exceeding air conditioners identified in step 502), and the abnormal category air conditioner set C (the abnormal air conditioners identified by clustering analysis in step 401). Then, the system performs set union operation to obtain the abnormal air conditioner set D = A ∪ B ∪ C. To avoid misjudgment, the system also verifies the union result: if a certain air conditioner only appears in a single subset, the system will re-evaluate its abnormality; if it appears in two or more subsets at the same time, it is directly confirmed as an abnormal air conditioner. This multi-dimensional abnormality determination method can improve the accuracy and reliability of abnormality identification.

[0101] Step 504: Generate a running report corresponding to the set of abnormal air conditioners and perform visual display.

[0102] Specifically, the system generates a special running report for the set of abnormal air conditioners and adopts a multi-level visual display scheme. The running report contains the following core content: basic information of abnormal air conditioners (device number, location, model, etc.), abnormal type analysis (poor health, abnormal energy consumption, or specific determination basis of abnormal category), abnormal parameter details (abnormal value of specific indicators and their trend), and recommended maintenance measures. In terms of visual display, the system adopts an interactive dashboard design: the top shows the statistical chart of the total number and distribution of abnormal air conditioners, the middle uses a heat map to show the spatial distribution of abnormal air conditioners, and the bottom uses multi-dimensional charts to show specific abnormal indicators. For example, for an air conditioner that has both a low health score (71.5) and an abnormal energy consumption (30% over standard), the system will highlight the abnormal conditions in these two dimensions in the report and show the trend of these indicators through a line chart, while providing detailed abnormal reason analysis and treatment suggestions. This visual report can help managers quickly locate problem air conditioners, understand the specific situation of abnormalities, and develop targeted maintenance solutions.

[0103] As an optional embodiment, the step of generating a running report corresponding to the set of abnormal air conditioners and performing visual display can further include steps 601-603:

[0104] Step 601: Analyze the abnormal characteristics of each abnormal air conditioner in the set of abnormal air conditioners.

[0105] Specifically, the system analyzes each abnormal air conditioner in the set of abnormal air conditioners and extracts its abnormal characteristics. The system first divides the abnormal characteristics into three categories: hardware abnormal characteristics, energy consumption abnormal characteristics, and environmental factor characteristics. For hardware abnormal characteristics, the system analyzes the deviation of key operating parameters from the normal range, such as compressor current fluctuation (normal range is ±20% of rated value), refrigerant pressure abnormality (high-side pressure exceeds 2.5 MPa or low-side pressure is less than 0.4 MPa), heat exchanger temperature abnormality (condenser temperature exceeds 55°C or evaporator temperature is less than 5°C), etc. For energy consumption abnormal characteristics, the system calculates the energy consumption distribution of each operating link, including refrigeration system energy consumption ratio (which should be within 65%-75% of total energy consumption), fan system energy consumption ratio (which should be within 15%-25%), etc. For environmental factor characteristics, the system analyzes the impact of room usage (such as door and window opening, personnel density), outdoor temperature changes, etc. on device operation. For example, when the system detects that the condenser temperature of a certain air conditioner continuously exceeds 60°C and the refrigeration system energy consumption ratio reaches 85%, it will determine that it has the abnormal characteristic of low heat exchange efficiency.

[0106] Step 602: Match the abnormal features with the preset strategy library to obtain the maintenance strategy of each abnormal air conditioner, which at least includes cleaning maintenance scheme, equipment repair scheme, and operation optimization strategy.

[0107] Specifically, the system intelligently matches the extracted abnormal features with the preset strategy library to generate targeted maintenance strategies. The strategy library contains three basic maintenance schemes: cleaning maintenance scheme (for problems such as component dust and blockage), equipment repair scheme (for problems such as component performance decline and failure), and operation optimization strategy (for problems such as energy consumption anomaly and low running efficiency). The system selects the corresponding maintenance strategy based on the type and degree of abnormal features. For example, for a device with low heat exchange efficiency, if it is detected that it is due to dust accumulation in the heat exchanger, the system will generate a cleaning maintenance scheme, including cleaning cycle (recommended to be cleaned once every 3 months), cleaning method (using professional cleaning agent combined with high-pressure water gun), and cleaning precautions. For a device with a declining compressor performance, the system will generate an equipment repair scheme, detailing the items to be checked (such as compressor current, starting capacitor, etc.), and the components that may need to be replaced and their specification requirements. For a device with energy consumption anomaly but normal hardware status, the system will generate an operation optimization strategy, such as adjusting the temperature setting range (recommended to be set at 26-28°C), optimizing the fan operation mode (automatically adjusting the fan speed according to the load), and other specific measures.

[0108] Step 603: Display the abnormal features and maintenance strategies through a visual interface, and establish an execution progress tracking mechanism for maintenance tasks.

[0109] Specifically, the system visually displays abnormal features and maintenance strategies through a visual interface, and establishes a complete maintenance task tracking mechanism. In terms of visual interface design, the system uses a hierarchical display method: the top layer displays the overall distribution of abnormal air conditioners, using different colors to identify different types of abnormalities (red for hardware abnormalities, yellow for energy consumption abnormalities, and blue for environmental factor abnormalities); the middle layer displays the specific abnormal features of each abnormal air conditioner, using a radar chart to display the abnormal degree of each parameter; the bottom layer displays the details of the corresponding maintenance strategy, including the specific steps of the maintenance scheme, required tools and materials, estimated completion time, etc. At the same time, the system establishes an execution progress tracking mechanism for maintenance tasks: through the operation and maintenance management APP, it pushes the maintenance tasks to relevant personnel in real time, requiring them to respond and perform maintenance work within a specified time (such as within 48 hours); maintenance personnel need to record the maintenance process through the APP, including submitting on-site photos, filling in maintenance records, uploading detection data, etc.; the system tracks the execution status of the maintenance tasks in real time, and continuously monitors the equipment operation parameters after the maintenance is completed to evaluate the maintenance effect. This visual display and task tracking mechanism can ensure the timeliness and effectiveness of maintenance work, and improve the efficiency of equipment management.

[0110] Reference Figure 3 The application also provides a structural diagram of an intelligent agent control system of an unattended air conditioner, which specifically comprises:

[0111] a data acquisition module, configured to acquire operation data of each target air conditioner, wherein the operation data comprises environmental parameters, energy consumption data and equipment state data;

[0112] a data processing module, configured to perform edge computing processing on the operation data to obtain time sequence features, and upload the time sequence features to a cloud platform;

[0113] an intelligent agent construction module, configured to construct an air conditioner control intelligent agent and an air conditioner working condition intelligent agent based on the time sequence features;

[0114] a strategy generation module, configured to generate an air conditioner control strategy according to the air conditioner control intelligent agent, and control each target air conditioner to operate according to the air conditioner control strategy;

[0115] a report generation module, configured to analyze the health status of each target air conditioner based on the air conditioner working condition intelligent agent, generate an operation report of each target air conditioner, and perform visual display.

[0116] Optionally, the data processing module is further configured to perform data cleaning on the environmental parameters, the energy consumption data and the equipment state data in the operation data, wherein the data cleaning comprises removing outliers, supplementing missing values and data smoothing processing;

[0117] the cleaned environmental parameters are normalized to obtain an environmental parameter change curve;

[0118] the cleaned energy consumption data are analyzed by using a sliding window to extract energy consumption fluctuation features, and an energy consumption change curve is obtained;

[0119] the cleaned equipment state data are subjected to Fourier transform to extract noise spectrum features, and a state feature curve is obtained;

[0120] the environmental parameter change curve, the energy consumption change curve and the state feature curve are combined according to timestamps to form time sequence features.

[0121] Optionally, the intelligent agent construction module is further configured to obtain time-aligned training parameters based on the time sequence features, wherein the training parameters comprise a state space, an action space and a reward function, the state space comprises a current temperature, a target temperature, a power load and an outdoor temperature, the action space comprises an air conditioner switch state, a temperature gear and a running time length, and the reward function is determined in combination with temperature comfort and energy consumption indexes.

[0122] train the preset initial air conditioner control agent based on the training parameters until the preset stop condition is met, to obtain a trained air conditioner control agent, the preset stop condition being that a result of the reward function after continuous multiple training iterations is greater than a preset threshold.

[0123] Optionally, the agent construction module is further configured to extract key features in the time sequence features as independent variables, and construct a regression model with an air conditioner health degree index and an energy consumption index as dependent variables,

[0124] train the regression model based on historical data of the target air conditioners, to establish a corresponding relationship between the independent variables and the dependent variables, and obtain a health degree evaluation model;

[0125] classify target air conditioners with similar environmental parameter change curves and energy consumption change curves by using a clustering analysis method;

[0126] train a clustering analysis model based on health data and energy consumption data of target air conditioners in different categories;

[0127] combine the health degree evaluation model and the clustering analysis model to form a composite air conditioner working condition agent.

[0128] Optionally, the report generation module is further configured to input the time sequence features of the target air conditioners into the air conditioner working condition agent, obtain health degrees and energy consumption degrees of the target air conditioners by using the health degree evaluation model, and classify and identify the target air conditioners by using the clustering analysis model, to obtain abnormal category air conditioners;

[0129] generate a running report according to the health degrees, the energy consumption degrees, and the abnormal category air conditioners of the target air conditioners, and perform visual display.

[0130] Optionally, the report generation module is further configured to quantitatively score the target air conditioners based on the health degrees, and determine unhealthy air conditioners based on the scores;

[0131] mark a target air conditioner whose energy consumption degree exceeds a preset proportion of average energy consumption of air conditioners of the same type as an energy consumption abnormal air conditioner;

[0132] determine an abnormal air conditioner set based on the unhealthy air conditioners, the energy consumption abnormal air conditioners, and the abnormal category air conditioners;

[0133] generate a running report corresponding to the abnormal air conditioner set, and perform visual display.

[0134] Optionally, the report generation module is further configured to analyze abnormal features of the abnormal air conditioners in the abnormal air conditioner set;

[0135] The abnormal features are matched with a preset strategy library, to obtain a maintenance strategy of each of the abnormal air conditioners, the maintenance strategy at least including a cleaning maintenance scheme, an equipment maintenance scheme and an operation optimization strategy;

[0136] The abnormal features and the maintenance strategy are displayed through a visual interface, and an execution progress tracking mechanism of a maintenance task is established.

[0137] It should be noted that, when the device provided in the above embodiments implements its functions, only the division of the above functional modules is exemplified, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be described here.

[0138] The embodiment of the application further provides a computer storage medium, which can store a plurality of instructions, the instructions being suitable for being loaded by a processor and executing the intelligent agent control method of the unattended air conditioner provided in the above embodiment, and the specific execution process can be referred to the specific description of the above embodiment, which will not be described here.

[0139] Please refer to Figure 4 The application further discloses an electronic device. Figure 4 is a structural schematic diagram of an electronic device disclosed by the embodiment of the application. The electronic device 300 can include at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302.

[0140] The communication bus 302 is configured to realize the connection and communication between the components.

[0141] The user interface 303 can include a display screen (Display) and a camera (Camera), and the optional user interface 303 can further include a standard wired interface and a wireless interface.

[0142] The network interface 304 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0143] The processor 301 can include one or more processing cores. The processor 301 connects various parts within the server through various interfaces and lines, performs various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Alternatively, the processor 301 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 301 can integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes operating systems, user interfaces, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 301, but can be realized by a separate chip.

[0144] The memory 305 can include a random access memory (RAM) and a read-only memory (ROM). Alternatively, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 305 can alternatively be at least one storage device located away from the aforementioned processor 301. Referring to Figure 4 The memory 305 as a computer storage medium can include an operating system, a network communication module, a user interface module, and an application program of the unattended air conditioner intelligent agent control method.

[0145] In Figure 4In the electronic device 300 shown, the user interface 303 is mainly used to provide an interface for the user to input, and obtain data input by the user; and the processor 301 can be used to invoke an application program of the intelligent agent control method of the unattended air conditioner stored in the memory 305, and when executed by one or more processors 301, the electronic device 300 performs the method described in one or more of the above embodiments. It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a combination of a series of actions, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, some steps can be performed in other order or at the same time. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0146] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0147] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner for actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different parts can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical or other forms.

[0148] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0149] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0150] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable memory. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned memory includes: a U disk, a mobile hard disk, a magnetic disk or an optical disk and various media that can store program codes.

[0151] The above described are only exemplary embodiments of the present disclosure, and cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon considering the specification and the practical true disclosure.

[0152] The present application is intended to cover any variations, uses or adaptive changes of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not recorded in the present disclosure.

Claims

1. A method for intelligent control of an unattended air conditioner, characterized in that, The method includes: Acquire the operating data of each target air conditioner, including environmental parameters, energy consumption data, and equipment status data; Edge computing is performed on the running data to obtain time-series features, and the time-series features are uploaded to the cloud platform; Based on the aforementioned temporal characteristics, an air conditioning control intelligent agent and an air conditioning operating condition intelligent agent are constructed. The air conditioning control agent generates an air conditioning control strategy and controls the operation of each target air conditioner according to the air conditioning control strategy. Based on the analysis of the target air conditioner's health status by the air conditioner operating condition intelligent agent, an operation report for each target air conditioner is generated and displayed visually.

2. The intelligent control method for unattended air conditioners according to claim 1, characterized in that, Edge computing processing is performed on the running data to obtain time-series features, including: The environmental parameters, energy consumption data, and equipment status data in the operational data are cleaned. The data cleaning includes removing outliers, supplementing missing values, and data smoothing. The environmental parameters after cleaning were normalized to obtain the environmental parameter change curves. A sliding window analysis was performed on the cleaned energy consumption data to extract energy consumption fluctuation characteristics and obtain the energy consumption change curve. Fourier transform is performed on the equipment status data after cleaning to extract noise spectrum features and obtain status feature curves; The environmental parameter change curve, the energy consumption change curve, and the state characteristic curve are aligned and combined according to timestamps to form a time-series feature.

3. The intelligent control method for unattended air conditioners according to claim 1, characterized in that, Based on the aforementioned temporal features, an intelligent agent for air conditioning control is constructed, including: Based on the aforementioned temporal features, time-aligned training parameters are obtained. The training parameters include a state space, an action space, and a reward function. The state space includes the current temperature, target temperature, power load, and outdoor temperature. The action space includes the air conditioner's on / off status, temperature setting, and running time. The reward function is determined by combining temperature comfort and energy consumption indicators. The initial air conditioning control agent is trained based on the training parameters until the training meets the preset stopping condition, and the trained air conditioning control agent is obtained. The preset stopping condition is that the calculation result of the reward function is greater than the preset threshold after multiple consecutive training iterations.

4. The intelligent control method for unattended air conditioners according to claim 1, characterized in that, Based on the aforementioned time-series features, an intelligent agent for air conditioning operation conditions is constructed, including: Key features were extracted from the time-series characteristics as independent variables, and a regression model was constructed using air conditioner health indicators and energy consumption indicators as dependent variables. The regression model is trained based on the historical data of each target air conditioner, and the correspondence between the independent variables and the dependent variables is established to obtain the health assessment model. Cluster analysis was used to classify target air conditioners with similar environmental parameter variation curves and energy consumption variation curves. A clustering analysis model was trained based on health data and energy consumption data of different categories of target air conditioners; The health assessment model and the cluster analysis model are combined to form a composite intelligent agent for air conditioning operation.

5. The intelligent control method for unattended air conditioners according to claim 4, characterized in that, Based on the analysis of the health status of the target air conditioners by the air conditioning operating condition intelligent agent, an operation report for each target air conditioner is generated and visualized, including: The temporal characteristics of each target air conditioner are input into the air conditioner operating condition intelligent agent. The health and energy consumption of each target air conditioner are obtained through the health assessment model. The target air conditioners are classified and identified through the cluster analysis model to obtain abnormal category air conditioners. An operation report is generated based on the health status, energy consumption, and abnormal category of each target air conditioner, and then visualized.

6. The intelligent control method for unattended air conditioners according to claim 5, characterized in that, An operation report is generated based on the health status, energy consumption, and abnormal category of each target air conditioner, and then visualized, including: The target air conditioners are quantitatively scored based on their health status, and air conditioners with poor health are identified based on their scores. The target air conditioner whose energy consumption exceeds the preset ratio of the average energy consumption of air conditioners of the same type is marked as an air conditioner with abnormal energy consumption. Based on the air conditioners with poor health, air conditioners with abnormal energy consumption, and air conditioners of the abnormal category, a set of abnormal air conditioners is determined; Generate an operation report corresponding to the abnormal air conditioner set and display it visually.

7. The intelligent control method for unattended air conditioners according to claim 6, characterized in that, Generate an operation report corresponding to the set of abnormal air conditioners and display it visually, including: Analyze the abnormal characteristics of each abnormal air conditioner in the abnormal air conditioner set; The abnormal features are matched with a preset strategy library to obtain the maintenance strategy for each abnormal air conditioner. The maintenance strategy includes at least a cleaning and maintenance plan, an equipment repair plan, and an operation optimization strategy. The abnormal characteristics and maintenance strategies are displayed through a visual interface, and a mechanism for tracking the execution progress of maintenance tasks is established.

8. An intelligent control system for an unattended air conditioner, characterized in that, The system includes: The data acquisition module is used to acquire the operating data of each target air conditioner, including environmental parameters, energy consumption data and equipment status data; The data processing module is used to perform edge computing processing on the running data to obtain time-series features, and upload the time-series features to the cloud platform; The intelligent agent construction module is used to construct an air conditioning control intelligent agent and an air conditioning operating condition intelligent agent based on the aforementioned temporal features; The strategy generation module is used to generate an air conditioning control strategy based on the air conditioning control agent, and control the operation of each target air conditioner according to the air conditioning control strategy; The report generation module is used to analyze the health status of the target air conditioner based on the air conditioner operating condition intelligent agent, generate an operation report for each target air conditioner, and display it visually.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.

Images