A central air-conditioning intelligent energy-saving management control system and method

Abstract

The present invention discloses an intelligent energy-saving management and control system and method for central air conditioning, which relates to the field of intelligent control technology. The system includes the following steps: collecting and preprocessing indoor monitoring data; dividing the control area based on the processed indoor monitoring data using a heat balance control technology; optimizing the ventilation direction and air volume level in the control area using an airflow simulation algorithm, reducing pollutant concentration and balancing the heat distribution of personnel, and generating an energy-saving optimization plan; collecting multi-source data, fusing the multi-source data with indoor monitoring data using a Kalman filter algorithm, generating a comprehensive environmental state vector, constructing a load prediction model, and predicting future heat load and pollutant concentration trends; and dynamically adjusting central air conditioning parameters using a reinforcement learning algorithm using load prediction, indoor monitoring data, and the energy-saving optimization plan. The present invention achieves accurate prediction of future heat load and pollutant concentration by constructing a load prediction model based on LSTM.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to an intelligent energy-saving management control system and method for a central air conditioner. Background Art

[0002] Central air conditioning, a core component of modern building environmental management, has seen rapid development in recent years in intelligent and energy-saving optimization technologies. Early central air conditioning systems relied primarily on simple thermostats for on / off control. These systems have evolved into automated control systems based on sensors and programmable logic controllers, enabling basic temperature and air volume adjustments. Furthermore, with advances in the Internet of Things, big data analytics, and artificial intelligence, central air conditioning systems have begun integrating multi-source environmental data and optimizing operating parameters through pre-set algorithms.

[0003] However, existing technologies still have significant deficiencies in the dimension of dynamic spatial control: the existing central air-conditioning system adopts a static zoning control method, which has the core defect of insufficient dynamic adaptability. The temperature control area divided by fixed geometric boundaries is solidified during the equipment deployment stage and cannot be topologically reconstructed according to the real-time distribution of personnel. There is a spatial mismatch between the gathering area and the air-conditioning supply area; when the flow of personnel inside the building causes changes in the heat load distribution, the static zoning method cannot dynamically adjust the air supply parameters, which not only causes local cooling or heating, but also cannot effectively track the diffusion path of pollutants and perform directional ventilation. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an intelligent energy-saving management and control method for central air conditioning to solve the problem of being unable to effectively track the diffusion path of pollutants and perform directional ventilation.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides an intelligent energy-saving management and control method for central air conditioning, which includes collecting indoor monitoring data and preprocessing it, wherein the indoor monitoring data includes personnel location coordinates and pollutant concentration data; dividing the control area based on the processed indoor monitoring data through thermal balance control technology, and using an airflow simulation algorithm to optimize the ventilation direction and air volume level of the control area, reduce the pollutant concentration and balance the heat distribution of personnel, and generate an energy-saving optimization plan; collecting multi-source data, using a Kalman filter algorithm to fuse the multi-source data and indoor monitoring data, generate a comprehensive environmental state vector, construct a load prediction model, and predict future heat load and pollutant concentration trends; through a reinforcement learning algorithm, using load prediction, indoor monitoring data and energy-saving optimization plans, dynamically adjust central air-conditioning parameters, and optimize pollutant control, heat balance and energy consumption of central air-conditioning operating parameters.

[0008] As a preferred solution of the central air-conditioning intelligent energy-saving management and control method of the present invention, wherein: the heat balance control technology is used to divide the control area based on indoor monitoring data, and the specific steps are:

[0009] Edge computing devices scan the coordinates of people in each control area, count the number of people, calculate the average number of people in the meeting room, assess the heat load in crowded areas, set heat thresholds based on the average number of people, and determine high-heat and low-heat areas.

[0010] Compare pollutant concentration data with the normal concentration range of the preset spectral database, classify the pollution degree of the control area, and determine the high pollution area and low pollution area;

[0011] Based on the classification of heat zones and pollution levels, the purifiers in the control areas are prioritized, with high-priority areas and low-priority areas marked.

[0012] As a preferred solution of the intelligent energy-saving management and control method for central air conditioning described in the present invention, the use of airflow simulation algorithm to optimize the ventilation direction and air volume level of the control area refers to constructing a three-dimensional model of the conference room, dividing the air flow grid, mapping pollutant concentration data and personnel heat, analyzing the influence of the outlet wind speed and outlet angle, calculating the air speed and pollutant diffusion speed of the grid points in the air flow grid, adjusting the ventilation direction and air volume level, optimizing the air flow field, reducing the pollutant concentration and balancing the heat distribution of personnel.

[0013] As a preferred solution of the central air-conditioning intelligent energy-saving management and control method of the present invention, wherein: the Kalman filter algorithm is used to fuse multi-source data and indoor monitoring data, the specific steps are:

[0014] Calculate external heat input value based on outdoor temperature data, indoor temperature data and window area;

[0015] Define the personnel position coordinates, pollutant concentration data, indoor temperature data, outdoor temperature data, window area and external heat input value as the initial environmental state vector;

[0016] Calculate the noise covariance for the dynamic data in the initial environment state vector, while the static data remain fixed;

[0017] The dynamic data includes personnel location coordinates, pollutant concentration data, indoor temperature data, outdoor temperature data and external heat input value;

[0018] The static data includes window area;

[0019] Based on the Kalman filter algorithm, the posterior state estimate of the previous moment is fused with the multi-source observation data of the current moment to generate the comprehensive environment state vector of the current moment.

[0020] As a preferred solution of the intelligent energy-saving management and control method for central air conditioning described in the present invention, the construction of a load forecasting model refers to constructing a load forecasting model based on an LSTM model, defining an input layer, an LSTM layer, a fully connected layer and an output layer. The input layer processes indoor monitoring data and a comprehensive environmental state vector through standardization and a sliding window to generate a time series. The LSTM layer captures the long-term trend and short-term fluctuation of the time series. The fully connected layer maps it as a prediction target, and the output layer generates a load forecast.

[0021] As a preferred solution of the intelligent energy-saving management and control method for central air conditioning described in the present invention, the reinforcement learning algorithm refers to defining the intelligent agent and environment, setting the state space and action space, and optimizing pollutant control, heat balance and energy consumption of central air conditioning operating parameters based on initialization strategy and reward function.

[0022] As a preferred solution of the central air-conditioning intelligent energy-saving management and control method of the present invention, the multi-source data includes outdoor temperature data, outdoor temperature data and window area.

[0023] In a second aspect, the present invention provides a central air-conditioning intelligent energy-saving management and control system, comprising:

[0024] An acquisition module is used to collect and pre-process indoor monitoring data, wherein the indoor monitoring data includes personnel location coordinates and pollutant concentration data;

[0025] The partitioning module is used to divide the control area based on the processed indoor monitoring data through heat balance control technology, and use the airflow simulation algorithm to optimize the ventilation direction and air volume level in the control area, reduce the concentration of pollutants and balance the heat distribution of people, and generate energy-saving optimization solutions;

[0026] The prediction module is used to collect multi-source data, fuse the multi-source data with indoor monitoring data using the Kalman filter algorithm, generate a comprehensive environmental state vector, build a load prediction model, and predict future heat load and pollutant concentration trends;

[0027] The adjustment module is used to dynamically adjust central air-conditioning parameters through reinforcement learning algorithms, using load forecasting, indoor monitoring data and energy-saving optimization schemes to optimize pollutant control, heat balance and energy consumption of central air-conditioning operating parameters.

[0028] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the central air-conditioning intelligent energy-saving management and control method as described in the first aspect of the present invention is implemented.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the central air-conditioning intelligent energy-saving management and control method as described in the first aspect of the present invention is implemented.

[0030] The present invention achieves the following beneficial effects: By constructing a load forecasting model based on LSTM, it accurately predicts future heat loads and pollutant concentrations. The load forecasting model uses two layers of LSTM to analyze time series, capturing long-term trends and short-term fluctuations, and outputs prediction results for each time step. This provides a forward-looking basis for dynamic central air conditioning control, suitable for complex indoor scenarios with frequent personnel movements and changing external environments. It also avoids air quality deterioration or heat imbalance caused by delayed response. Furthermore, the prediction results support dynamic adjustment of energy-saving plans, prioritizing resource allocation to high-demand areas and reducing inefficient operation time. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 The figure is a flow chart of the intelligent energy-saving management and control method for central air conditioning.

[0033] Figure 2 This is a schematic diagram of the intelligent energy-saving management and control system for central air conditioning.

[0034] Figure 3 This is the flow chart for determining the dynamic partition of thermal balance.

[0035] Figure 4This is the structure diagram of the LSTM time series prediction model. DETAILED DESCRIPTION

[0036] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0037] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0038] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0039] In this embodiment, refer to Figures 1 to 4 This embodiment provides a central air-conditioning intelligent energy-saving management and control method, comprising the following steps:

[0040] S1. Collect indoor monitoring data and pre-process it.

[0041] Going a step further, a LiFi transmitter is installed on the top of the conference room, a quantum dot sensor is installed on the return air vent of the central air conditioner, and a photosensitive receiver is installed below the air outlet;

[0042] The LiFi transmitter transmits a modulated visible light signal at a frequency of 120 Hz. The modulated visible light signal contains a timestamp and the LiFi transmitter number.

[0043] The photosensitive receiver receives the modulated visible light signal, collects multiple groups (e.g., 120 groups) of modulated visible light signal intensity and arrival time data per second, and records the modulated visible light signal intensity and arrival time data. For example, photosensitive receiver 1 records the arrival time 2×10⁻ 9 Seconds, photosensitive receiver 2 records 3×10⁻ 9 Seconds, photosensitive receiver 3 records 2.5×10⁻ 9 Seconds etc.

[0044] Calculate the person's location coordinates based on the arrival time. Read the arrival time data of at least three photosensitive receivers. Calculate the person's location coordinates based on the LiFi transmitter's location coordinates, updating every 3 seconds. The person's location coordinates are calculated based on the arrival time of the photosensitive receiver and the distance. This is obtained by triangulation using the LiFi transmitter's coordinates (e.g., X=0m, Y=0m, Z=3m). The expression is:

[0045] Distance = speed of light × arrival time;

[0046] Among them, the speed of light is the propagation speed of the modulated visible light signal and is a physical constant;

[0047] The quantum dot sensor collects reflectance spectral data at a sampling rate of 100 times per second and identifies pollutant concentration data by comparing it with a preset spectral database. The preset spectral database is a storage of spectral characteristics of various pollutants (such as PM2.5, VOC, CO2), which includes the normal concentration range based on indoor air quality standards (PM2.5: 0-35 micrograms per cubic meter, VOC: 0-0.6 milligrams per cubic meter, CO2: 400-1000 ppm).

[0048] The personnel location coordinates and pollutant concentration data are transmitted to the edge computing unit via Ethernet.

[0049] The personnel position coordinates are filtered using the sliding average method, as follows:

[0050] Set the window size to 3 seconds, collect multiple groups (e.g., 360 groups) of modulated visible light signal intensity and arrival time data, check the intensity of each group of modulated visible light signal, and remove the entire group of modulated visible light signal intensity and arrival time data below decibel milliwatts (e.g., -95);

[0051] Using the arrival time of the remaining valid data, the personnel position coordinates are calculated by triangulation to obtain the X, Y, and Z values ​​of each group;

[0052] Process the X, Y, and Z coordinate values ​​separately. For example, if the X coordinate has 350 values, such as 4.4 meters, 4.6 meters, and 4.5 meters, add these values ​​and divide them by 350 to get the average value of X. Similarly, process the Y and Z coordinates. After processing, the smoothed coordinates of the person's position are generated.

[0053] The pollutant concentration data are smoothed using the sliding average method, as follows:

[0054] Set the window size to 6 seconds, collect pollutant concentration data (PM2.5, VOC, CO2), check the concentration value of each pollutant, and remove instantaneous abnormal values. For example, remove the PM2.5 instantaneous value of 80 micrograms per cubic meter, the VOC instantaneous value of 1.0 ppm, and the CO2 instantaneous value of 1500 ppm. Keep the values ​​within the normal range, such as PM2.5 of 50-60 micrograms per cubic meter, VOC of 0.6-0.7 ppm, and CO2 of 1000-1100 ppm;

[0055] The average of the remaining normal concentration values ​​for each pollutant in the pollutant concentration data is calculated. For example, the remaining values ​​of PM2.5 include 54, 55, and 56 micrograms per cubic meter, all of which are within the normal range of 50-60 micrograms per cubic meter. These values ​​are added together and divided by the number of valid data points to obtain the average value. VOC and CO2 are processed similarly. After completion of this processing, the smoothed pollutant concentration data is generated.

[0056] The smoothed personnel location coordinates and pollutant concentration data are encapsulated as preprocessed indoor monitoring data in the format of "meeting room-timestamp-personnel location coordinates-pollutant concentration data". The indoor monitoring data is stored in the temporary buffer of the edge computing device.

[0057] S2. Divide the control area based on the processed indoor monitoring data through heat balance control technology, and use the airflow simulation algorithm to optimize the ventilation direction and air volume level of the control area, reduce the concentration of pollutants and balance the heat distribution of people, and generate an energy-saving optimization plan.

[0058] Furthermore, indoor monitoring data is obtained from the temporary buffer of the edge computing device;

[0059] Using the positioning function of the LiFi transmitter, the conference room is divided into multiple control areas, each of which is centered on the LiFi transmitter;

[0060] Through heat balance control technology, control areas are divided according to personnel location coordinates and pollutant concentration data, and the heat and air quality requirements of personnel in each control area are evaluated as follows:

[0061] The edge computing device scans the location coordinates of personnel in each control area and counts the number of personnel in each control area;

[0062] Based on the number of people, calculate the average number of people in the meeting room and evaluate the heat load in crowded areas (that is, the more people there are, the higher the heat load). The expression for the average number of people is:

[0063] Average number of personnel = total number of personnel ÷ number of control areas;

[0064] Set heat thresholds based on the average number of people to determine heat zones and heat loads;

[0065] If the number of people in the control area is 5 or more (10 square meters), it is determined to be a high-heat area, indicating a high heat load;

[0066] If the number of people in the control area is 0-4 (10 square meters), it is determined to be a low-heat area, indicating that the heat load is small;

[0067] Compare pollutant concentration data with the normal concentration range of the preset spectral database, classify the pollution degree of the control area, and determine the air volume level;

[0068] If the concentration of any pollutant in the pollutant concentration data exceeds the normal concentration range, it will be marked as a high-pollution area, indicating that high ventilation volume is required to purify the air;

[0069] If all pollutant concentrations in the pollutant concentration data are within the normal concentration range, it is marked as a low-pollution area, indicating that low ventilation volume is required to purify the air;

[0070] The heat threshold of 5 people per 10 square meters is chosen because the corresponding occupant density (0.5 people per square meter) is the critical point at which the conference room's heat load (50 watts per square meter) and pollutant concentrations (such as CO2) increase significantly. This meets the ventilation rate requirements of ASHRAE 62.1 (an international ventilation standard) (25-50 liters per second per 10 square meters), requiring high ventilation rates for air purification.

[0071] Prioritize purifiers in controlled areas based on the classification of personnel heat and pollutant concentrations;

[0072] If the control area is a high-heat area or a high-pollution area, it will be marked as a high-priority area, indicating that priority control is required (such as increasing ventilation and turning on the purifier first);

[0073] If the regulated area is a low-heat area and a low-pollution area, it is marked as a low-priority area, indicating that routine regulation is required (such as maintaining low ventilation and regular purification);

[0074] Airflow simulation algorithms are used to optimize ventilation direction and air volume levels in each controlled area, as follows:

[0075] Use a laser rangefinder to measure the length, width and height of the conference room to generate the conference room space parameters;

[0076] The LiFi transmitter positioning function is combined with the personnel's location coordinates to determine the location of the central air conditioner's air outlet and return air outlet;

[0077] Use spatial modeling software to associate the conference room's spatial parameters, air outlet locations, and return air outlet locations to generate a three-dimensional model of the conference room.

[0078] Divide the 3D model of the conference room into air flow grids and define multiple grid points (e.g., 20,000) to cover the 3D model of the conference room. Each grid point records the spatial coordinates of the 3D model of the conference room.

[0079] Map pollutant concentration data and human heat to multiple grid points in the air flow grid, and analyze the impact of outlet wind speed and outlet angle on pollutant concentration data and human heat distribution in the air flow grid. For example, define the test outlet wind speed as 1.5 to 2.5 meters per second and the outlet angle as 30 to 60 degrees, and observe the changes in pollutant concentration data and human heat distribution in the air flow grid. Assume that the outlet wind speed in control area A is 2.0 meters per second and the outlet angle is 45 degrees, which reduces the PM2.5 concentration from 40 micrograms per cubic meter to 37 micrograms per cubic meter and the human heat from 600 watts to 550 watts, indicating that high wind speed drives pollutant concentration and human heat to diffuse outward;

[0080] Based on the air outlet wind speed, air outlet angle and pollutant concentration data, the air speed at each grid point in the air flow grid and the pollutant diffusion rate from the grid point to the adjacent grid points are calculated. The air outlet wind speed and air outlet angle are adjusted to optimize the air flow grid, reduce the pollutant concentration and balance the heat distribution of people.

[0081] The air velocity expression is:

[0082] ;

[0083] in, is a grid point Air velocity (meters per second), is the initial wind speed at the air outlet (meters per second), is the cosine function, which is used to calculate the cosine value of the outlet angle. is the angle between the outlet angle and the line connecting the grid points (radians), is the grid point Euclidean distance to the air outlet (meters), is the attenuation coefficient, is the base of natural logarithms, approximately equal to 2.71828, is the grid point index, which is the number of the grid point in the air flow grid;

[0084] The pollutant diffusion rate expression is:

[0085] ;

[0086] in, For pollutants from grid points To the neighboring grid point Diffusion rate (micrograms per cubic meter per second), is the diffusion coefficient (m² per second), is the grid point in the air flow grid The concentration of pollutants in micrograms per cubic meter, is the adjacent grid point in the air flow grid The concentration of pollutants, is the index of the neighboring grid point, The number of another adjacent grid point;

[0087] The air outlet controller adjusts the ventilation direction and air volume level of adjacent control areas, so that the high ventilation volume in high-priority areas drives the adjacent areas, forming a continuous air flow field, promoting the uniform distribution of human heat and pollutant concentrations among the control areas, and reducing the high load operation of the air conditioner in a single control area.

[0088] Summarize the ventilation direction, air volume level, and purifier priority of each control area and generate an energy-saving optimization plan, including the control area number, ventilation direction, air volume level, and purifier priority.

[0089] S3. Collect multi-source data, use the Kalman filter algorithm to fuse multi-source data and indoor monitoring data, generate a comprehensive environmental state vector, build a load prediction model, and predict future heat load and pollutant concentration trends.

[0090] Furthermore, outdoor temperature data is obtained through the Internet meteorological API interface, temperature and humidity sensors are deployed in the corners of the conference room to collect indoor temperature data, and building data, including window area, is obtained from the property database;

[0091] Based on the outdoor temperature data, indoor temperature data and window area, the external heat input value is calculated as follows:

[0092] ;

[0093] in, is the external heat input value (W), is the heat transfer coefficient (W / m2·K), is the window area (square meters), is the temperature difference between indoor and outdoor (in degrees Celsius); for example, if the window area is 5 m2, the temperature difference between indoor and outdoor is 4 degrees Celsius, and the heat transfer coefficient is 1 W / m2·K, the external heat input is calculated to be about 200 W;

[0094] The Kalman filter algorithm is used to fuse the personnel position coordinates, pollutant concentration data, indoor temperature data, outdoor temperature data, window area, and external heat input value to generate a comprehensive environmental state vector, as follows:

[0095] Define the personnel position coordinates (converted to personnel density), pollutant concentration data, indoor temperature data, outdoor temperature data, window area and external heat input value as the initial environmental state vector;

[0096] Calculate the noise covariance for the dynamic data in the initial environment state vector. The window area is the static data, and the noise covariance is 0, which remains fixed.

[0097] Noise covariance example:

[0098] The occupant density is 6 people measured by the LiFi positioning function. This is compared with the calibration value of 7 people. Dividing this by 10 square meters gives a density error of 0.1 people per square meter, and a noise covariance of 0.01 (people per square meter squared).

[0099] The PM2.5 concentration is based on the pollutant sensor measurement value of 40 micrograms per cubic meter. Compared with the calibration value of 42 micrograms per cubic meter, the error is 2 micrograms per cubic meter, and the noise covariance is determined to be 4 (micrograms per cubic meter squared).

[0100] The VOC concentration is based on the pollutant sensor measurement value of 0.5 mg / m3, which is compared with the calibration value of 0.54 mg / m3, resulting in an error of 0.04 mg / m3 and a noise covariance of 0.0016 (milligrams per cubic meter squared).

[0101] The CO2 concentration is based on the pollutant sensor measurement value of 850ppm, which is compared with the calibration value of 910ppm, resulting in an error of 60ppm and a noise covariance of 3600 (ppm squared).

[0102] The indoor temperature is measured by the temperature sensor at 24 degrees Celsius. Compared with the calibration value of 24.7 degrees Celsius, the error is 0.7 degrees Celsius, and the noise covariance is 0.49 (squared of degrees Celsius).

[0103] The outdoor temperature is measured using the internet weather API as 28 degrees Celsius. Compared to the calibrated value of 29 degrees Celsius, the error is 1 degree Celsius, and the noise covariance is determined to be 1 (degrees Celsius squared).

[0104] The measured value of the external heat input is 200W, the calibrated value of the external heat input is 210W, the error is 10W, and the noise covariance is determined to be 100 (W)²;

[0105] The window area is obtained from the property database as a fixed value of 5 square meters. Compared with the calibration value of 5 square meters, there is no error, and the noise covariance of the window area is determined to be 0 (square meters squared);

[0106] Based on the Kalman filter algorithm, the posterior state estimate of the previous moment (the comprehensive environmental state vector optimized by the Kalman filter at the previous moment) is predicted and updated with the multi-source observation data at the current moment (the observation data at the current moment, including the coordinates of the person's location, pollutant concentration, indoor temperature, outdoor temperature, and external heat input). The observation error is calculated, and the Kalman gain is allocated based on the observation error and noise covariance. The dynamic data and static data are fused through weighted averaging to generate the comprehensive environmental state vector at the current moment.

[0107] Build a load forecasting model based on the LSTM model to generate load forecasts;

[0108] The load forecasting model is defined as an input layer, two LSTM layers, a fully connected layer, and an output layer;

[0109] The input layer inputs indoor monitoring data and comprehensive environmental state vectors. Through minimum-maximum normalization, the indoor monitoring data and comprehensive environmental state vectors are mapped to [0, 1] to eliminate dimensional differences. Through a sliding window, the indoor monitoring data and comprehensive environmental state vectors are divided into 12 time series.

[0110] The first LSTM layer (64 dimensions) processes 12 time series through the LSTM gating mechanism (forget gate, input gate, output gate) to capture long-term trends;

[0111] The forget gate uses the sigmoid function to decide to forget the irrelevant time series among the 12 time series;

[0112] The input gate updates the cell state of the current LSTM layer through sigmoid and tanh to incorporate the new time series;

[0113] The output gate generates the 64-dimensional hidden state of the current LSTM layer through sigmoid and tanh, which represents the long-term trend of the 12 time series (such as CO2 800-900ppm and heat load 600 watts / square meter mode);

[0114] The second LSTM layer (32 dimensions) again uses the LSTM gating mechanism to focus on the short-term fluctuations of the 12 time series and extract key features (such as the short-term stability of heat load, the slight increase in PM2.5, the stability of VOC, and the short-term growth of CO2).

[0115] The forget gate uses the sigmoid function to filter out minor long-term features in the first LSTM layer (such as outdoor temperature and population density at earlier time steps in the 12 time series);

[0116] The input gate updates the cell state of the second LSTM layer through sigmoid and tanh, focusing on short-term fluctuations (such as PM2.5±0.5);

[0117] Integrate the long-term trend relationship and short-term fluctuations in the first LSTM layer;

[0118] The output gate generates a 32-dimensional hidden state vector, which represents short-term fluctuations and provides key information for prediction;

[0119] The fully connected layer receives a 32-dimensional hidden state vector and maps the 32-dimensional vector to a 4-dimensional prediction target through a linear transformation;

[0120] The output layer outputs a 6×4-dimensional load forecast (heat load, PM2.5, VOC, and CO2) for the next hour through denormalization. 6 represents the six time steps in the next hour (one every 10 minutes, for a total of 60 minutes), corresponding to the time resolution of the forecast. 4 represents the four prediction targets for each time step, corresponding to the 4-dimensional prediction target of the fully connected layer.

[0121] Run the load forecasting model to predict the heat load and pollutant concentration trends for the next hour and generate a load forecast;

[0122] It should be noted that the input layer inputs indoor monitoring data and comprehensive environmental state vectors because the two provide complementary information. The comprehensive environmental state vector is high-precision fusion data optimized by Kalman filtering, which is suitable for long-term trend prediction; the indoor monitoring data (after preprocessing) retains real-time fluctuation details and is suitable for capturing short-term changes. The LSTM model uses the difference between the two to improve prediction robustness and accuracy.

[0123] S4. Through reinforcement learning algorithms, load forecasting, indoor monitoring data and energy-saving optimization schemes are used to dynamically adjust central air-conditioning parameters to optimize pollutant control, heat balance and energy consumption of central air-conditioning operating parameters.

[0124] Furthermore, through reinforcement learning algorithms, combined with load forecasting, indoor monitoring data, and energy-saving optimization plans, central air conditioning operating parameters (including compressor frequency to control cooling intensity, air volume to adjust air circulation, ventilation angle to determine coverage area, and purifier power to treat pollutant concentration) are dynamically adjusted to achieve the multi-objective goals of pollutant control, heat balance, and energy optimization, maintaining indoor comfort levels (such as temperature 22-26°C, PM2.5 <35µg / m³, VOC <0.6mg / m³, and CO2 <1000ppm), as follows:

[0125] Define the agent and environment. The agent runs on the edge computing device, observes the environmental status, and decides on the air conditioning parameter adjustment. The environment refers to the control area divided in the conference room and the corresponding central air conditioning equipment, including the air outlet, return air outlet, and related control components, which are used to respond to the parameter adjustment instructions of the agent.

[0126] Define the comprehensive environmental state vector, load forecast, indoor monitoring data and energy-saving optimization plan as the state space to provide a decision basis for the intelligent agent;

[0127] Define the action space as the adjustable central air conditioning operating parameters;

[0128] Set initialization strategy based on energy-saving optimization plan, load forecast and external heat input value;

[0129] The initialization action for high-priority areas (such as area A) is high air volume (such as 2.0m / s) and high purification power (such as 100%);

[0130] The initialization action for low-priority areas (such as area B) is low air volume (such as 0.5m / s) and low purification power (such as 50%);

[0131] If the predicted heat load is high (e.g. 600W / m²) or the external heat input value is large (e.g. 198W), the initialization action tends to be forced cooling;

[0132] Through the reward function, the central air-conditioning operating parameters are controlled and multiple objectives are balanced;

[0133] If all pollutant concentration values ​​in the pollutant concentration data are within the normal range (e.g. PM2.5 < 35µg / m³, VOC < 0.6mg / m³, CO2 < 1000ppm), then the reward is positive (+1). If any pollutant concentration value in the pollutant concentration data exceeds the normal range (PM2.5 > 35µg / m³, VOC > 0.6mg / m³, CO2 > 1000ppm), then a negative reward is given, with a deduction of 0.5 for each pollutant exceeding the standard, and the negative rewards are accumulated. For example, a 0.5 deduction for PM2.5 exceeding the standard, and a 1.0 deduction for PM2.5 and VOC exceeding the standard;

[0134] If the heat load of the controlled area is close to the average heat load (calculated based on the number of people, such as 300W / m²), the reward is positive (+1). If the heat load of the controlled area deviates greatly from the average heat load, a negative reward (-0.5 for every 100W / m² deviation) is given as a penalty.

[0135] Penalties are set for compressor frequency, air volume, and purifier power. For every 10Hz increase in compressor frequency, 0.5m / s air volume, and 20% increase in purifier power, the reward is deducted by 0.1. The total penalty multiplied by the weight is 0.2. For example, a 60Hz compressor frequency deducts 0.6, an air volume deducts 0.4, and a 100% purifier power deducts 0.5, for a total deduction of 0.3.

[0136] If the temperature is controlled at 22-26°C, a positive reward (+1) will be given, and if it deviates from 22-26°C, a negative reward (0.2 for every 0.5°C deviation) will be given as a penalty;

[0137] The intelligent agent reads the current state space from the environment through edge computing devices, integrates the comprehensive environmental state vector, prediction data, indoor monitoring data, and energy-saving optimization solutions to provide a basis for selecting actions;

[0138] The agent generates optimized actions based on the current state and initialization strategy through the actor network. For example, if the state of control area A shows high heat load and excessive PM2.5, the initialization recommendation is to deliver high air volume. The actor network selects a compressor frequency of 60Hz, an air volume of 2.0m / s, an angle of 50°, and a purifier power of 100% to achieve cooling, purification, and personnel coverage.

[0139] The actor network is a neural network in the deep learning algorithm. It is responsible for generating optimized actions based on the current state and initialization strategy. By learning the reward function, it outputs continuously optimized action parameters to control the central air conditioning.

[0140] The intelligent agent sends optimization actions to the environment through the controller (software in the edge computing device), adjusting the compressor frequency, air volume, air outlet angle, and purifier power, such as implementing high air volume to cope with heat load;

[0141] After the environment performs the action, it recollects indoor monitoring data, generates the next state, and gives rewards according to the reward function;

[0142] Set a time (e.g. 5 minutes) as the termination condition. If the termination condition is reached, a new optimization action is triggered to complete the adjustment of the central air conditioning operating parameters.

[0143] The intelligent agent integrates the current optimization action and the next state to generate optimization results (including compressor frequency, air volume, ventilation angle, purifier power, pollutant concentration and temperature), which are stored in the edge computing device.

[0144] This embodiment also provides a central air-conditioning intelligent energy-saving management and control system, including:

[0145] An acquisition module is used to collect and pre-process indoor monitoring data, wherein the indoor monitoring data includes personnel location coordinates and pollutant concentration data;

[0146] The partitioning module is used to divide the control area based on the processed indoor monitoring data through heat balance control technology, and use the airflow simulation algorithm to optimize the ventilation direction and air volume level in the control area, reduce the concentration of pollutants and balance the heat distribution of people, and generate energy-saving optimization solutions;

[0147] The prediction module is used to collect multi-source data, fuse the multi-source data with indoor monitoring data using the Kalman filter algorithm, generate a comprehensive environmental state vector, build a load prediction model, and predict future heat load and pollutant concentration trends;

[0148] The adjustment module is used to dynamically adjust central air-conditioning parameters through reinforcement learning algorithms, using load forecasting, indoor monitoring data and energy-saving optimization schemes to optimize pollutant control, heat balance and energy consumption of central air-conditioning operating parameters.

[0149] This embodiment also provides a computer device suitable for the intelligent energy-saving management and control method of central air conditioning, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent energy-saving management and control method of central air conditioning proposed in the above embodiment.

[0150] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0151] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the program implements the intelligent energy-saving management and control method for a central air conditioner as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0152] In summary, this paper achieves accurate predictions of future heat loads and pollutant concentrations by constructing a load forecasting model based on LSTM. This model analyzes time series using two layers of LSTM, capturing long-term trends and short-term fluctuations, and outputs prediction results for each time step. This model provides a forward-looking basis for dynamic central air conditioning control, making it suitable for complex indoor scenarios with frequent personnel movements and changing external environments. It also avoids air quality deterioration or heat imbalances caused by delayed response. Furthermore, the prediction results support dynamic adjustment of energy-saving plans, prioritizing resource allocation to high-demand areas and reducing inefficient operation time.

[0153] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A central air-conditioning intelligent energy-saving management and control method, characterized by: include, Collecting and preprocessing indoor monitoring data, including personnel location coordinates and pollutant concentration data; Thermal balance control technology divides control areas based on processed indoor monitoring data, and uses airflow simulation algorithms to optimize ventilation direction and air volume levels in control areas, reducing pollutant concentrations and balancing heat distribution among occupants, generating energy-saving optimization solutions. Collect multi-source data, use the Kalman filter algorithm to fuse multi-source data and indoor monitoring data, generate a comprehensive environmental state vector, build a load forecasting model, and predict future heat load and pollutant concentration trends; Through reinforcement learning algorithms, load forecasting, indoor monitoring data and energy-saving optimization schemes are used to dynamically adjust central air-conditioning parameters to optimize pollutant control, heat balance and energy consumption of central air-conditioning operating parameters; The heat balance control technology is used to divide the control area based on indoor monitoring data, and the specific steps are: Edge computing devices scan the coordinates of people in each control area, count the number of people, calculate the average number of people in the meeting room, assess the heat load in crowded areas, set heat thresholds based on the average number of people, and determine high-heat and low-heat areas. Compare pollutant concentration data with the normal concentration range of the preset spectral database, classify the pollution degree of the control area, and determine the high pollution area and low pollution area; Based on the classification of heat zones and pollution levels, the purifiers in the control areas are prioritized, and high-priority areas and low-priority areas are marked; Optimizing the ventilation direction and air volume level of the control area using an airflow simulation algorithm involves constructing a three-dimensional model of the conference room, dividing the air flow grid, mapping pollutant concentration data and occupant heat, analyzing the influence of outlet wind speed and outlet angle, calculating the air velocity and pollutant diffusion rate at the grid points in the air flow grid, adjusting the ventilation direction and air volume level, optimizing the air flow field, reducing pollutant concentration, and balancing occupant heat distribution; The Kalman filter algorithm is used to fuse multi-source data and indoor monitoring data. The specific steps are: Calculate external heat input value based on outdoor temperature data, indoor temperature data and window area; Define the personnel position coordinates, pollutant concentration data, indoor temperature data, outdoor temperature data, window area and external heat input value as the initial environmental state vector; Calculate the noise covariance for the dynamic data in the initial environment state vector, while the static data remain fixed; The dynamic data includes personnel location coordinates, pollutant concentration data, indoor temperature data, outdoor temperature data and external heat input value; The static data includes window area; Based on the Kalman filter algorithm, the posterior state estimate of the previous moment is fused with the multi-source observation data of the current moment to generate the comprehensive environment state vector of the current moment.

2. The central air-conditioning intelligent energy-saving management and control method according to claim 1, characterized in that: The load forecasting model is constructed based on the LSTM model, and an input layer, an LSTM layer, a fully connected layer, and an output layer are defined. The input layer processes indoor monitoring data and a comprehensive environmental state vector through standardization and a sliding window to generate a time series. The LSTM layer captures the long-term trend and short-term fluctuation of the time series. The fully connected layer maps the prediction target. The output layer generates a load forecast.

3. The central air-conditioning intelligent energy-saving management and control method according to claim 1, characterized in that: The reinforcement learning algorithm is to define the intelligent agent and environment, set the state space and action space, and optimize the energy consumption of pollutant control, heat balance and central air-conditioning operating parameters based on the initialization strategy and reward function.

4. The central air-conditioning intelligent energy-saving management and control method according to claim 1, characterized in that: The multi-source data includes outdoor temperature data, outdoor temperature data and window areas.

5. A central air-conditioning intelligent energy-saving management and control system, based on the central air-conditioning intelligent energy-saving management and control method according to any one of claims 1 to 4, characterized in that: include, An acquisition module is used to collect and pre-process indoor monitoring data, wherein the indoor monitoring data includes personnel location coordinates and pollutant concentration data; The partitioning module is used to divide the control area based on the processed indoor monitoring data through heat balance control technology, and use the airflow simulation algorithm to optimize the ventilation direction and air volume level in the control area, reduce the concentration of pollutants and balance the heat distribution of people, and generate energy-saving optimization solutions; The prediction module is used to collect multi-source data, fuse the multi-source data with indoor monitoring data using the Kalman filter algorithm, generate a comprehensive environmental state vector, build a load prediction model, and predict future heat load and pollutant concentration trends; The adjustment module is used to dynamically adjust central air-conditioning parameters through reinforcement learning algorithms, using load forecasting, indoor monitoring data and energy-saving optimization schemes to optimize pollutant control, heat balance and energy consumption of central air-conditioning operating parameters.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the central air-conditioning intelligent energy-saving management and control method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the central air-conditioning intelligent energy-saving management and control method according to any one of claims 1 to 4 are implemented.

Images