Air conditioning unit fresh air volume adjusting method for high-density crowd building space-time perception prediction

Abstract

The invention relates to an air conditioning unit fresh air volume adjusting method for high-density crowd building space-time perception prediction, and the method comprises the steps: carrying out the weighted integration of an LSTM time sequence prediction result and a ConvLSTM space-time prediction result, so as to calculate the target fresh air volume of an air conditioning unit, and carrying out the regulation and control of the air volume through the balance time sequence and space-time prediction results. According to the method, the periodical law and short-term spatial-temporal change of the flow of people can be fully utilized, it is ensured that regulation and control of the fresh air volume conform to the long-term trend, instant fluctuation can be quickly responded, the dynamic nature and accuracy of regulation and control of the fresh air volume are ensured in the process, meanwhile, unnecessary energy consumption is reduced, and it is ensured that the air quality is stabilized at the appropriate level; and the per capita fresh air volume is combined with the identified building type or business state, and the per capita minimum fresh air volume meeting the national existing standard requirements is automatically selected in a matched mode.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method for adjusting the fresh air volume of an air-conditioning unit based on spatiotemporal perception prediction of a building with high-density population. Background Art

[0002] The amount of fresh air required by the air conditioning system should be determined based on the requirements of indoor air quality, the activities of personnel, the nature and time of use of the building, the pollution source and the condition of the building. Usually, when designing most air conditioning systems, the amount of fresh air is selected based on the minimum fresh air volume that only meets the hygiene requirements of personnel. Generally, carbon dioxide is used as an indicator to keep its daily average concentration within 1000ppm to determine the minimum fresh air volume. However, this method may lead to deviations, and the delayed feedback of carbon dioxide concentration and other indoor pollutants should also be considered.

[0003] The fresh air volume of conventional air conditioning units is mainly based on the average flow of people in the service area. However, for buildings with high population density, such as shopping malls, supermarkets, museums, exhibition halls, waiting halls, etc., the flow of people varies greatly, the duration of peak flow is short, and it is significantly affected by factors such as work and rest, holidays, seasons, and climate. If the fresh air volume cannot be adjusted in time, it cannot ensure that the indoor air quality meets the hygiene requirements of personnel, the comfort of the environment where personnel live cannot be guaranteed, and energy conservation cannot be maximized according to the changes in personnel density.

[0004] According to relevant information, the air volume of some fresh air fans is mainly regulated based on real-time crowd flow data, and fails to consider the temporal law of crowd flow and short-term spatiotemporal changes. This leads to the lag of air volume regulation and cannot effectively adapt to the changes in demand during peak hours or holidays. Existing methods mainly focus on the spatial distribution of real-time crowds and the average crowd flow, but lack the prediction of periodic fluctuations and short-term spatiotemporal changes. Summary of the invention

[0005] To solve the above problems, a method for adjusting the fresh air volume of air-conditioning units in high-density buildings based on spatiotemporal perception and prediction was proposed. Through real-time video stream collection, spatiotemporal prediction and weighted air volume control, intelligent fresh air volume adjustment was achieved to ensure indoor air quality and comfort while improving energy efficiency.

[0006] The technical solution of the present invention is: a method for adjusting the fresh air volume of an air conditioning unit based on spatiotemporal perception prediction in a high-density building, which specifically comprises the following steps:

[0007] 1) Video stream collection and target detection: The cameras in the service area of ​​the air-conditioning unit continuously collect video stream data, which is input into the YOLOv5 model for target detection to identify the number and location distribution of people in the service area;

[0008] 2) Time series prediction of personnel: Use step 1) to identify the flow of people in the past week, use the LSTM time series model to model the time series and train and optimize it. After training, the LSTM time series model inputs the number of people in the past week and the time characteristics, and outputs the predicted flow of people every hour in the next week; 3) Spatiotemporal density prediction of personnel: The ConvLSTM spatiotemporal model inputs the regional flow density distribution map and time characteristics obtained by YOLOv5 target detection for training and optimization. After training, the ConvLSTM spatiotemporal model inputs the identified flow data within a day, and outputs the spatial distribution of flow density at different time points in the future day;

[0009] 4) Weighted air volume control: The predicted data results of each day and hour of this week predicted by the LSTM time series model based on the previous week’s passenger flow data are weighted and integrated with the predicted data results of each hour of the day predicted by the ConvLSTM spatiotemporal model based on yesterday’s passenger flow data to calculate the target fresh air volume of the air-conditioning unit; based on the target air volume obtained by weighted calculation, the system instantly adjusts the opening of the fresh air electric valve to achieve fresh air volume control in the service area of ​​the air-conditioning unit.

[0010] Furthermore, the specific method of fresh air volume control in step 4) is as follows: the fresh air volume target value directly affects the actual fresh air output through the change in the opening of the fresh air electric valve, and the actual output effect of the fresh air volume is monitored through a feedback mechanism during the control process, and the air quality and crowd density data collected in real time are compared. If the deviation between the actual situation and the prediction is greater than the set value, the weighting coefficient, the prediction model, or the weighting coefficient and the prediction model are dynamically adjusted simultaneously to ensure the accuracy and continuity of the control.

[0011] Furthermore, the target fresh air volume in step 4) is combined with the building type or business format of the air-conditioning unit service area identified in step 1) to automatically adapt and select the minimum fresh air volume per capita that meets the requirements of existing specifications.

[0012] Furthermore, in step 1), the camera continuously collects video stream data: the camera installation position ensures full coverage of the service area and accurately obtains real-time data of pedestrian flow. The collected video stream data is processed through a computer vision framework and converted into analyzable image frames to ensure that each frame is clear and effective, providing basic data for subsequent pedestrian flow identification and statistics.

[0013] Furthermore, the LSTM time series model in step 2) captures the periodic change characteristics of the flow of people, analyzes and predicts the changing rules of the flow of people in different time periods, and adjusts the air volume in advance before the peak period arrives to achieve on-demand control.

[0014] Furthermore, the ConvLSTM spatiotemporal model in step 3) captures the dynamic distribution changes of the flow of people in space in different time periods, makes a detailed analysis of the short-term fluctuations in the density of people flow, and responds to the fluctuations in the density of people flow in a timely and accurate manner.

[0015] The beneficial effects of the present invention are as follows: the method for adjusting the fresh air volume of air-conditioning units based on spatiotemporal perception prediction in high-density crowd buildings of the present invention balances the time series and spatiotemporal prediction results to regulate the air volume, and can make full use of the periodic laws of human flow and short-term spatiotemporal changes to ensure that the regulation of the fresh air volume is in line with the long-term trend and can quickly respond to immediate fluctuations. This process ensures the dynamics and accuracy of the fresh air volume regulation, while reducing unnecessary energy consumption and ensuring that the air quality is stable at an appropriate level; accordingly, the fresh air volume of the air-conditioning unit is adjusted in real time and dynamically, and the per capita fresh air volume is automatically adapted to select the per capita minimum fresh air volume that meets the requirements of existing national standards in combination with the aforementioned identified building type or business format. DETAILED DESCRIPTION

[0016] The present invention is described in detail below in conjunction with specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0017] The method for adjusting the fresh air volume of an air conditioning unit based on time-space perception in a high-density crowd building of the present invention specifically comprises the following steps:

[0018] Step 1: Video stream acquisition and target detection

[0019] 1. Video stream acquisition: The cameras in the service area of ​​the air-conditioning unit continuously collect video stream data. The cameras are installed in appropriate locations to ensure full coverage of the service area and to accurately obtain real-time data on the flow of people. The collected video stream data is processed through a computer vision framework and converted into analyzable image frames to ensure that each frame is clear and effective, providing basic data for subsequent identification and statistics of people flow.

[0020] 2. Target detection (based on YOLOv5): Video stream data is input into the YOLOv5 model for target detection. The YOLOv5 model uses a deep learning algorithm to quickly and efficiently identify the number and location distribution of people in the service area. The detection results are converted into real-time crowd density information in different sub-areas and recorded in the database. The purpose of this step is to provide accurate basic crowd flow data for subsequent time series and spatiotemporal predictions.

[0021] Step 2: Time Series Forecast of Personnel

[0022] 1. Construction of LSTM time series model: Using the historical passenger flow data within a week, the LSTM time series (long short-term memory network) model is used to model the time series. The LSTM time series model can capture the periodic change characteristics of personnel flow and is suitable for the analysis and prediction of the change rules of personnel flow in different time periods. After the model is built, it is continuously updated and optimized using real-time passenger flow data to ensure the accuracy of the prediction results.

[0023] 2. LSTM time series prediction process: The input of the LSTM time series model is the number of people data in the past week and time characteristics, such as hour and holiday marks. The output of the model is the predicted data of the flow of people every hour in the next week. Through this prediction, the system can adjust the air volume in advance before the peak period arrives to achieve on-demand control. This process ensures that the air conditioning system can be fully prepared when high-density people flow arrives, avoiding the decrease in comfort and possible increase in energy consumption due to delayed response.

[0024] Step 3: Prediction of the spatiotemporal density of people

[0025] 1. Construction of ConvLSTM spatiotemporal model: The ConvLSTM spatiotemporal model combines convolutional neural networks and long short-term memory networks to process the spatiotemporal density changes of people flow within a day. The input data of the ConvLSTM spatiotemporal model is the regional crowd density distribution map and time characteristics obtained by YOLOv5 target detection. Through this model, the dynamic distribution changes of people flow in space in different time periods can be captured at the same time, which is particularly suitable for the detailed analysis of short-term crowd density fluctuations.

[0026] 2. ConvLSTM spatiotemporal prediction process: The output of the ConvLSTM spatiotemporal model is the spatial distribution of the density of people flow at different time points in the future day. Based on the prediction results of each sub-area, the predicted fresh air volume of the air-conditioning unit is accumulated to achieve accurate air volume control. This process ensures that the system can respond to short-term fluctuations in the density of people flow in a timely and accurate manner.

[0027] Step 4: Weighted air volume control

[0028] 1. Implementation of weighted strategy: Weighted integration of the predicted data results of the previous week's passenger flow data through LSTM time series prediction and the predicted data results of yesterday's passenger flow data through ConvLSTM spatiotemporal model prediction for each hour of the day to calculate the target fresh air volume of the air-conditioning unit. The weighted calculation formula is as follows: W_i(t)=α·LSTM_t+β·ConvLSTM_{i,t}

[0029] Among them, W_i(t) is the target fresh air volume of sub-area i at time t after the air-conditioning unit service area is divided into several sub-areas; α and β are weighting coefficients, which are used to balance the contribution of time series and spatiotemporal prediction results to air volume control. This weighting strategy can make full use of the periodic law and short-term spatiotemporal changes of the flow of people, ensuring that the control of fresh air volume is in line with the long-term trend and can quickly respond to immediate fluctuations.

[0030] 2. Air volume control execution: According to the target air volume obtained by weighted calculation, the system immediately adjusts the opening of the fresh air electric valve to achieve fresh air volume control in the service area of ​​the air conditioning unit. The fresh air volume in the service area of ​​the air conditioning unit is adjusted in advance according to the predicted change trend. When the predicted results show that the density of human traffic will increase, the system automatically increases the air volume in the corresponding area; when the predicted density of human traffic shows a decrease, the system automatically reduces the air volume output. This process ensures the dynamic and accurate control of fresh air volume, while reducing unnecessary energy consumption and ensuring that the air quality is stable at an appropriate level.

[0031] Step 5: Fresh air output control

[0032] 1. Input conversion of the fresh air system: The fresh air volume target value directly affects the actual fresh air output through the change in the opening of the fresh air electric valve. The system converts the weighted air volume result into the specific opening control parameter of the electric valve to ensure that the actual fresh air output of the air-conditioning unit is consistent with the predicted target air volume. The purpose of this is to ensure that the fresh air supply in the service area is adapted to the population density, improve the overall energy utilization efficiency, reduce energy waste, and reduce the workload of the operation and maintenance department of buildings with high population density. The per capita fresh air volume is automatically adapted to the minimum per capita fresh air volume that meets the requirements of the existing national specifications in combination with the aforementioned identified building type or business format.

[0033] 2. Feedback and optimization of fresh air volume control: During the control process, the system monitors the actual output effect of fresh air volume through the feedback mechanism, and compares the real-time collected air quality with the crowd density data. If the actual situation deviates from the prediction by more than the set value, the weighting coefficient, prediction model, or the weighting coefficient and prediction model are dynamically adjusted to ensure the accuracy and sustainability of the control. This optimization mechanism improves the system's adaptive ability and makes the fresh air volume control more intelligent.

[0034] The above-mentioned embodiments only express the specific implementation of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. A method for adjusting the fresh air volume of an air conditioning unit based on spatiotemporal perception prediction in a high-density building, characterized in that: The specific steps include: 1) Video stream collection and target detection: The cameras in the service area of ​​the air-conditioning unit continuously collect video stream data, which is input into the YOLOv5 model for target detection to identify the number and location distribution of people in the service area; 2) Time series prediction of personnel: Use step 1) to identify the passenger flow data within a week in history, use the LSTM time series model to model the time series and train and optimize it. After training, the LSTM time series model inputs the number of personnel data and time characteristics within a week in history, and outputs the passenger flow prediction data for each hour in the next week; 3) Prediction of the spatiotemporal density of people: The ConvLSTM spatiotemporal model inputs the regional crowd density distribution map and time characteristics obtained by YOLOv5 target detection for training and optimization. After training, the ConvLSTM spatiotemporal model inputs the recognized crowd flow data within one day and outputs the spatial distribution of crowd density at different time points in the future day; 4) Weighted air volume control: The predicted data results of each day and hour of this week predicted by the LSTM time series model based on the previous week’s passenger flow data are weighted and integrated with the predicted data results of each hour of the day predicted by the ConvLSTM spatiotemporal model based on yesterday’s passenger flow data to calculate the target fresh air volume of the air-conditioning unit; based on the target air volume obtained by weighted calculation, the system instantly adjusts the opening of the fresh air electric valve to achieve fresh air volume control in the service area of ​​the air-conditioning unit.

2. The method for adjusting fresh air volume of air conditioner units based on spatiotemporal perception prediction in high-density crowd buildings according to claim 1 is characterized in that: The specific method of fresh air volume control in step 4) is as follows: the fresh air volume target value directly affects the actual fresh air output through the change in the opening of the fresh air electric valve. During the control process, the actual output effect of the fresh air volume is monitored through a feedback mechanism, and the air quality and crowd density data collected in real time are compared. If the deviation between the actual situation and the prediction is greater than the set value, the weighting coefficient, the prediction model, or the weighting coefficient and the prediction model are dynamically adjusted simultaneously to ensure the accuracy and continuity of the control.

3. The method for adjusting fresh air volume of air conditioner units based on spatiotemporal perception prediction of high-density crowd buildings according to claim 1 or 2, characterized in that: The target fresh air volume in step 4) is combined with the building type or business format of the air-conditioning unit service area identified in step 1), and automatically adapts and selects the minimum fresh air volume per capita that meets the requirements of existing specifications.

4. According to the method for adjusting fresh air volume of air conditioner units based on spatiotemporal perception prediction in high-density crowd buildings according to claim 3, it is characterized in that: In step 1) the camera continuously collects video stream data: the camera installation position ensures comprehensive coverage of the service area and accurately obtains real-time data of human flow. The collected video stream data is processed through a computer vision framework and converted into analyzable image frames to ensure that each frame is clear and effective, providing basic data for subsequent human flow identification and statistics.

5. According to claim 3, the method for adjusting fresh air volume of air conditioner units based on spatiotemporal perception prediction in high-density crowd buildings is characterized in that: The LSTM time series model in step 2) captures the periodic change characteristics of the flow of people, analyzes and predicts the changing rules of the flow of people in different time periods, adjusts the air volume in advance before the peak period arrives, and realizes on-demand regulation.

6. The method for adjusting fresh air volume of air conditioner units based on spatiotemporal perception prediction in high-density crowd buildings according to claim 3 is characterized in that: The ConvLSTM spatiotemporal model in step 3) captures the dynamic distribution changes of the flow of people in space in different time periods, makes a detailed analysis of the short-term fluctuations in the density of the flow of people, and responds to the fluctuations in the density of the flow of people in a timely and accurate manner.