Real-time Dynamic Estimation Of Occupancy Load And An Air-conditioning Predictive Control Method Based On Image Information Fusion

Building energy consumption accounts for more than 30% of the global total energy consumption,and the energy consumption of air conditioning system occupies the first place in building energy consumption.In the current research on optimal operation and regulation of air conditioning in public buildings,the influence of personnel change and regional distribution on system dynamic regulation and energy consumption has attracted more and more attention.Due to the limitation of control strategy,the acquisition of personnel load is mostly based on empirical judgment and presetting,which makes it impossible to realize real-time detection of the change of personnel load inside the building.At the same time,due to the thermal inertia of buildings,indoor heat diffusion process,the existence of delay,such as gas inertia changes than indoor personnel load,air conditioning system temperature sensor is a fairly long time lag,cause the system real-time regulating action to follow the indoor load changes,the serious influence of the air conditioning system running effect.Therefore,this paper USES the computer vision technology to take the image information as the "soft sensor" for the detection of the personnel inside the building,so as to realize the real-time estimation of the personnel load.In addition,the influence factor of personnel load is introduced into the control of air conditioning system to establish a predictive control strategy of air conditioning system based on the estimation of personnel load and optimize the operation process of the system,which isof great significance for guiding the energy saving optimization of air conditioning system and reducing building energy consumption.This paper will conduct research from the following aspects:(1)Aiming at the problem that it is difficult to achieve the real-time estimation in the process of estimating the internal personnel load of the building,which leads to the response lag of the air-conditioning system,this paper proposes a real-time estimation model of the construction personnel load based on the image information.Firstly,an end-to-end space personnel detection model based on convolutional neural network is established by using computer vision and deep learning technology to collect the image information inside the building through the camera.Secondly,the personnel load estimation model is established by using the personnel detection model to realize the real-time estimation of the personnel load and its variation,and to improve the deficiency of the traditional personnel load estimation method.Finally,through experimental analysis,the real-time estimation model of construction personnel load based on image information is proposed in this paper,and the real-time detection and detection accuracy are greatly improved compared with the traditional detection method.(2)The accuracy of personnel detection determines the performance of personnel load estimation.For complex building environments such as shopping malls and stations,where the density of personnel changes when the density is high,the performance of the head detection method needs to be further improved.Therefore,this paper puts forward a kind of High and low density HSDnet space perspective people counting convolution neural network model(High and low spatial perspective density convolutional neural network,HSDnet),solve the head detection methods are susceptible to keep out of the single,Camera Angle change and other factors affect the problem.Firstly,the Density Map Estimation Module(DME)is used to extract the features of the input images to obtain the Density regression Map of the images.Secondly,according to the color change of the density map,use the idea of segmentation to divide the density map Density area division,get the boundary of the high and low density area,map to the original image,get the high density module map and the low density module map.Finally,by using different crowd counting methods,the high-density module map was counted by density map regression,while the low-density module map was counted by direct head detection.Through the experimental analysis,it can be seen that the HSDnet model combines the detection method with the density map regression method to improve the defects of the single personnel counting method,improve the personnel detection accuracy,and the model has higher robustness and accuracy.(3)In view of the inertia of the gas and the inherent delay of the sensor during the dynamic adjustment of the air conditioning system,the system response is difficult to follow the real-time change of the personnel load,resulting in high output energy consumption and low-quality internal environment problems in the building A predictive control strategy for air-conditioning system based on real-time estimation of personnel load.First,establish a personnel detection model based on building image information to calculate the real-time change in personnel load;secondly,introduce a personnel load control factor to predict the indoor temperature change trend brought by the load change;finally,calculate the system compensation cold capacity and adjust the air conditioning system cold supply,Improve the indoor environment quality problems caused by the large hysteresis under the conventional control method,save energy and reduce consumption.Simulation results show that compared with conventional PID control,the temperature control effect of predictive control is higher,and the energy saving effect can reach 6.70%.At the same time,when the amount of personnel load change is larger,the predictive control strategy shows better control effect,the system responds faster,the higher the temperature control performance,the more obvious the energy saving effect.