Energy output power adjusting method and system based on artificial intelligence and big data

Abstract

The invention discloses an energy output power adjusting method based on artificial intelligence and big data. The method comprises the following steps of collecting environment data and personnel flow information in a space in real time; generating a people flow model through big data analysis and a machine learning algorithm; the operation state and the output power of the energy equipment are automatically adjusted according to the people flow model and the environment data, and precise environment control and energy saving are achieved. According to the invention, environment data is collected in real time through the data acquisition module, and an optimal energy adjustment strategy is generated by using big data analysis and a machine learning algorithm. The system has self-learning and self-adaptive capabilities, and can automatically optimize an adjustment strategy according to environmental changes and user habits, so that energy consumption is remarkably reduced, and user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent energy management technology, and in particular to an energy output power regulation method and system based on artificial intelligence and big data. Background Art

[0002] Heating, ventilation, and air conditioning (HVAC) systems account for over 40% of energy consumption in commercial and public buildings, with as much as 15%-30% of this energy wasted due to imbalances in supply and demand. The core issue lies in the fact that traditional environmental control systems generally employ static threshold control strategies (e.g., fixed temperature setpoints) that fail to respond to dynamic variables such as crowd density and behavioral patterns. When occupancy distribution exhibits temporal and spatial imbalances (e.g., sudden gatherings in meeting areas, vacant offices at night), the systems often fail due to over-powering (constantly outputting rated power in low-traffic areas) or lag control (relying on passive feedback from environmental parameters). This problem stems from a single-dimensional data perception system. Existing systems rely on basic sensors such as temperature, humidity, and CO2 concentration, but lack the ability to accurately perceive crowd dynamics. This leads to inherent flaws in single-point monitoring technology. Video analysis is affected by light obstruction, resulting in counting errors exceeding 20% in dense scenes. Wi-Fi probes cannot distinguish between devices and actual people, and their signals are susceptible to interference from building structures. Infrared sensors can only detect human presence, but struggle to capture movement trajectories and dwell time. This makes it impossible to construct a three-dimensional "people-space-time" model of heat load demand. Although technologies like mobile communications (cell phone signaling) and RFID positioning can supplement human trajectory data, multi-source data fusion faces algorithmic bottlenecks: It's difficult to establish a spatiotemporal alignment benchmark for devices with different sampling frequencies (e.g., 30fps video vs. 5s / second signaling). Pressure sensor mis-touch and infrared false alarms generate noise pollution. Existing technologies only count the number of people but fail to analyze dwell patterns (rapid passage vs. long-stay) and spatial thermal coupling (the localized temperature rise effect in crowded areas). This ultimately leads to a fundamental conflict between energy efficiency and comfort. Coarse-grained energy-saving strategies (e.g., a global temperature setpoint offset of ±2°C) lead to degraded thermal comfort in core functional areas (conference rooms, data centers). Furthermore, independent control of air conditioning, fresh air, and lighting systems lacks coordination, making joint optimization based on spatial thermal distribution impossible (e.g., when high-density areas require increased cooling, adjacent areas still maintain high-power air supply). Summary of the Invention

[0003] Purpose of the Invention: The first purpose of the present invention is to provide an energy output power regulation method based on artificial intelligence and big data. This method uses a data acquisition module to collect environmental data in real time, and uses big data analysis and machine learning algorithms to generate an optimal energy regulation strategy. This method can automatically optimize the regulation strategy based on environmental changes and user habits, significantly reducing energy consumption and improving user experience.

[0004] The second purpose is to provide an energy output power regulation system based on artificial intelligence and big data.

[0005] Technical solution: The energy output power regulation method based on artificial intelligence and big data provided by the present invention includes the following steps:

[0006] S1, real-time collection of environmental data and personnel flow information within the space;

[0007] S2, generate a crowd flow model through big data analysis and machine learning algorithms;

[0008] S3. Automatically adjust the operating status and output power of energy equipment according to the crowd flow model and environmental data to achieve precise environmental control and energy conservation.

[0009] Among them, S1 collects wireless signal strength, personnel movement trajectory, identity tags, visual images, environmental temperature and humidity, regional pressure changes and human thermal radiation data, and integrates them to build a three-dimensional monitoring network covering the physical environment status and human flow dynamics.

[0010] Furthermore, S2 receives and stores multi-source data from S1, and relies on the artificial intelligence platform to dynamically compare and pattern match real-time data with typical crowd flow models in the big data model group. It also trains prediction algorithms based on historical data to generate accurate prediction results of the spatiotemporal distribution of crowd flow.

[0011] Convert the analysis conclusions into control parameters to provide a decision-making basis for S3.

[0012] Furthermore, S3 uses the spatiotemporal distribution of crowd density predictions, combined with real-time ambient temperature and humidity data, to establish a dynamic cooling and heating load regulation model for the crowd-space-environment coupling. By analyzing the mapping relationship between crowd density and spatial heat load in different areas, it calculates the cooling and heating output requirements of the air conditioning system and dynamically adjusts the energy supply strategy based on real-time environmental parameters.

[0013] When the predicted crowd density is higher than the threshold and the regional temperature deviates from the set range, the variable frequency compressor is activated first to increase the cooling / heating power output. At the same time, the air supply direction and wind speed are adjusted according to the spatial heterogeneity of the occupant distribution to achieve precise control of local thermal comfort.

[0014] On the contrary, when it is detected that the crowd density is decreasing and the environmental parameters are in a comfortable range, the operating frequency of the air-conditioning host is automatically reduced and switched to low-power mode, and the basic temperature control requirements are maintained by gradient adjustment of the supply air volume and return water temperature.

[0015] Furthermore, the establishment of the crowd flow model in S2 includes the following steps:

[0016] Step 1: Construct a personnel information sample data packet matrix

[0017] Assume that there are m samples of people in the building, and each sample has q indicator explanatory variables, and the indicator variable Y = (y1, y2, y3, ... y q ), the characteristic value Y of the information η =(y η1 ,y η2 ,y η3 ,…y ηq ), establish the personnel information sample data matrix:

[0018]

[0019] In the formula, the sample data matrix Y belongs to the target of text classification, where Y ηi Describe the characteristic value of the ξ-th indicator in the η-th person information;

[0020] Step 2: Calculate the Euclidean distance between people, two person samples y g 、y f The similarity is determined by the Euclidean distance as follows:

[0021]

[0022] Among them, e gf Indicates person y g ,y f In the multidimensional feature space, smaller values of distance indicate more similar behavior patterns;

[0023] Step 3: According to the number of clusters, set the initial cluster center and cluster the personnel information samples. The clustering standard is: cluster m personnel information samples y g 、y f Assume d types, as follows

[0024]

[0025] The clustering standard matrix has the following properties, as shown in the formula:

[0026] v ηξ ∈{0,1} and

[0027] where v ηξ Indicates whether the sample ξ belongs to the category η;

[0028] Step 4: Calculate the class center and the difference between classes

[0029] The number of personnel information samples in the ηth category is m η :

[0030]

[0031] The central eigenvalue of the nth type of personnel information sample is

[0032]

[0033] The inter-class difference of the nth type of personnel information samples:

[0034]

[0035] The overall difference is:

[0036]

[0037] in represents the average behavioral characteristics of the ηth type of personnel, Z(v * ) is the advantage indicator of the clustering result, the smaller the value, the more reasonable it is;

[0038] Step 5: Iterative Optimization

[0039] By minimizing the overall difference Z(v * ), solve the optimal clustering result v * :

[0040] Z(v ＊ )=min{Z(v)}

[0041] When the number of iterations reaches the same as the set number of iterations, clustering is stopped and the personnel information classification results are obtained;

[0042] Step 6: Solve the crowd density model

[0043]

[0044] Where, T pred (t) is the predicted value of the output crowd density, Y is the description matrix; c is a vector, is the weight coefficient.

[0045] Furthermore, the S3 priority algorithm prioritizes core areas and prioritizes energy supply to areas with higher scores. This includes the following steps:

[0046] Step 1: Let the priority score P of the i-th region be i for:

[0047]

[0048] in, is the normalized crowd density, ΔT i is the absolute deviation between the current temperature and the set temperature, C i is the unit energy consumption cost, α1, β1, and γ1 are dynamic weight coefficients, reflecting the priority of the control target;

[0049] Step 2: Data normalization, mapping parameters of different dimensions to the interval 0,1 to ensure comparability:

[0050]

[0051] Among them D max , ΔT max 、C max is the preset maximum reference value;

[0052] Step 3: Adjust the weight coefficient according to time, season, and operation strategy:

[0053] Peak hours: Prioritize comfort, with weights set as:

[0054] α1=0.4,β1=0.5,γ1=0.1

[0055] Energy-saving period: focuses on reducing energy consumption costs, with weights set as:

[0056] α1=0.1,β1=0.3,γ1=0.6

[0057] Step 4: Calculate P for all regions i After the value is entered, it is sorted in descending order, and energy is allocated first to the area with higher priority:

[0058] Priority queue = Sort(P1,P2,...,P n )

[0059] Step 5: Score P based on priority i , dynamically adjust the device output power:

[0060] Air conditioning power distribution:

[0061] W i AC =W base +k·(P i -P threshold )

[0062] Among them, W i AC represents the electric power of the air conditioner in the i-th area, k is the gain coefficient, P threshold is the activation threshold; then the actual power output of each air conditioner is:

[0063]

[0064] in is the actual power output of the jth device in area i, and M represents the total number of adjustable air-conditioning devices in the system;

[0065] Exhaust fan speed:

[0066]

[0067] The speed is allocated according to the proportion of pedestrian density to ensure ventilation needs in high-density areas. The real-time exhaust volume of the fan is:

[0068]

[0069] in, is the real-time speed of the exhaust fan in the ith area, S max is the maximum rated speed of the exhaust fan, n is the total number of zones, k q is the air volume-speed ratio coefficient.

[0070] Furthermore, a multi-objective optimization objective function for efficient energy utilization is established, and energy optimization is performed according to target changes:

[0071]

[0072] Where α is the energy consumption optimization weight, E(t) is the total energy consumption of the system, β is the weight of the crowd flow prediction accuracy, N is the total number of sampling points in the time series, and T pred (t) is the crowd density predicted by the fusion of infrared sensor and pressure sensor, T real (t) is the actual crowd density, γ is the equipment control accuracy weight, M represents the total number of adjustable air-conditioning equipment in the system, is the actual power output of the jth device in region i, is the theoretical optimal power of the device, ΔH and ΔT are the allowable fluctuation range of ambient temperature and humidity, T(t) is the real-time ambient temperature, H(t) is the real-time ambient humidity, Q flow (t) is the real-time equipment air volume, θ press is the static pressure of the space, T set is the set ambient temperature, H set is the set ambient humidity, Q max is the maximum air volume, ∈ flow is the minimum fresh air volume coefficient, θ low is the minimum static pressure in the space, θ high is the maximum static pressure in the space.

[0073] Correspondingly, an energy output power regulation system based on artificial intelligence and big data is also disclosed, including: a data acquisition module, a data processing module and an intelligent control module;

[0074] The data acquisition module is used to collect environmental data and personnel flow information in real time;

[0075] The data processing module includes a data storage unit and a data analysis and processing unit, which generates a crowd flow model through big data analysis and machine learning algorithms;

[0076] The intelligent control module includes an environmental control system and an air conditioning energy supply system, which is used to automatically adjust the operating status and output power of energy equipment according to the crowd flow model and environmental data to achieve precise environmental control and energy conservation.

[0077] Beneficial effects: Compared with the prior art, the present invention has the following significant improvements:

[0078] 1. Through multi-dimensional data fusion and artificial intelligence prediction models, the output power of air conditioning, exhaust and other equipment is dynamically optimized to achieve on-demand energy supply;

[0079] 2. Based on the crowd density-spatial heat load coupling model, resources are dynamically allocated through a priority algorithm. While ensuring the stability of temperature and humidity in the core area, wind direction adjustment and local compensation technology are used to eliminate temperature deviations. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 Schematic diagram of the system of the present invention;

[0081] Figure 2 This is a flow chart of the crowd density algorithm of the present invention. DETAILED DESCRIPTION

[0082] like Figure 1 The energy output power regulation system based on artificial intelligence and big data shown in the figure includes an interconnected data acquisition module, a data processing module and an intelligent control module; the data acquisition module includes a temperature and humidity sensor, a pressure sensor, a video camera, an infrared sensor, a Wi-Fi probe, a mobile phone signaling and a radio frequency identification device, which is used to collect environmental data and personnel flow information in real time; the data processing module includes a data storage unit and a data analysis and processing unit, which generates a crowd flow model through big data analysis and machine learning algorithms; the intelligent control module includes an environmental control system and an air-conditioning power supply system, which is used to automatically adjust the operating status and output power of equipment such as air conditioners and exhaust fans according to the crowd flow model and environmental data, so as to achieve precise environmental control and energy saving.

[0083] The WiFi probes, mobile phone signaling, radio frequency identification, video cameras, temperature and humidity sensors, pressure sensors, and infrared sensors in the data acquisition module work together to collect multi-dimensional data such as wireless signal strength, personnel movement trajectory, identity tags, visual images, ambient temperature and humidity, regional pressure changes, and human thermal radiation. These data are aggregated and integrated through the "data collection" module pointed by the arrow to build a three-dimensional monitoring network covering the physical environment status and human flow dynamics. Among them, the video camera and WiFi probe cross-verify the human flow density, the infrared sensor and pressure sensor complement each other to eliminate interference, the mobile phone signaling and radio frequency identification are integrated to analyze the personnel behavior patterns, and the real-time temperature and humidity parameters are combined to form a high-reliability data stream, which provides accurate input for subsequent intelligent analysis and ultimately supports the system to dynamically adjust the power of air conditioning, ventilation and other equipment.

[0084] The data processing module receives and stores multi-source data (including temperature and humidity, pressure, infrared signals, Wi-Fi probes, mobile phone signals, video images, etc.) pre-processed and corrected by the data acquisition module through a database. Relying on the artificial intelligence platform, it dynamically compares and matches the real-time data with typical crowd flow models (such as peak hours, regional distribution, behavioral patterns, etc.) in the big data model group. It combines historical data to train the prediction algorithm to generate accurate prediction results of the spatiotemporal distribution of crowd flow. At the same time, the analysis conclusions (such as crowd density trends and changes in environmental demand) are converted into control parameters to provide a decision-making basis for the intelligent control module, driving air conditioning, exhaust equipment and other equipment to adjust their working status and power output as needed.

[0085] The intelligent control module converts the prediction model and control instructions output by the data processing module into device actions to achieve dynamic optimization of energy output: by receiving crowd density predictions, environmental comfort requirements (such as temperature and humidity thresholds) and energy consumption constraint parameters, combined with real-time feedback sensor data, the intelligent control module uses built-in priority algorithms (for example, prioritizing energy supply to core areas during peak hours) and multi-device collaboration strategies to dynamically adjust the cooling / heating power of the air conditioner, the speed of the exhaust fan, the brightness of the lighting system, etc. For example, when the crowd flow drops sharply, the air conditioner power is automatically reduced and switched to low-frequency air supply mode, or when it is detected that the temperature in the local area deviates from the set value, the output compensation of adjacent devices is adjusted in a coordinated manner.

[0086] The intelligent control module establishes a dynamic cooling and heating load adjustment model for the coupling of human flow, space and environment based on the spatiotemporal distribution prediction results of human flow density generated by the data processing module and the real-time collected ambient temperature and humidity data. By analyzing the mapping relationship between human flow density and spatial heat load in different areas, the cooling and heating output requirements of the air-conditioning system are calculated, and the energy supply strategy is dynamically corrected in combination with real-time environmental parameters: when the predicted human flow density is higher than the threshold and the regional temperature deviates from the set range, the variable frequency compressor is preferentially started to increase the cooling / heating power output, and the air supply direction and wind speed are adjusted according to the spatial heterogeneity of the human distribution to achieve precise control of local thermal comfort. Conversely, when the human flow density is monitored to decrease and the environmental parameters are in the comfortable range, the operating frequency of the air-conditioning host is automatically reduced and switched to low-power mode, and the basic temperature control requirements are maintained by gradient adjustment of the air supply volume and return water temperature.

[0087] The core goal of this application is to achieve efficient energy utilization while ensuring environmental comfort. To this end, the following multi-objective optimization objective function is established:

[0088]

[0089]

[0090] Where α is the energy consumption optimization weight, E(t) is the total energy consumption of the system, β is the weight of the crowd flow prediction accuracy, N is the total number of sampling points in the time series, and T pred (t) is the crowd density predicted by the fusion of infrared sensor and pressure sensor, T real (t) is the actual crowd density, γ is the equipment control accuracy weight, M represents the total number of adjustable devices in the system, The actual power output of the jth device, is the theoretical optimal power of the device, ΔH and ΔT are the allowable fluctuation range of ambient temperature and humidity, T(t) is the real-time ambient temperature, H(t) is the real-time ambient humidity, Q flow (t) is the real-time equipment air volume, θ press is the static pressure of the space, T set is the set ambient temperature, H set is the set ambient humidity, Q max is the maximum air volume, ∈ flow is the minimum fresh air volume coefficient, θ low is the minimum static pressure in the space, θ high is the maximum static pressure in the space.

[0091] The core algorithm used to establish the crowd flow model in the data processing module is implemented as follows: Figure 2 :

[0092] Step 1: Construct a personnel information sample data packet matrix

[0093] Assume that there are m personnel samples in the building, and each sample has q indicator explanatory variables, such as location coordinates, movement speed, and residence time. The indicator variable Y = (y1, y2, y3, ... y q ), the characteristic value of information, Y η =(y η1 ,y η2 ,y η3 ,…y ηq ) Establish a personnel information sample data matrix as follows:

[0094]

[0095] In the formula, the sample data matrix Y belongs to the target of text classification, where Y ηi Describes the characteristic value of the ξ-th indicator in the η-th person information. ξ1 is the current position of the person, y ξ2 This is the person's movement speed at this time.

[0096] Step 2: Calculate the Euclidean distance between people, two person samples y g 、y f The similarity is determined by the Euclidean distance as follows:

[0097]

[0098] where e gf Indicates person y g ,y f In the multidimensional feature space, smaller distance values indicate more similar behavior patterns.

[0099] Step 3: According to the number of clusters, set the initial cluster center and cluster the personnel information samples. The clustering standard is: g 、y f Assume d types, as follows

[0100]

[0101] Then the clustering standard matrix has the following properties, as shown in the formula:

[0102] v ηξ ∈{0,1} and

[0103] where v ηξ Indicates whether the sample ξ belongs to the category η

[0104] Step 4: Calculate the class center and the difference between classes

[0105] The number of personnel information samples in the ηth category is m η , then we have the following formula:

[0106]

[0107] The central eigenvalue of the nth type of personnel information sample is Then there is a formula

[0108]

[0109] The inter-class difference of the ηth type of personnel information samples is as follows (measuring the degree of dispersion of samples of the same type).

[0110]

[0111] The overall difference is:

[0112]

[0113] in represents the average behavioral characteristics of the ηth type of personnel (such as average moving speed), Z(v * ) is the advantage indicator of clustering results, and the smaller the value, the more reasonable it is.

[0114] Step 5: Iterative Optimization

[0115] By minimizing the overall difference Z(v * ), solve the optimal clustering result v * :

[0116] Z(v ＊ )=min{Z(v)}

[0117] When the number of iterations reaches the same as the set number of iterations, clustering is stopped and the personnel information classification results are obtained.

[0118] Step 6: Solve the crowd density model

[0119]

[0120] Where, T pred (t) is the predicted value of the output crowd density, Y is the description matrix; c is a vector, is the weight coefficient.

[0121] The intelligent control module uses a built-in priority algorithm to prioritize core areas. Areas with higher scores are given priority for energy supply. The built-in priority algorithm steps are as follows:

[0122] Step 1: Let the priority score P of the i-th region be i for:

[0123]

[0124] in, is the normalized crowd density, ΔT i is the absolute deviation between the current temperature and the set temperature, C i is the unit energy consumption cost, α1, β1, and γ1 are dynamic weight coefficients, reflecting the priority of the control target (e.g., α1 increases during peak hours and γ1 increases during nighttime energy-saving mode).

[0125] Step 2: Data normalization, mapping parameters of different dimensions to the interval 0,1 to ensure comparability:

[0126]

[0127] Among them D max , ΔT max 、C max is the preset maximum reference value.

[0128] Step 3: Adjust the weight coefficient according to time, season, operation strategy, etc.:

[0129] Peak hours (e.g. 10:00-12:00 on weekdays): Prioritize comfort, with the following weights:

[0130] α1=0.4,β1=0.5,γ1=0.1

[0131] Energy-saving period (e.g., 11:00 PM to 6:00 AM): Focus on reducing energy consumption costs. The weights are set as follows:

[0132] α1=0.1,β1=0.3,γ1=0.6

[0133] Step 4: Calculate P for all regions i After the value is calculated, it is sorted in descending order, and energy is allocated first to the area with higher priority:

[0134] Priority queue = Sort(P1,P2,...,P n )

[0135] Step 5: Score P based on priority i , dynamically adjust the device output power:

[0136] Air conditioning power distribution:

[0137] W i AC =W base +k·(P i -P threshold )

[0138] Among them, W i AC represents the electric power of the air conditioner in the i-th area, k is the gain coefficient, P thresholdis the activation threshold; then the actual power output of each air conditioner is:

[0139]

[0140] in is the actual power output of the jth device in area i, and M represents the total number of adjustable air-conditioning devices in the system;

[0141] Exhaust fan speed:

[0142]

[0143] The speed is allocated according to the proportion of pedestrian density to ensure ventilation needs in high-density areas. The real-time exhaust volume of the fan is:

[0144]

[0145] in, is the real-time speed of the exhaust fan in the i-th area, S max is the maximum rated speed of the exhaust fan, n is the total number of zones, k q is the air volume-speed ratio coefficient.

Claims

1. An energy output power regulation method based on artificial intelligence and big data, characterized in that: The following steps are involved: S1, real-time collection of environmental data and personnel flow information within the space; S2, generate a crowd flow model through big data analysis and machine learning algorithms; S3. Automatically adjust the operating status and output power of energy equipment according to the crowd flow model and environmental data to achieve precise environmental control and energy conservation.

2. The energy output power regulation method based on artificial intelligence and big data according to claim 1 is characterized in that: S1 collects data on wireless signal strength, personnel movement trajectories, identity tags, visual images, ambient temperature and humidity, regional pressure changes, and human thermal radiation, and integrates them to build a three-dimensional monitoring network covering the physical environment status and human flow dynamics.

3. The energy output power regulation method based on artificial intelligence and big data according to claim 1 is characterized in that: S2 receives and stores multi-source data from S1, and uses the artificial intelligence platform to dynamically compare and pattern-match real-time data with typical crowd flow models in the big data model group. It then trains prediction algorithms based on historical data to generate accurate prediction results for the spatiotemporal distribution of crowd flow. Convert the analysis conclusions into control parameters to provide a decision-making basis for S3.

4. The energy output power regulation method based on artificial intelligence and big data according to claim 1 is characterized in that: S3 uses the spatiotemporal distribution of crowd density predictions and real-time ambient temperature and humidity data to establish a dynamic cooling and heating load regulation model for the coupled crowd, space, and environment. By analyzing the mapping relationship between crowd density and spatial heat load in different areas, it calculates the cooling and heating output requirements of the air conditioning system and dynamically adjusts the energy supply strategy based on real-time environmental parameters. When the predicted crowd density is higher than the threshold and the regional temperature deviates from the set range, the variable frequency compressor is activated first to increase the cooling / heating power output. At the same time, the air supply direction and wind speed are adjusted according to the spatial heterogeneity of the occupant distribution to achieve precise control of local thermal comfort. On the contrary, when it is detected that the crowd density is decreasing and the environmental parameters are in a comfortable range, the operating frequency of the air-conditioning host is automatically reduced and switched to low-power mode, and the basic temperature control requirements are maintained by gradient adjustment of the supply air volume and return water temperature.

5. The energy output power regulation method based on artificial intelligence and big data according to claim 3 is characterized in that: The establishment of the crowd flow model in S2 includes the following steps: Step 1: Construct a personnel information sample data packet matrix Assume that there are m samples of people in the building, and each sample has q indicator explanatory variables, and the indicator variable Y = (y1, y2, y3, ... y q ), the characteristic value Y of the information η =(y η1 ,y η2 ,y η3 ,…y ηq ), establish the personnel information sample data matrix: In the formula, the sample data matrix Y belongs to the target of text classification, where Y ηi Describe the characteristic value of the ξ-th indicator in the η-th person information; Step 2: Calculate the Euclidean distance between people, two person samples y g 、y f The similarity is determined by the Euclidean distance as follows: Among them, e gf Indicates person y g ,y f In the multidimensional feature space, smaller values of distance indicate more similar behavior patterns; Step 3: According to the number of clusters, set the initial cluster center and cluster the personnel information samples. The clustering standard is: cluster m personnel information samples y g 、y f Assume d types, as follows The clustering standard matrix has the following properties, as shown in the formula: v ηξ ∈{0,1} and where v ηξ Indicates whether the sample ξ belongs to the category η; Step 4: Calculate the class center and inter-class difference The number of personnel information samples in the ηth category is m η : The central eigenvalue of the nth type of personnel information sample is The inter-class difference of the nth type of personnel information samples: The overall difference is: in represents the average behavioral characteristics of the ηth type of personnel, Z(v * ) is the advantage indicator of the clustering result, the smaller the value, the more reasonable it is; Step 5: Iterative Optimization By minimizing the overall difference Z(v * ), solve the optimal clustering result v * : With(in ★ )=min{Z(v)} When the number of iterations reaches the same as the set number of iterations, clustering is stopped and the personnel information classification results are obtained; Step 6: Solve the crowd density model Where, T pred (t) is the predicted value of the output crowd density, Y is the description matrix; c is a vector, is the weight coefficient.

6. The energy output power regulation method based on artificial intelligence and big data according to claim 5 is characterized in that: The S3 priority algorithm prioritizes core areas and prioritizes energy supply to areas with higher scores. The algorithm includes the following steps: Step 1: Let the priority score P of the i-th region be i for: in, is the normalized crowd density, ΔT i is the absolute deviation between the current temperature and the set temperature, C i is the unit energy consumption cost, α1, β1, and γ1 are dynamic weight coefficients, reflecting the priority of the control target; Step 2: Data normalization, mapping parameters of different dimensions to the interval 0,1 to ensure comparability: Among them D max , ΔT max 、C max is the preset maximum reference value; Step 3: Adjust the weight coefficient according to time, season, and operation strategy: Peak hours: Prioritize comfort, with weights set as: α1=0.4,β1=0.5,γ1=0.1 Energy-saving period: focuses on reducing energy consumption costs, with weights set as: α1=0.1,β1=0.3,γ1=0.6 Step 4: Calculate P for all regions i After the value is entered, it is sorted in descending order, and energy is allocated first to the area with higher priority: Priority queue = Sort(P1,P2,...,P n ) Step 5: Score P based on priority i , dynamically adjust the device output power: Air conditioning power distribution: W i AC =W base +k·(P i -P threshold ) Among them, W i AC represents the electric power of the air conditioner in the i-th area, k is the gain coefficient, P threshold is the activation threshold; then the actual power output of each air conditioner is: in is the actual power output of the jth device in area i, and M represents the total number of adjustable air-conditioning devices in the system; Exhaust fan speed: The speed is allocated according to the proportion of pedestrian density to ensure ventilation needs in high-density areas. The real-time exhaust volume of the fan is: in, is the real-time speed of the exhaust fan in the i-th area, S max is the maximum rated speed of the exhaust fan, n is the total number of zones, k q is the air volume-speed ratio coefficient.

7. The energy output power regulation method based on artificial intelligence and big data according to claim 6 is characterized in that: Establish a multi-objective optimization objective function for efficient energy utilization and perform energy optimization according to target changes: Where α is the energy consumption optimization weight, E(t) is the total energy consumption of the system, β is the weight of the crowd flow prediction accuracy, N is the total number of sampling points in the time series, and T pred (t) is the crowd density predicted by the fusion of infrared sensor and pressure sensor, T real (t) is the actual crowd density, γ is the equipment control accuracy weight, M represents the total number of adjustable air-conditioning equipment in the system, is the actual power output of the jth device in region i, is the theoretical optimal power of the device, ΔH and ΔT are the allowable fluctuation range of ambient temperature and humidity, T(t) is the real-time ambient temperature, H(t) is the real-time ambient humidity, Q flow (t) is the real-time equipment air volume, θ press is the static pressure of the space, T set is the set ambient temperature, H set is the set ambient humidity, Q max is the maximum air volume, ∈ flow is the minimum fresh air volume coefficient, θ low is the minimum static pressure in the space, θ high is the maximum static pressure in the space.

8. An energy output power regulation system based on artificial intelligence and big data, characterized in that: include: Data acquisition module, data processing module and intelligent control module; The data acquisition module is used to collect environmental data and personnel flow information in real time; The data processing module includes a data storage unit and a data analysis and processing unit, which generates a crowd flow model through big data analysis and machine learning algorithms; The intelligent control module includes an environmental control system and an air conditioning energy supply system, which is used to automatically adjust the operating status and output power of energy equipment according to the crowd flow model and environmental data to achieve precise environmental control and energy conservation.

9. The energy output power regulation system based on artificial intelligence and big data according to claim 8 is characterized in that: The data acquisition module includes: temperature and humidity sensors, pressure sensors, video cameras, infrared sensors, Wi-Fi probes, mobile phone signaling and radio frequency identification equipment.

Images