A smart power consumption behavior analysis and management method

Abstract

The application discloses a kind of wisdom power consumption behavior analysis and management method, it is related to energy management technical field, it aims at through load correlation analysis technology, improve the energy consumption synergistic effect identification and management between devices in smart home.Method through the real-time acquisition of multiple electric appliance equipment in family of internet of things equipment or sensor power consumption data;Based on the load fluctuation between devices, a load correlation graph is constructed, and the load fluctuation strength relationship between devices is calculated through cross correlation analysis, mutual information analysis and other methods, so as to identify the synergistic effect and potential failure between devices.Through graph theory algorithm, the load correlation graph is optimized, the device priority is sorted, and the energy scheduling and management strategy is optimized.The application can monitor the load fluctuation of the device in real time, accurately identify abnormal power consumption behavior, provide intelligent load scheduling and energy efficiency management scheme, effectively improve the safety, energy efficiency and intelligent level of home intelligent power management.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, and in particular to a method for intelligent electricity consumption behavior analysis and management. Background Technology

[0002] With the increasing number and variety of electrical appliances, the electrical load in homes and commercial buildings is becoming increasingly complex. Multiple appliances, such as air conditioners, refrigerators, washing machines, and water heaters, often operate simultaneously, and their energy consumption is affected not only by their individual operating conditions but also by synergistic effects. Especially when high-energy-consuming appliances, such as air conditioners and refrigerators, are operating together, it can lead to drastic fluctuations in electrical load, thus placing enormous pressure on energy management in homes or buildings.

[0003] Most existing methods for analyzing electricity consumption behavior focus on monitoring and analyzing the energy consumption of individual devices. By collecting and processing the energy consumption data of each device, it is possible to predict the operating mode of the device and manage it. However, these traditional methods often overlook the synergistic effects between devices. That is, when multiple devices are running together, their energy consumption fluctuations are interconnected, and this mutual influence cannot be accurately grasped by simply analyzing the load characteristics of a single device. For example, in the summer when air conditioners and refrigerators are used together, the increased cooling load of the air conditioner may overlap with the frequent start-up of the refrigerator, resulting in drastic changes in load fluctuations. Traditional methods only analyze the behavior of individual devices and cannot identify the potential problems caused by such synergistic fluctuations.

[0004] Furthermore, existing energy consumption monitoring typically employs static models based on single devices or simple statistics to predict energy consumption fluctuations and anomalies. This presents significant limitations when dealing with complex load fluctuations caused by multiple devices operating simultaneously. Traditional monitoring methods fail to effectively capture the temporal and functional correlations between devices, as well as the characteristics of load fluctuations. Especially when devices start simultaneously and frequently switch operating modes, traditional methods often fail to detect abnormal load fluctuations in a timely manner, leading to energy waste, insufficient early warning of equipment failures, and even impacting the electrical safety of the entire building or household.

[0005] Therefore, a smart electricity consumption behavior analysis and management method is proposed. Summary of the Invention

[0006] The purpose of this invention is to address the problem that when multiple electrical devices operate together, their energy consumption fluctuations are interconnected, and this mutual influence cannot be accurately grasped simply by analyzing the load characteristics of a single device. Therefore, this invention proposes a smart electricity behavior analysis and management method.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A smart electricity consumption behavior analysis and management method includes the following steps:

[0009] Collect device information: Collect electricity consumption data of multiple electrical appliances in the home in real time through IoT devices or sensors. The electricity consumption data includes the power, running time, voltage, current and switch status of the device, which serves as the basic data for each device.

[0010] Device load correlation modeling: Each device is treated as a node, and the load correlation between devices is treated as an edge to construct a device load correlation network. The load fluctuation of the devices is reflected by time series data, which includes changes in device power, start-up and stop times between devices, etc.

[0011] Calculate the strength of load fluctuations: Based on the power consumption data of the devices, calculate the strength relationship of load fluctuations between the devices. The calculation includes:

[0012] Cross-correlation analysis is used to determine whether there is a temporal correlation between the power consumption fluctuations of two devices; the formula for calculating the cross-correlation function is:

[0013]

[0014] in, and These represent the time intervals of device A and device B, respectively. Power data, and These are their means, It's a time delay;

[0015] The mutual information content or correlation coefficient is used to determine whether the load fluctuations between two devices have a non-linear correlation; the formula for calculating mutual information content is:

[0016]

[0017] in, It is the joint probability distribution of power data for device A and device B. and It is the marginal probability distribution of equipment A and equipment B;

[0018] Based on the correlation of load fluctuations, determine the load correlation between equipment A and equipment B, and assess whether there is a synergistic effect between them; the comprehensive correlation score formula is as follows:

[0019]

[0020] in, and These are weighting coefficients used to balance the contributions of cross-correlation and mutual information analysis. It is cross-correlation. It is mutual information content;

[0021] Generate a load correlation graph: Based on the strength of load fluctuations between devices, generate a load correlation graph between devices. In the graph, each device is a node, and the load correlation between each pair of devices is an edge. The weight of the edge is assigned according to the intensity of load fluctuations.

[0022] Anomaly detection based on load correlation: By monitoring changes in the load correlation graph, abnormal power consumption behavior can be identified. The abnormal behavior includes abnormal correlation of load fluctuations between devices, thereby detecting equipment failure, energy waste or abnormal power consumption.

[0023] Preferably, the inter-device load correlation modeling includes:

[0024] Each electrical device is considered as a node in the graph, and the device's power consumption data is used as the node's attribute information.

[0025] The time-series data of the power consumption of the equipment includes changes in equipment power, the on and off status of the equipment, the operating time of the equipment, the frequency, and the temporal relationships between them;

[0026] By calculating the power fluctuation data between devices and performing cross-correlation analysis based on the device time-series data, the load correlation between devices is constructed, and the correlation is represented as edges in the graph.

[0027] The edge weights between each pair of devices are assigned values ​​based on the strength of load fluctuations. The edge weight values ​​reflect the correlation of load fluctuations between devices, further demonstrating the mutual influence and correlation strength of load fluctuations between devices.

[0028] Preferably, the inter-device load correlation modeling further includes:

[0029] Based on the power data between devices, a spatiotemporal model of device load correlation is constructed, further considering the geographical distribution and time period characteristics of the devices to capture the interaction between devices;

[0030] By using time window sliding technology to segment data, we can more efficiently analyze the load change trends of equipment in different time periods and improve the accuracy of load correlation modeling.

[0031] Preferably, the calculation of load fluctuation strength includes: cross-correlation analysis: by calculating the time-series correlation of power fluctuations between devices, the time dependency of load fluctuations between device A and device B is analyzed. The specific steps are as follows:

[0032] Collect power data from devices A and B. The power data is recorded in time series form, reflecting the power consumption of the devices within a specific time period.

[0033] The power time series of device A and device B are calculated using the cross-correlation function to measure the similarity between the two time series under different time delays.

[0034] If the correlation value is significant at a certain time delay, it indicates that there is a significant temporal correlation between load fluctuations between devices, and further analysis is needed to determine whether there is a synergistic effect.

[0035] Preferably, the strength of the computational load fluctuation also includes:

[0036] Mutual information analysis: By calculating mutual information, the nonlinear relationship of load fluctuations between devices is measured, revealing potential synergistic effects between devices. The steps are as follows:

[0037] Acquire the power time series data of device A and device B, and ensure that there is sufficient overlap between the time series;

[0038] By discretizing the power data and calculating the mutual information through the joint probability distribution and marginal probability distribution, the nonlinear dependence between device A and device B is revealed.

[0039] By identifying the synergistic effect between devices through mutual information values, mutual information can capture the nonlinear effects between devices.

[0040] Preferably, the strength of the computational load fluctuation also includes:

[0041] Correlation determination: Combining the results of cross-correlation analysis and mutual information analysis, the strength of load fluctuations between devices is comprehensively evaluated. The specific steps are as follows:

[0042] The results of cross-correlation analysis and mutual information analysis are combined, and a weighted calculation is used to form a comprehensive load correlation score.

[0043] Based on the overall score, the strength of the load fluctuation relationship between devices is determined, with high scores indicating strong correlation and low scores indicating weak correlation.

[0044] Based on the comprehensive score and a predetermined threshold, it is determined whether there is a significant correlation in load fluctuations between devices, thereby identifying potential synergistic load effects.

[0045] Preferably, the generation of the load correlation graph includes:

[0046] Based on the relative strength of load fluctuations among devices, a weighted graph algorithm is used to generate a device load correlation graph.

[0047] In the diagram, each device is a node, and the load association between each pair of devices is an edge in the diagram. The weight of the edge is assigned according to the intensity of load fluctuation.

[0048] In the load relationship diagram, the dynamic adjustment of edge weights reflects the changes in load fluctuations between devices in real time, ensuring that the load relationship of devices in the diagram can adapt to the dynamically changing power environment.

[0049] Preferably, the generation of the load correlation graph further includes:

[0050] Based on the initial load correlation graph, graph theory algorithms are applied to further optimize the graph structure in order to identify key correlation nodes between devices and prioritize the device group.

[0051] By analyzing the strength of load fluctuations among devices, the importance weights of devices are determined. Based on these weights, devices are grouped to optimize energy dispatching and equipment management strategies, thereby improving overall power efficiency.

[0052] In the load correlation diagram, graph analysis tools are used to analyze the changing trends of load correlation between devices, and the scheduling and management strategies of devices are adjusted according to the load fluctuation patterns of devices, thereby improving the efficiency of energy management and optimizing fault detection and early warning capabilities.

[0053] Preferably, the anomaly detection based on load correlation includes:

[0054] Monitor changes in the load correlation diagram to determine abnormal changes in load fluctuations between devices and identify abnormal power consumption behavior;

[0055] The abnormal behavior includes an abnormal correlation of load fluctuations between device A and device B, indicating device failure, device overload, energy waste, and abnormal power consumption.

[0056] Preferably, the anomaly detection further includes:

[0057] Based on historical data from the load correlation diagram, the threshold for load fluctuations between devices is dynamically adjusted. By comparing the deviation between the current load fluctuation and the historical fluctuation, it is determined whether the fluctuation is abnormal.

[0058] By training a deep neural network model and combining it with the characteristics of load fluctuations, abnormal patterns of load fluctuations between devices can be automatically identified, thereby improving the accuracy and robustness of anomaly detection.

[0059] The present invention has the following beneficial effects:

[0060] 1. This invention significantly improves the energy efficiency synergy among devices in smart power management by introducing load correlation modeling and its calculation methods. By combining cross-correlation analysis and mutual information analysis, the temporal correlation between devices can be captured, and the potential synergistic effects of nonlinear load fluctuations can be revealed. By constructing a load correlation graph between devices and optimizing the load correlation using graph theory algorithms, dynamic analysis and optimization of multi-device synergistic effects are achieved. By real-time monitoring of load fluctuations and their correlations among devices, synergistic load fluctuations among devices can be accurately captured. By introducing a graph convolutional network (GCN) graph theory optimization algorithm, not only can key correlation nodes between devices be identified, but the scheduling priority of devices can also be dynamically adjusted according to the intensity and frequency of device load fluctuations. Furthermore, it promptly identifies abnormal synergistic load fluctuations.

[0061] 2. In this invention, by collecting real-time electricity consumption data from household devices and combining it with load correlation modeling and anomaly detection between devices, the load fluctuation of devices can be comprehensively and in real-time reflected. The impact of load fluctuations when multiple devices operate collaboratively is considered, and optimized scheduling is performed based on this correlation. It can monitor load fluctuations between devices in real time and dynamically adjust energy management strategies according to the changing trends of load correlations. Furthermore, the anomaly detection mechanism, by introducing historical data to dynamically adjust the detection threshold and combining it with a deep neural network (DNN) model, effectively improves the accuracy and robustness of anomaly fluctuation detection. Attached Figure Description

[0062] Figure 1 This is a flowchart of a smart electricity consumption behavior analysis and management method proposed in this invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] This implementation plan proposes a smart electricity consumption behavior analysis and management method. Through data acquisition, modeling, analysis, anomaly detection, and intelligent scheduling, it achieves load fluctuation monitoring, fault diagnosis, and energy optimization for electrical equipment. The modeling and analysis process of this method is described in detail below.

[0065] like Figure 1 As shown, the present invention proposes a smart electricity consumption behavior analysis and management method, which includes the following steps:

[0066] Collect device information: Collect real-time power consumption data of multiple electrical appliances in the home through IoT devices or sensors. The power consumption data includes the power, running time, voltage, current and switch status of the device, which serves as the basic data for each device.

[0067] Device load correlation modeling: Each device is treated as a node, and the load correlation between devices is treated as an edge to construct a device load correlation network. The load fluctuation of the devices is reflected by time series data, which includes changes in device power, start-up and stop times between devices, etc.

[0068] Calculate the strength of load fluctuations: Based on the power consumption data of the equipment, calculate the relationship between the strength of load fluctuations between the equipment. The calculation includes:

[0069] Cross-correlation analysis is used to determine whether there is a temporal correlation between the power consumption fluctuations of two devices; the formula for calculating the cross-correlation function is:

[0070]

[0071] in, and These represent the time intervals of device A and device B, respectively. Power data, and These are their means, It's a time delay;

[0072] The mutual information content or correlation coefficient is used to determine whether the load fluctuations between two devices have a non-linear correlation; the formula for calculating mutual information content is:

[0073]

[0074] in, It is the joint probability distribution of power data for device A and device B. and It is the marginal probability distribution of equipment A and equipment B;

[0075] Based on the correlation of load fluctuations, determine the load correlation between equipment A and equipment B, and assess whether there is a synergistic effect between them; the comprehensive correlation score formula is as follows:

[0076]

[0077] in, and These are weighting coefficients used to balance the contributions of cross-correlation and mutual information analysis. It is cross-correlation. It is mutual information content;

[0078] Generate a load correlation graph: Based on the strength of load fluctuations between devices, generate a load correlation graph between devices. In the graph, each device is a node, and the load correlation between each pair of devices is an edge. The weight of the edge is assigned according to the strength of the load fluctuation.

[0079] Anomaly detection based on load correlation: By monitoring changes in the load correlation graph, abnormal power consumption behavior can be identified. Abnormal behavior includes abnormal correlation of load fluctuations between devices, thereby detecting equipment failure, energy waste or abnormal power consumption.

[0080] 1. Data Acquisition and Preprocessing:

[0081] In this method, the data input is electricity consumption data collected in real time from multiple home appliances via IoT devices and sensors. The specific steps are as follows:

[0082] Data acquisition: Using sensor devices installed in the home (such as smart sockets, power meters, smart meters, etc.) to periodically or in real time monitor basic information such as power, current, voltage, and switch status of multiple electrical appliances.

[0083] Power (W): Reflects the operating intensity of the equipment; Current (A): Reflects the power consumption of the equipment; Voltage (V): The voltage requirement of the equipment; Switch status: Reflects whether the equipment is in the on or off state.

[0084] Data cleaning and preprocessing: Data denoising and supplementation: Using data cleaning algorithms to remove noise that occurs during the data collection process, fill in missing values, and ensure data integrity.

[0085] Standardization and normalization: In order to ensure that the data of different electrical devices are in the same dimension, the collected power, current, voltage and other values ​​need to be standardized.

[0086] Time alignment: Since the data collection time points of different devices may differ, it is necessary to perform time alignment processing on the data to ensure that data within the same time period can be effectively compared and analyzed.

[0087] 2. Modeling the load correlation between devices: Analyzing the load fluctuation relationship between devices. Specific steps are as follows:

[0088] Construction of the load correlation network between devices:

[0089] Each device is considered a node in the graph, and the device's power consumption data (such as power, current, etc.) is considered an attribute of the node.

[0090] The load fluctuation relationships between devices are represented by edges in the graph. The weight of each edge is determined by the strength of the load fluctuation relationship between each pair of devices.

[0091] The load fluctuations of the equipment are reflected by time series data, which includes changes in the equipment's power, the timing of its on and off cycles, and the frequency of its use.

[0092] Correlation analysis: Through cross-correlation analysis and mutual information analysis, the load fluctuation relationship between devices is revealed.

[0093] The spatiotemporal model treats each device as a node, and the load correlation between devices is dynamically weighted based on the following factors:

[0094] Geographical distance: The spatial distance between devices affects the degree of mutual influence of their load fluctuations. Devices that are closer together may have a stronger load correlation.

[0095] Time period relationship: The load fluctuation of equipment within the same time period has time period characteristics. By modeling the time period characteristics, the model can capture the periodic pattern of equipment load fluctuation.

[0096] Spatiotemporal modeling can more accurately describe the spatiotemporal relationships between device loads and identify complex interaction patterns across time and geographical regions.

[0097] Cross-correlation analysis:

[0098] The cross-correlation function is used to calculate the temporal correlation between power data sequences of two devices. Cross-correlation reflects the synchronicity and lag of power fluctuations between devices over time.

[0099] For example, suppose the power data of device A is The power data for device B is Cross-correlation can be calculated using the following formula: The formula for calculating the cross-correlation function is:

[0100]

[0101] in, and These represent the time intervals of device A and device B, respectively. Power data, and These are their means, It's a time delay;

[0102] Mutual information analysis measures the non-linear dependency between device A and device B. It aims to uncover potential associations that cannot be revealed by cross-correlation.

[0103] Assume the power data of device A and device B are respectively and Then mutual information It can be calculated using the joint probability distribution and the marginal probability distribution:

[0104]

[0105] in, It is the joint probability distribution of power data for device A and device B. and It is the marginal probability distribution of equipment A and equipment B;

[0106] Determining the strength of load correlation:

[0107] Based on the results of cross-correlation analysis and mutual information calculation, the strength of load fluctuations between devices is comprehensively evaluated.

[0108] If the cross-correlation or mutual information value is high, it indicates that there is a strong correlation between the load fluctuations of the two devices, and there may be a synergistic effect.

[0109] Application of sliding time window technology: To more efficiently analyze the load change trends of equipment over different time periods, a sliding time window technique is used to segment the data. This technique dynamically segments the data and progressively updates the load change trends within each time window, thereby improving the accuracy of load correlation modeling. Data within each time window is analyzed according to specific time intervals, thus revealing the load change patterns of each device and their mutual influences.

[0110] 3. Generation of load correlation diagram:

[0111] Based on the results of the load correlation analysis, a load correlation diagram between devices is generated. The specific steps are as follows:

[0112] Graph Construction: Load fluctuation relationships between devices are represented by edges in the graph, with each device acting as a node, and the edge weight representing the strength of the load fluctuation. The edge weight can be calculated using the following formula:

[0113]

[0114] in, It is equipment With equipment The load correlation strength between them and These are weighting coefficients used to balance the contributions of cross-correlation and mutual information analysis. It is cross-correlation. It is mutual information content;

[0115] Dynamic updates of the graph: The load correlation graph is a dynamic graph. As the power consumption data of the devices is updated in real time, the structure and edge weights of the graph will change. In order to monitor the status of electrical equipment in real time, the edge weights in the graph need to be dynamically adjusted to reflect changes in load fluctuations.

[0116] Graph structure optimization: Optimization is performed using load relationship graphs constructed using graph theory algorithms. Specific optimization methods include:

[0117] Minimum Spanning Tree (MST) algorithm: used to select key nodes from all device nodes, remove redundant nodes and edges, and ensure that the connection between devices with strong load fluctuations is the shortest and most direct.

[0118] Maximum flow algorithm: Optimizes the power flow path between devices, ensuring efficient energy transmission and avoiding unnecessary power loss.

[0119] Shortest path algorithm: Determine the shortest path for load fluctuations between devices and optimize load scheduling strategies.

[0120] Optimization process of Graph Convolutional Networks (GCN):

[0121] In the graph optimization section, the load correlation graph, after incorporating the spatiotemporal model, will be optimized using a Graph Convolutional Network (GCN). GCN learns the neighborhood information of device nodes and optimizes the features of each node, thereby improving the accuracy of the load correlation model. The specific optimization process is as follows:

[0122]

[0123] in, Indicates the first Layer device node characteristics, For nodes Neighbors For nodes The degree, For the first The weight matrix of the layer, It is the activation function. Through optimization, the representational power of the device load correlation graph is improved, thereby more accurately reflecting the load fluctuation relationship between devices and enhancing the predictive power of the model.

[0124] Critical Node Identification: By analyzing the graph structure, critical nodes can be identified, which are devices with high load fluctuation intensity. Critical nodes are usually the "central" devices that affect the load of other devices.

[0125] 4. Anomaly detection based on load correlation graph:

[0126] By monitoring the load correlation diagram, abnormal power consumption behavior can be identified in a timely manner. The anomaly detection process includes the following steps:

[0127] Monitor load fluctuations: By tracking changes in edge weights in the load correlation graph in real time, abnormal changes in load fluctuations between devices can be detected. For example, if a device suddenly changes its load fluctuation pattern, it may indicate that the device has malfunctioned.

[0128] Identification of Abnormal Behavior:

[0129] When the load fluctuation between two devices no longer conforms to the previous correlation pattern, a threshold is set to determine whether the change is abnormal.

[0130] Abnormal behavior may include equipment malfunction (such as large power fluctuations), equipment overload (such as power exceeding the maximum load) or abnormal power consumption (such as current flowing through the equipment when it is turned off).

[0131] Output anomaly alarm: When an anomaly is detected, an alarm can be output to alert the user that the equipment is malfunctioning, there is abnormal power consumption, or there is potential energy waste.

[0132] Historical data is used to dynamically adjust the threshold for load fluctuations between devices. Based on the statistical characteristics of historical data (such as average value and standard deviation), a dynamic threshold range is set, and the deviation between current load fluctuation data and historical fluctuation data is compared in real time to determine whether it is an abnormal fluctuation. Dynamic threshold adjustment improves the accuracy of anomaly detection and avoids false alarms and missed alarms.

[0133] A deep neural network (DNN) model is introduced. This model is trained on historical load fluctuation data and combined with the load fluctuation characteristics between devices to automatically identify abnormal patterns of load fluctuations between devices.

[0134] 5. Energy Management and Optimization:

[0135] Based on the load correlation diagram, further optimizations can be made to energy scheduling and equipment management:

[0136] Equipment scheduling: By analyzing the load correlation between equipment, priority is given to scheduling equipment with smaller load fluctuations to avoid energy waste caused by high-load equipment working at the same time.

[0137] Optimize device group management: Based on the load correlation strength, devices can be divided into different priority groups, and those devices with strong load correlation can be scheduled first to achieve energy saving and load balancing.

[0138] Intelligent control: By combining the prediction of load fluctuations and the results of anomaly detection, the operation of the equipment is intelligently controlled to optimize energy consumption and extend the service life of the equipment.

[0139] In one specific embodiment: A user has installed various smart home devices in their home, such as air conditioners, refrigerators, washing machines, televisions, and water heaters. All devices' power consumption is monitored in real time via smart sockets and electricity meters. The user's total power capacity is 10kW, and peak daily electricity consumption is expected to occur in the summer afternoons (12:00-14:00) and winter evenings (18:00-20:00). During these peak periods, air conditioners, water heaters, and electric heating appliances operate simultaneously, often leading to power shortages and even equipment overload and malfunctions.

[0140] 1. Data acquisition and real-time monitoring:

[0141] Real-time data collection is achieved through smart sockets and electricity meters. Each device uploads data such as power, current, and voltage in real time via sensors. The following is an example of electricity consumption data collection for a user's household appliances:

[0142] Air conditioner: power consumption is 1200W, current is 5.5A, voltage is 220V, and sampling frequency is once per second.

[0143] Refrigerator: power consumption is 200W, current is 0.9A, voltage is 220V, and sampling frequency is once every 10 seconds.

[0144] Television: power consumption is 150W, current is 0.7A, voltage is 220V, and sampling frequency is once per second.

[0145] Washing machine: power consumption is 1500W, current is 6.8A, voltage is 220V, and sampling frequency is once every 5 seconds.

[0146] Water heater: power consumption is 3000W, current is 13.6A, voltage is 220V, and sampling frequency is once every 5 seconds.

[0147] By collecting data from these devices, the energy consumption of each appliance can be tracked in real time, ensuring that electricity is allocated reasonably during peak power periods.

[0148] 2. Load correlation analysis between devices:

[0149] Based on the power consumption data between devices, analyze the load fluctuation relationship of each device. Assume the following are the results of the data analysis:

[0150] Air conditioners and water heaters: During peak summer hours (12:00-14:00), the load fluctuations of air conditioners and water heaters showed a significant correlation. The peak power of the air conditioner was 1200W, while the peak power of the water heater during the same period was 3000W. A strong cross-correlation was found between their load fluctuations, with a cross-correlation value of 0.85, indicating a strong association.

[0151] Refrigerator and television: The load fluctuations of these two devices have no obvious correlation, and the cross-correlation is 0.2, so the load relationship between them is considered to be weak.

[0152] Example of equipment load fluctuation relationship:

[0153] Air conditioners and water heaters:

[0154] Cross-correlation: 0.85, indicating a very strong correlation between the load fluctuations of air conditioners and water heaters. This means that during peak hours, the power consumption of air conditioners and water heaters usually changes synchronously, therefore their load management requires special attention.

[0155] Mutual information: 0.45, indicating a high degree of information sharing between the two devices. When the load of one device fluctuates, the power consumption of the other device may change significantly.

[0156] Air conditioners and refrigerators:

[0157] Cross-correlation: 0.2, indicating a weak correlation between the load fluctuations of air conditioners and refrigerators. Although they are both common household appliances, their load changes are not always synchronized, thus allowing for some flexibility in scheduling.

[0158] The mutual information level is 0.1, indicating that their electricity consumption behaviors do not have a strong mutual influence.

[0159] Washing machines and televisions:

[0160] Cross-correlation: 0.1, indicating that the load fluctuations of the washing machine and the television are almost unrelated. Their operating modes are unlikely to cause mutual interference, so their scheduling can be independent.

[0161] Mutual information: 0.05, indicating that the information correlation between them is very weak and can be almost ignored.

[0162] Based on the above correlation analysis, it is possible to intelligently determine which devices are high-load devices and which devices need to be operated in staggered shifts to avoid overload.

[0163] 3. Construction and dynamic updating of the load correlation graph:

[0164] A load correlation graph is constructed using a graph algorithm, and the load relationships between devices are updated in real time. The weight of the edge connecting each device node represents the intensity of load fluctuations, and the graph is updated every 5 seconds. An example is shown below:

[0165] The edge weights for air conditioners and water heaters are 0.85, indicating a high correlation between their load fluctuations.

[0166] The edge weights for air conditioners and refrigerators are 0.2, indicating that their load fluctuations are almost unrelated.

[0167] Each time new load data is collected, the graph structure is updated, and the edge weights between devices are adjusted. For example, when the air conditioner is running at increased speed, power dispatch is optimized by adjusting the edge weights in the load relationship graph.

[0168] 4. Intelligent scheduling and load optimization:

[0169] Based on the load correlation diagram and device power requirements, load optimization scheduling is performed. The following is an example of a scheduling strategy:

[0170] Priority scheduling: During the summer peak hours of 12:00-14:00, the load of air conditioners and water heaters fluctuates greatly. Air conditioners are scheduled to run from 9:00-11:00 in the morning, while water heaters are scheduled to run from 3:00-5:00 in the afternoon to avoid overlap between the two.

[0171] Load balancing: When household appliances are under high load (such as between 6:00 PM and 8:00 PM), the order in which devices operate is adjusted based on priority and the intensity of load fluctuations. For example, when a user turns on the TV, the washing machine, which has a higher power consumption, will be turned off first to avoid overloading the power supply by running simultaneously.

[0172] Time slot optimization: Based on historical data analysis, it is known that users usually use the washing machine between 9:00 pm and 10:00 pm, so the washing machine will be automatically scheduled to run during this time slot.

[0173] 5. Anomaly detection and fault early warning:

[0174] Real-time monitoring of equipment load fluctuations and setting thresholds based on historical data are used to detect abnormal power consumption behavior. The specific operations for data monitoring and anomaly detection are as follows:

[0175] Air Conditioner Overload Detection: The system monitors a sudden increase in the air conditioner's power to 2000W (exceeding the normal 1200W) and maintains high power operation for 5 minutes. At this point, it determines that the air conditioner may be malfunctioning (e.g., compressor overload) and sends an alarm notification to the user.

[0176] Abnormal water heater start-up: It was detected that the water heater was still drawing current (0.5A) even when the user did not turn it on, and this continued for more than 1 hour. This was detected as abnormal power consumption, and a warning was sent to the user, suggesting that they check the status of the water heater.

[0177] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart power utilization behavior analysis and management method, characterized in that, The method comprises the following steps: Collecting device information: collecting power consumption data of multiple electrical devices in the home in real time through Internet of Things devices or sensors, including device power, runtime, voltage, current, and switch state as basic data for each device; Modeling load correlation between devices: each device is taken as a node, and the load correlation between devices is taken as an edge to construct a device load correlation network, wherein the load fluctuation of the device is reflected through time series data, including the change of device power and the start and stop time between devices; Calculating load fluctuation strength: based on the power consumption data of the device, the strength of the load fluctuation between devices is calculated, which includes: Whether the power fluctuation of two devices has a time sequence correlation is determined through cross-correlation analysis; a cross-correlation function calculation formula is: wherein, and denote the power data of device A and device B at time respectively, and are their mean values respectively, is the time delay; Whether the load fluctuation between two devices has nonlinear correlation is judged by mutual information or correlation coefficient; the mutual information calculation formula is: wherein, is the joint probability distribution of the power data of device A and device B, and is the marginal probability distribution of device A and device B; According to the correlation degree of the load fluctuation, the load correlation relationship between the device A and the device B is determined, and whether the device A and the device B exist the synergistic effect is determined; the comprehensive correlation relationship score formula is: wherein, and are weight coefficients balancing the contribution of cross-correlation and mutual information analysis, is the cross-correlation, is the mutual information; Generating a load correlation graph: based on the strength of the load fluctuation between devices, a load correlation graph between devices is generated, in which each device is a node and the load correlation between each pair of devices is an edge, and the weight of the edge is assigned according to the load fluctuation intensity; Abnormality detection based on load correlation: abnormal power consumption behavior is identified by monitoring changes in the load correlation graph, which includes abnormal correlation of load fluctuation between devices, and further detects device failure, energy waste or abnormal power consumption. 2.The method of claim 1, wherein, The device load correlation modeling includes: Each electrical device is regarded as a node in the graph, and the power consumption data of the device is taken as the attribute information of the node; The time series data of the power consumption data of the device includes the change of device power, the start and stop state of the device, the runtime of the device, the frequency and the time sequence relationship between them; By calculating the power fluctuation data between devices, the cross-correlation analysis based on device time series data is used to construct the load correlation between devices, which is represented as an edge in the graph; The weight of the edge between each pair of devices is assigned according to the strength of the load fluctuation, and the weight value of the edge reflects the correlation of the load fluctuation between devices, further embodying the mutual influence and correlation strength of the load fluctuation between devices. 3.The method of claim 1, wherein, The device load correlation modeling further includes: Based on the power data between devices, a space-time model of device load correlation is constructed, further considering the geographical distribution and time period characteristics of the device to capture the interaction between devices; Use time window sliding technology to segment data to more efficiently analyze the load change trend of devices in different time periods and improve the accuracy of load correlation modeling.

4. The method of claim 1, wherein the method further comprises: The calculation of load fluctuation strength includes cross-correlation analysis: by calculating the time sequence correlation of power fluctuation between devices, the time-dependent relationship of load fluctuation between device A and device B is analyzed, and the specific steps are as follows: Collecting power data of device A and device B, which is recorded in time series form to reflect the power consumption of the device in a specific time period; Using cross-correlation function to calculate the power time series of device A and device B to measure the similarity of the two time series at different time delays; If the correlation value at a certain time delay is significant, it indicates that there is a significant time sequence correlation between the load fluctuation of the devices, and further analysis is needed to determine whether there is a synergistic effect.

5. The method of claim 1, wherein the method further comprises: The calculation of load fluctuation strength also includes: Mutual information analysis: By calculating mutual information, the nonlinear relationship between the load fluctuations of devices is measured, and the potential synergistic effects between devices are revealed. The steps are as follows: Obtain the power time series data of device A and device B, and ensure that there is enough overlap between the time series; Discretize the power data, calculate the mutual information through joint probability distribution and marginal probability distribution, and reveal the nonlinear dependence between device A and device B; Identify the synergistic effect between devices through mutual information value. Mutual information can capture the nonlinear influence between devices.

6. The method of claim 1, wherein the method further comprises: The calculation of the strength of load fluctuation also includes: Correlation determination: Combine the results of cross-correlation analysis and mutual information analysis to comprehensively evaluate the strength of the load fluctuation relationship between devices. The specific steps are as follows: Combine the results of cross-correlation analysis and mutual information analysis, and use weighted calculation to form a comprehensive load correlation score; According to the comprehensive score, determine the strength of the load fluctuation relationship between devices. High scores indicate strong correlation, and low scores indicate weak correlation. Based on the comprehensive score and the predetermined threshold, determine whether there is a significant load fluctuation correlation between devices, and further identify potential synergistic load effects.

7. The method of claim 1, wherein the method further comprises: The generation of the load correlation relationship graph includes: Based on the strength of the load fluctuation relationship between devices, use a weighted graph algorithm to generate a device load correlation graph; In the graph, each device is a node, and the load correlation between each pair of devices is an edge in the graph. The weight of the edge is assigned according to the load fluctuation strength; In the load correlation relationship graph, through dynamic adjustment of the edge weight, the change of the load fluctuation between devices is reflected in real time, ensuring that the load correlation relationship of devices in the graph can adapt to the dynamically changing power environment. 8.The method of claim 1, wherein, The generation of the load correlation relationship graph further includes: Based on the constructed preliminary load correlation relationship graph, apply graph theory algorithms to further optimize the graph structure to identify key correlation nodes between devices and prioritize device groups; By analyzing the strength of the load fluctuation relationship between devices, determine the importance weight between devices, group devices according to these weights, and optimize energy scheduling and device management strategies to improve overall power efficiency; In the load correlation relationship graph, use graph analysis tools to analyze the trend of load correlation between devices, and adjust the scheduling and management strategies of devices according to their load fluctuation patterns, thereby improving the efficiency of energy management and optimizing fault detection and early warning capabilities. 9.The method of claim 1, wherein, The abnormality detection based on the load correlation relationship includes: Monitor changes in the load correlation relationship graph to determine abnormal changes in load fluctuations between devices and identify abnormal power usage behaviors; The abnormal power usage behavior includes abnormal correlation of load fluctuations between device A and device B, indicating device failure, device overload, energy waste, and non-normal power usage.

10. The method of claim 1, wherein, The abnormality detection further includes: Based on the historical data of the load correlation relationship graph, dynamically adjust the threshold of the load fluctuation between devices, and compare the deviation of the current load fluctuation from the historical fluctuation to determine whether it is an abnormal fluctuation; Through training of a deep neural network model, combine the characteristics of load fluctuations to automatically identify abnormal patterns of load fluctuations between devices, improving the accuracy and robustness of abnormality detection.

Images