Internet of Things equipment energy consumption optimization method based on artificial intelligence

Abstract

The invention discloses an Internet of Things equipment energy consumption optimization method based on artificial intelligence, and the method comprises the steps: S1, integrating the data of Internet of Things equipment, and forming a multi-dimensional data fusion set through employing a multi-source data fusion technology; s2, dynamically establishing an energy consumption baseline model by using an adaptive regression algorithm; s3, monitoring fluctuation deviation of the energy consumption equipment through a fluctuation identification technology, quickly positioning an energy consumption abnormal position by using an energy consumption detection mechanism, and diagnosing an energy consumption abnormal reason; s4, transmitting the energy consumption data and the user behavior data into an intelligent energy consumption scheduling system to optimize task scheduling of the Internet of Things equipment, and reducing an energy consumption peak value; and S5, constructing an energy consumption peak optimization mechanism, optimizing the weight and the threshold value of the Internet of Things, and realizing continuous energy consumption reduction. Through multi-source data fusion, dynamic energy consumption baseline modeling, energy consumption anomaly detection and diagnosis, intelligent task scheduling and a continuous optimization mechanism, effective management and reduction of energy consumption of the Internet of Things equipment are realized, and the overall energy efficiency of the system and the user experience are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy consumption management of Internet of Things devices, and in particular to an energy consumption optimization method for Internet of Things devices based on artificial intelligence. Background Art

[0002] With the rapid increase in the number of IoT devices and the complex and diverse application scenarios, traditional static or semi-static energy management strategies can no longer meet the needs of efficient and flexible energy management. With the popularization of IoT devices, the collaborative work of massive devices has generated huge data traffic and energy consumption. Especially on resource-constrained edge devices, how to balance performance and energy consumption has become an urgent problem to be solved.

[0003] Traditional energy consumption optimization methods often ignore the dynamic changes in the operating status of equipment and the influence of external environmental factors, resulting in poor energy consumption optimization effects and limited equipment performance and endurance. In addition, IoT devices are often deployed in a wide range of changing environments, from indoors to outdoors, from static to mobile, from low temperature to high temperature. These unpredictable conditions require that the equipment must have intelligent self-adjustment capabilities to adapt to different usage scenarios. Moreover, in traditional technologies for energy consumption management of IoT devices, devices usually adopt preset energy consumption management strategies, fixed operating frequencies, power levels, and sleep cycles. These strategies are often determined during the device design phase and remain unchanged throughout the device life cycle. In addition, devices usually operate independently, lack collaboration and information sharing with other devices, and energy consumption optimization is limited to the level of a single device. It fails to fully utilize environmental data and status information of other devices for optimization and energy reduction.

[0004] Chinese patent CN118210240A (publication date: June 18, 2024) discloses a multi-device energy consumption monitoring and management method based on the Internet of Things, which uses Internet of Things sensors to collect energy consumption data, performs preliminary analysis on the collected data through edge computing devices, transmits the processed data to the central processing server, and performs in-depth analysis of the data through algorithms. Based on the analysis results of the algorithms, the operating parameters are automatically adjusted, network redundancy technology is used to ensure the stability of data transmission, and blockchain technology is applied to record and verify all data transactions, including verifying the integrity, transparency and immutability of the data. The settings are automatically adjusted through self-learning and adaptive algorithms, and renewable energy and smart battery systems are integrated. The above invention uses edge computing for data processing, reduces dependence on central servers, and improves the real-time and efficiency of data processing. At the same time, the integrated AI algorithm can achieve more accurate energy consumption prediction and management.

[0005] However, in terms of the dynamics and accuracy of energy consumption optimization, although the above patent uses artificial intelligence algorithms for energy consumption pattern recognition and user behavior analysis, its adjustment of the device's operating frequency, power and sleep cycle may rely on the analysis results of the central processing server, which may lead to limited response speed and adjustment accuracy. At the same time, the rapid information exchange and two-way collaborative work between devices are mainly based on localized data processing and swarm intelligence network optimization mechanisms, which may not fully tap the potential for deep collaboration between devices, especially in the use of swarm intelligence for advanced decision-making. Although self-learning and adaptive algorithms have been introduced, their adaptive capabilities may be mainly focused on the prediction of user behavior and seasonal changes, and their ability to respond immediately to changes in the device's own state and sudden environmental events may be insufficient.

[0006] In order to solve the above problems, the present invention provides an energy consumption optimization method for IoT devices based on artificial intelligence to achieve high dynamics and accuracy in energy consumption optimization. Through the optimization mechanism, the device can make advanced decisions based on new data and environmental changes, optimize energy consumption models and strategies, and combine multi-dimensional energy consumption analysis models to ensure the comprehensiveness and intelligence of energy consumption management. An adaptive energy consumption optimization adjustment mechanism is designed, which enables the system to automatically adjust the optimization strategy based on real-time feedback, realizing truly adaptive energy consumption management. Summary of the invention

[0007] An artificial intelligence-based method for optimizing energy consumption of IoT devices comprises the following steps:

[0008] S1, integrate energy consumption data, user behavior data and environmental interaction data from IoT devices, and use multi-source data fusion technology to form a multi-dimensional data fusion set;

[0009] S2. Dynamically establish an energy consumption baseline model using an adaptive regression algorithm based on the multi-dimensional data fusion set to reflect the normal fluctuation range of equipment energy consumption;

[0010] S3. Monitor the fluctuation deviation of energy-consuming equipment through fluctuation identification technology, use energy consumption detection mechanism to quickly locate abnormal energy consumption, and diagnose the cause of abnormal energy consumption;

[0011] S4, transferring energy consumption data and user behavior data to the intelligent energy consumption scheduling system to optimize the task scheduling of IoT devices, solve the causes of abnormal energy consumption, and reduce energy consumption peaks;

[0012] S5. Build an energy consumption peak optimization mechanism, combine user feedback and system performance data, optimize the weights and thresholds of the Internet of Things, and achieve continuous energy consumption reduction and system performance improvement.

[0013] Preferably, the S1 specifically includes: unifying the energy consumption data E, user behavior data U and environmental interaction data A in the IoT device to the same scale to eliminate the influence of the dimension;

[0014] The energy consumption data E, user behavior data U and environment interaction data A are given an original data point X with feature j j , which is converted to a standard data point X by the formula j ':

[0015]

[0016] Among them, min(X j ) is the minimum value of the jth feature, max(X j ) is the maximum value of the jth feature;

[0017] According to the standard data point X j 'Select feature j in the IoT device and calculate the absolute correlation coefficient |ρ j ∣, according to ∣ρ j ∣>r threshold Determine the correlation between feature j and energy consumption, when |ρ j ∣>r threshold , then feature j is directly related to energy consumption, otherwise, it is not related;

[0018]

[0019] Among them, X j ' .i is the i-th sample of feature j after standardization, Y i is the energy consumption value of the i-th sample, and are the mean of feature j and energy consumption, n is the number of samples, r threshold is the set correlation coefficient threshold;

[0020] The device E 、w U and w A are the weights of energy consumption data, user behavior data, and environmental interaction data, respectively. The multidimensional fusion data set Z is:

[0021] Z=(w E E+w U U+w A A)

[0022] Among them, w E is the weight of energy consumption data, w U is the weight of user behavior data, w A is the weight of the environmental interaction data.

[0023] Preferably, the S2 specifically includes: extracting a feature set F directly related to energy consumption from the multi-dimensional fusion data set Z, and adding a feature j that meets the conditions to the feature set F;

[0024] Assume that the feature set F contains k features and establish an energy consumption baseline model to predict the energy consumption value k i :

[0025]

[0026] Among them, k0 is the intercept term, k j is the weight coefficient of the jth feature, F j is the jth feature in the feature set F, k is the number of features in the feature set F;

[0027] The optimal energy consumption model is constructed by using a set of parameters so that the energy consumption value predicted by the optimal energy consumption model is close to the actually observed energy consumption value:

[0028]

[0029] Set the model's prediction value to k i , calculate the predicted value k i The mean and standard deviation σ p , the normal fluctuation range of energy consumption of IoT devices is Where c is the range constant.

[0030] Preferably, the energy consumption anomaly detection in S3 specifically includes:

[0031] The energy consumption anomaly is detected by the fluctuation deviation D between the energy consumption value and the predicted value;

[0032] The formula of the fluctuation deviation D is:

[0033] D=Y i -k i

[0034] When |D|>δ, an abnormal situation exists, where δ is a set deviation threshold.

[0035] Preferably, the abnormal energy consumption location in S3 specifically includes:

[0036] As mentioned above, the standard deviation σ of the fluctuation deviation D D Confirm the location of abnormal energy consumption;

[0037] The standard deviation σ D for:

[0038]

[0039] Where c is the detection constant, D i is the fluctuation deviation value of the ith one, is the fluctuation deviation D;

[0040] The mean of D is if |D|>c·σ D , then there is an abnormality in the energy consumption data. If |D|<c·σ D , then there are anomalies in user behavior data.

[0041] Preferably, the intelligent energy consumption scheduling system in S4 includes: an optimization module and an energy consumption peak smoothing module;

[0042] The optimization module will optimize the execution plan of IoT device tasks and reduce the peak energy consumption of IoT;

[0043] The energy consumption peak smoothing module dynamically adjusts the task execution time and sequence of the Internet of Things devices to reduce the energy consumption peak of the intelligent energy consumption scheduling system.

[0044] Preferably, the optimization module of the intelligent energy consumption scheduling system in S4 specifically includes:

[0045] The energy consumption data E and user behavior data U are used to optimize the execution time and sequence of device tasks through the optimization module, and the device task list T is scheduled using the optimization algorithm:

[0046] The equipment task list T = {t1, t2, ..., t i};

[0047] Among them, t i represents the i-th device task;

[0048] For each device task t i , using the energy consumption data E to evaluate its expected energy consumption E i :

[0049]

[0050] Among them, w j is the weight at the jth time point, E(t i , j) represents the expected energy consumption of the i-th device task at the j-th time point;

[0051] The optimization algorithm is used to sort and allocate the time of the equipment task list T:

[0052]

[0053] Among them, T' represents all possible task scheduling schemes, T opt is the optimal scheduling solution, E i(T') is the energy consumption of the ith task under schedule T'.

[0054] Preferably, the energy consumption peak smoothing module of the intelligent energy consumption scheduling system in S4 specifically includes:

[0055] The energy consumption peak smoothing module dynamically adjusts the task execution time and sequence of the equipment to reduce the energy consumption peak;

[0056] For each IoT device d, according to the optimized device scheduling scheme T opt , calculate the energy consumption E of IoT devices d ;

[0057]

[0058] Among them, I(t i , t) represents the indicator function, when device d performs task t at time t i Take 1 when it is, otherwise take 0;

[0059] The optimization algorithm is used to adjust the execution time and sequential calculation of the equipment to minimize the peak energy consumption E p ;

[0060]

[0061] The energy consumption peak E is minimized after the optimization. p .

[0062] Preferably, the weight optimization of the energy consumption peak optimization mechanism in S5 specifically includes:

[0063] The weight optimization collects user feedback to evaluate the actual performance of the system and combines it with the system performance indicators. The user feedback is quantitative information about energy consumption perception, equipment availability and satisfaction.

[0064] The setting defines the user feedback factor f u

[0065]

[0066] Among them, s i is the satisfaction score of the i-th user, s avg is the average satisfaction score, N is the number of users who participated in the feedback, and the user feedback factor f is used u Adjust and optimize weight parameters;

[0067] According to the user feedback factor f u Adjust the weights of energy consumption data E, user behavior data U, and environmental interaction data A, and update the optimization parameters;

[0068] wE '=α+f u ·λ1

[0069] w U '=β+f u ·λ2

[0070] w A '=γ+f u ·λ3

[0071] Among them, λ1, λ2, and λ3 are the learning rates used to update the weights of energy consumption data E, user behavior data U, and environmental interaction data A, respectively.

[0072] Preferably, the threshold optimization of the energy consumption peak optimization mechanism in S5 specifically includes:

[0073] The system performance data is used to determine the difference between actual energy consumption and predicted energy consumption, and the optimization parameters are adjusted accordingly.

[0074] Define the performance difference factor f p :

[0075]

[0076] Among them, E pi is the predicted energy consumption at the i-th time point, E ai is the actual energy consumption at the i-th time point;

[0077] The updating formula of the anomaly detection threshold is:

[0078] δ′=δ+f p ·η

[0079] Where η is the learning rate used to update the anomaly detection threshold;

[0080] The updating formula of the fluctuation range constant is:

[0081] c′=c+f p ·ξ

[0082] Where ξ is the learning rate used to update the fluctuation range constant.

[0083] The present invention uses multi-source data fusion technology to unify the different dimensions and scales of data from different sources to the same scale, which can eliminate the dimensional impact and enable different types of energy consumption related data to be effectively integrated and analyzed.

[0084] The present invention utilizes an adaptive regression algorithm to dynamically establish an energy consumption baseline model, which solves the problem that the energy consumption of equipment changes over time and it is difficult to reflect the normal energy consumption fluctuation range. It provides a model that can accurately reflect the energy consumption level of the equipment under normal working conditions and provides a benchmark for subsequent energy consumption anomaly detection.

[0085] The present invention detects and diagnoses energy consumption anomalies through fluctuation recognition technology and fluctuation deviation analysis, and can timely discover energy consumption anomalies and distinguish between energy consumption data itself and anomalies caused by user behavior.

[0086] The present invention dynamically adjusts the task execution time and sequence of the equipment through an optimization algorithm, effectively reduces the energy consumption peak and realizes more efficient energy consumption management.

[0087] The present invention combines user feedback and system performance data to continuously adjust the optimization strategy, which can continuously improve the energy consumption management strategy, improve system performance and enhance user experience.

[0088] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application so that it can be implemented in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following is a detailed description of the preferred embodiments of the present application in conjunction with the accompanying drawings as follows.

[0089] Based on the detailed description of the specific embodiments of the present application in combination with the accompanying drawings below, those skilled in the art will become more aware of the above and other objects, advantages and features of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings without creative work. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0091] Figure 1 A step diagram of an energy consumption management method for an IoT device with dynamic feedback;

[0092] Figure 2 It is a step diagram of multidimensional data fusion set;

[0093] Figure 3 It is a step diagram of the energy consumption baseline model;

[0094] Figure 4 Figure 1 is a diagram of the abnormality detection steps for fluctuation deviation;

[0095] Figure 5 Diagram of steps for energy consumption optimization. Specific embodiments

[0096] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, rather than all of the embodiments. In the following description, specific details such as specific configuration and components are provided only to help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, for clarity and brevity, the description of known functions and structures is omitted in the embodiment.

[0097] It should be understood that the references to "one embodiment" or "this embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present application. Therefore, the references to "one embodiment" or "this embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0098] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not in itself indicate the relationship between the various embodiments and / or settings discussed.

[0099] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" in this article describes another type of association object relationship, indicating that there can be two relationships. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0100] The term "at least one" in this article is merely a description of the association relationship of associated objects, indicating that there may be three relationships. For example, at least one of A and B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0101] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusions.

[0102] Example 1

[0104] This embodiment mainly describes in detail a specific implementation plan of an IoT device energy consumption optimization method based on artificial intelligence;

[0105] like Figure 1 As shown, a method for optimizing energy consumption of IoT devices based on artificial intelligence includes the following steps:

[0106] S1, integrate energy consumption data, user behavior data and environmental interaction data from IoT devices, and use multi-source data fusion technology to form a multi-dimensional data fusion set;

[0107] S2. Dynamically establish an energy consumption baseline model using an adaptive regression algorithm based on the multi-dimensional data fusion set to reflect the normal fluctuation range of equipment energy consumption;

[0108] S3. Monitor the fluctuation deviation of energy-consuming equipment through fluctuation identification technology, use energy consumption detection mechanism to quickly locate abnormal energy consumption, and diagnose the cause of abnormal energy consumption;

[0109] S4, transferring energy consumption data and user behavior data to the intelligent energy consumption scheduling system to optimize the task scheduling of IoT devices, solve the causes of abnormal energy consumption, and reduce energy consumption peaks;

[0110] S5. Build an energy consumption peak optimization mechanism, combine user feedback and system performance data, optimize the weights and thresholds of the Internet of Things, and achieve continuous energy consumption reduction and system performance improvement.

[0111] Further, if Figure 2 As shown in Figure 1, the energy consumption data E, user behavior data U, and environmental interaction data A in the IoT device are unified to the same scale to eliminate the impact of dimension. The energy consumption data E, user behavior data U, and environmental interaction data A are given an original data point X with feature j. j , which is converted to a standard data point X by the formula j ':

[0112]

[0113] Among them, min(X j ) is the minimum value of the jth feature, max(X j ) is the maximum value of the jth feature;

[0114] According to the standard data point X j 'Select feature j in the IoT device and calculate the absolute correlation coefficient |ρ j ∣, according to ∣ρ j ∣>r threshold Determine the correlation between feature j and energy consumption, r threshold The setting of is usually between [-1,1]. A value close to 1 indicates a positive correlation, a value close to -1 indicates a negative correlation, and a value close to 0 indicates almost no linear correlation. threshold Take 0.5-0.7; when ∣ρ j ∣>r threshold , then feature j is directly related to energy consumption, otherwise, it is not related;

[0115]

[0116] Among them, X j ' .i is the i-th sample of feature j after standardization, Y i is the energy consumption value of the i-th sample, and are the mean of feature j and energy consumption, n is the number of samples, r threshold is the set correlation coefficient threshold;

[0117] Assume w E 、w U and w A are the weights of energy consumption data, user behavior data, and environmental interaction data, respectively. The initial setting is w E The weight is greater than,w U and w A The weight of the multidimensional fusion data set Z is:

[0118] Z=(w E E+w U U+w A A)

[0119] Among them, w E is the weight of energy consumption data, w U is the weight of user behavior data, w A is the weight of the environmental interaction data;

[0120] After determining in S1 that the feature is directly related to energy consumption, the weight w of the feature is set E 、w U and w A , ensuring that in the subsequent energy consumption management process, the processing of characteristics is more reasonable and effective, so as to achieve better energy consumption management and optimization effects.

[0121] Further, if Figure 3As shown, S2 specifically includes: extracting a feature set F directly related to energy consumption from the multi-dimensional fusion data set Z, and adding a feature j that meets the conditions to the feature set F;

[0122] Assume that the feature set F contains k features and establish an energy consumption baseline model to predict the energy consumption value k i :

[0123]

[0124] Among them, k0 is the intercept term, k j is the weight coefficient of the jth feature, F j is the jth feature in the feature set F, k is the number of features in the feature set F;

[0125] Construct the optimal energy consumption model through the intercept term k0 and the feature weight coefficient k j Parameters, so that the energy consumption value predicted by the optimal energy consumption model is close to the actually observed energy consumption value:

[0126]

[0127] Set the model's prediction value to k i , calculate the predicted value k i The mean and standard deviation σ p , the normal fluctuation range of energy consumption of IoT devices is Wherein, c is a range constant. For level one energy consumption, c is usually 3, for level two energy consumption, c is usually 2, and for level three energy consumption, c is usually 1.

[0128] Further, if Figure 4 As shown, the energy consumption anomaly detection in S3 specifically includes:

[0129] Detect energy consumption anomalies through the fluctuation deviation D between the energy consumption value and the predicted value;

[0130] The formula for the fluctuation deviation D is:

[0131] D=Y i -k i

[0132] When |D|>δ, an abnormal situation exists, where δ is the set deviation threshold.

[0133] Furthermore, the energy consumption abnormality locations in S3 specifically include:

[0134] By the standard deviation σ of the fluctuation deviation D D Confirm the location of abnormal energy consumption;

[0135] Standard deviation D for:

[0136]

[0137] Where c is the detection constant, D i is the fluctuation deviation value of the ith one, is the fluctuation deviation D;

[0138] The mean of the fluctuation deviation D is ∣D∣>c·σ D , then there is an abnormality in the energy consumption data. If |D|<c·σ D , then there are anomalies in user behavior data.

[0139] Furthermore, the intelligent energy consumption scheduling system in S4 includes: an optimization module, an energy consumption peak smoothing module;

[0140] The optimization module will optimize the execution plan of IoT device tasks and reduce the peak energy consumption of IoT;

[0141] The energy consumption peak smoothing module dynamically adjusts the task execution time and sequence of IoT devices to reduce the energy consumption peak of the intelligent energy consumption scheduling system.

[0142] Further, if Figure 5 As shown, the optimization module steps of the intelligent energy consumption scheduling system in S4 include:

[0143] The energy consumption data E and user behavior data U are used to optimize the execution time and sequence of device tasks through the optimization module, and the device task list T is scheduled using the optimization algorithm:

[0144] Equipment task list T = {t1, t2, ..., t i};

[0145] Among them, t i represents the i-th device task;

[0146] For each device task t i , using the energy consumption data E to evaluate its expected energy consumption E i :

[0147]

[0148] Among them, w j is the weight at the jth time point, E(t i , j) represents the expected energy consumption of the i-th device task at the j-th time point;

[0149] Use the optimization algorithm to sort and allocate the time of the equipment task list T:

[0150]

[0151] Among them, T' represents all possible task scheduling schemes, T opt is the optimal scheduling solution, E i (T') is the energy consumption of the ith task under schedule T'.

[0152] Furthermore, the energy consumption peak smoothing module of the intelligent energy consumption scheduling system in S4 specifically includes:

[0153] The energy consumption peak smoothing module dynamically adjusts the task execution time and sequence of the equipment to reduce the energy consumption peak;

[0154] For each IoT device d, according to the optimized device scheduling scheme T opt , calculate the energy consumption E of IoT devices d ;

[0155]

[0156] Among them, I(t i , t) represents the indicator function, when device d performs task t at time t i Take 1 when it is, otherwise take 0;

[0157] The optimization algorithm is used to adjust the execution time and sequential calculation of the equipment to minimize the peak energy consumption E p ;

[0158]

[0159] After optimization, the energy consumption peak E is minimized. p .

[0160] Furthermore, the weight optimization of the energy consumption peak optimization mechanism in S5 specifically includes:

[0161] Weight optimization evaluates the actual performance of the system by collecting user feedback and combining it with system performance indicators. User feedback is quantitative information about energy consumption perception, equipment availability, and satisfaction.

[0162] Set the user feedback factor f u

[0163]

[0164] Among them, s i is the satisfaction score of the i-th user, s avg is the average satisfaction score, N is the number of users who participated in the feedback, and the user feedback factor f is used u Adjust and optimize weight parameters;

[0165] According to the user feedback factor f uAdjust the weights of energy consumption data E, user behavior data U, and environmental interaction data A, and update the optimization parameters;

[0166] w E '=α+f u ·λ1

[0167] w U '=β+f u ·λ2

[0168] w A '=γ+f u ·λ3

[0169] Among them, λ1, λ2, and λ3 are the learning rates used to update the weights of energy consumption data E, user behavior data U, and environmental interaction data A, respectively.

[0170] Furthermore, the threshold optimization of the energy consumption peak optimization mechanism in S5 specifically includes:

[0171] Use system performance data to determine the difference between actual energy consumption and predicted energy consumption, and adjust optimization parameters accordingly;

[0172] Define the performance difference factor f p :

[0173]

[0174] Among them, E pi is the predicted energy consumption at the i-th time point, E ai is the actual energy consumption at the i-th time point;

[0175] The update formula of the anomaly detection threshold is:

[0176] δ′=δ+f p ·η

[0177] Where η is the learning rate used to update the anomaly detection threshold;

[0178] The updating formula of the fluctuation range constant is:

[0179] c′=c+f p ·ξ

[0180] Where ξ is the learning rate used to update the fluctuation range constant.

[0181] This embodiment, through the specific description of an IoT device energy consumption optimization method based on artificial intelligence, realizes the effective management and reduction of IoT device energy consumption through multi-source data fusion, dynamic energy consumption baseline modeling, energy consumption anomaly detection and diagnosis, intelligent task scheduling and continuous optimization mechanism, thereby improving the overall energy efficiency of the system and user experience.

[0182] Example 2

[0183] Based on Example 1, this example is described in detail with examples;

[0184] There are two IoT devices d1 and d2 in a certain area. Each device generates energy consumption data E, user behavior data U, and environmental interaction data A.

[0185] The energy consumption data of devices d1 and d2 are 100 and 120 units of energy respectively;

[0186] The user behavior data U is active for 3 minutes and 4 minutes respectively;

[0187] The environmental interaction data A are temperatures of 25°C and 28°C respectively.

[0188] The data is standardized to eliminate the impact of dimension. The minimum value of energy consumption data is 100 and the maximum value is 120; the minimum value of user behavior data is 3 and the maximum value is 4; the minimum value of environmental interaction data is 25 and the maximum value is 28;

[0189] Converted into standardized data:

[0190] Device d1:X' E1 =0,X' U1 =0,X' A1 =0;

[0191] Device d2:X' E2 =1,X' U2 =1,X' A2 =1;

[0192] The correlation between features and energy consumption is calculated, assuming that the correlation coefficient between energy consumption data and energy consumption is 0.9, the correlation coefficient between user behavior data and energy consumption is 0.6, and the correlation coefficient between environmental interaction data and energy consumption is 0.75.

[0193] Correlation coefficient threshold r thresh old is set to 0.5, so all features are considered to be directly related to energy consumption.

[0194] Assume that the weight of energy consumption data is w E =0.5, the weight of user behavior data w U =0.3, weight of environmental interaction data w A =0.2;

[0195] The multidimensional fusion data set Z is:

[0196] Device d1: Z1 = 0.5 0 + 0.3 0 + 0.2 0 = 0;

[0197] Device d2: Z2 = 0.5 1 + 0.3 1 + 0.2 1 = 1;

[0198] Assume that the feature set F contains three features: energy consumption data, user behavior data, and environmental interaction data;

[0199] An energy consumption baseline model is established to predict energy consumption values. The intercept term k0 of the model is set to 50, and the feature weight coefficient k j k E =50,k U =30,k A =20;

[0200] The energy consumption baseline model is:

[0201] k i =50+50·Z E +30·Z U +20·Z A ;

[0202] Where Z E , Z U and Z A are the values ​​of energy consumption data, user behavior data, and environmental interaction data in the multidimensional fusion data set Z;

[0203] Calculate the mean of the model predictions Standard deviation p =10;

[0204] The range constant c is set to 2, so the normal fluctuation range is [90,130];

[0205] The actual observed energy consumption values ​​of devices d1 and d2 are 115 and 140 units of energy respectively.

[0206] According to the energy consumption baseline model, the predicted energy consumption values ​​are 100 and 120 units of energy consumption respectively.

[0207] The fluctuation deviations D are:

[0208] Device d1: D1 = 115-100 = 15;

[0209] Device d2: D2 = 140 - 120 = 20;

[0210] The deviation threshold δ is set to 10 units of energy consumption;

[0211] The fluctuation deviation of device d1 |D1|=15>δ, which indicates an abnormality.

[0212] The fluctuation deviation of device d2 |D2|=20>δ, which also shows abnormality.

[0213] The standard deviation σD of the volatility deviation D is assumed to be 5;

[0214] The detection constant c is set to 2;

[0215] Therefore, the standard deviation σ of the fluctuation deviation D D =5;

[0216] The fluctuation deviation of equipment d1 and d2 is greater than c·σ D =2·5=10, indicating that there are abnormalities in the energy consumption data.

[0217] The task list T of devices d1 and d2 includes four tasks:

[0218] T={t1,t2,t3,t4}

[0219] Each task t i Represents an operation that a device needs to perform;

[0220] For each device task t i , we need to use energy consumption data to evaluate its expected energy consumption E i ;

[0221] Each task can be executed at multiple time points in a day. We regard each time point as a period. A day is divided into two periods: morning and afternoon. Each period has a weight w. j , represents the impact of this period on energy consumption;

[0222] The weight of the first period (morning) is w1=0.6;

[0223] The weight of the second period (afternoon) is w2=0.4;

[0224] Assume that the expected energy consumption of each task in two periods is as follows:

[0225] t1: In the morning, E(t1,1) = 20 units of energy consumption, in the afternoon, E(t1,2) = 30 units of energy consumption;

[0226] t2: In the morning, E(t2,1) = 25 units of energy consumption, in the afternoon, E(t2,2) = 20 units of energy consumption;

[0227] t3: E(t3,1) = 35 units of energy consumption in the morning, E(t3,2) = 15 units of energy consumption in the afternoon;

[0228] t4: In the morning, E(t4,1) = 15 units of energy consumption, in the afternoon, E(t4,2) = 40 units of energy consumption;

[0229] Calculate expected energy consumption E i ;

[0230]

[0231] E1=0.6×20+0.4×30=12+12=24;

[0232] E2=0.6×25+0.4×20=15+8=23;

[0233] E3=0.6×35+0.4×15=21+6=27;

[0234] E4=0.6×15+0.4×40=9+16=25;

[0235] Use optimization algorithms to find the optimal scheduling solution T opt :

[0236] Calculate the total energy consumption of the current scheduling scheme:

[0237] E(T)=E1+E2+E3+E4=24+23+27+25=99;

[0238] T′={t1,t2,t3,t4}

[0239] Exchange the execution order of tasks. Assume that the first attempt is to exchange the positions of t1 and t4, obtain a new scheduling scheme, and evaluate the total energy consumption of the new scheduling scheme;

[0240] Calculate the total energy consumption of E(T'). If the total energy consumption of the new solution is lower, adopt the new solution; otherwise, keep the original solution unchanged until the lowest energy consumption solution is obtained.

[0241] Assume that the optimal scheduling solution is

[0242] T′={t4,t2,t3,t1}

[0243] Device d1 is responsible for executing tasks t1 and t3;

[0244] Device d2 is responsible for executing tasks t2 and t4;

[0245] Calculate the energy consumption E according to the optimized equipment scheduling plan Topt d ;

[0246] Energy consumption E of device d1 d1 :

[0247] The expected energy consumption of task t1 is E1=24;

[0248] The expected energy consumption of task t3 is E3=27.

[0249] Device d1 executes tasks t1 and t3 in the morning and afternoon respectively;

[0250] Therefore, the energy consumption of device d1 in the morning is Ed1 for:

[0251] E d1 =E1·I(t1, morning)+E3·I(t3, afternoon);

[0252] Since t1 is executed in the morning and t3 is executed in the afternoon, then: E d1 =24·1+27·1=51;

[0253] Energy consumption E of device d2 d2 :

[0254] The expected energy consumption of task t2 is E2 = 23;

[0255] The expected energy consumption of task t4 is E4=25.

[0256] Device d2 executes tasks t2 and t4 in the morning and afternoon respectively;

[0257] Therefore, the energy consumption of device d2 in the morning is E d2 For: E d2 =E2·I(t2, morning)+E4·I(t4, afternoon);

[0258] Since t2 is executed in the morning and t4 is executed in the afternoon, so: E d2 =23·1+25·1=48;

[0259] Calculate the minimized peak energy consumption E p ;

[0260] The maximum energy consumption of device d1 is 51;

[0261] The maximum energy consumption of device d2 is 48.

[0262] Minimize peak energy consumption E p :

[0263]

[0264] The energy consumption peak is minimized by dynamically adjusting the task execution time and sequence of the device.

[0265] Collect user feedback and optimize the energy consumption peak mechanism:

[0266] The user feedback factor is obtained through the feedback formula:

[0267]

[0268] User feedback factor f u =0.2;

[0269] Learning rate: λ1=0.1,λ2=0.1,λ3=0.1;

[0270] The updated weights are:

[0271] w E =w E +λ1·f u =0.5+0.1·0.2=0.52

[0272] w U =w U +λ2·f u =0.3+0.1·0.2=0.32

[0273] w A =w A +λ3·f u =0.2+0.1·0.2=0.22

[0274] The difference between actual energy consumption and predicted energy consumption is determined through system performance data, and the optimization parameters are adjusted accordingly to define the performance difference factor f p :

[0275]

[0276] Performance difference factor f p =5;

[0277] Learning rate η = 0.05, ξ = 0.05;

[0278] The update formula of the anomaly detection threshold is:

[0279] δ'=δ+f p ·η=δ+0.05·5;

[0280] The updating formula of the fluctuation range constant is:

[0281] c'=c+f p ·ξ=c+0.05·5;

[0282] The optimized anomaly detection threshold and fluctuation range constant will further optimize the scheduling scheme of the Internet of Things and reduce the peak energy consumption.

[0283] This embodiment uses a certain IoT device to introduce in detail how the intelligent energy consumption scheduling system optimizes the execution plan of device tasks, reduces energy consumption peaks, optimizes weights and thresholds through user feedback and system performance data, achieves continuous energy consumption reduction and system performance improvement, ensures that the overall system can operate efficiently and stably, and can adapt to the changing environment and user needs;

[0284] The above description is only a preferred embodiment of the present invention, which does not limit the scope of protection of the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any changes, modifications, replacements, integrations and parameter changes to these embodiments within the spirit and principles of the present invention through conventional substitutions or without departing from the principles and spirit of the present invention fall within the scope of protection of the present invention.

Claims

1. A method for optimizing energy consumption of IoT devices based on artificial intelligence, characterized in that: The following steps are involved: S1, integrate energy consumption data, user behavior data and environmental interaction data from IoT devices, and use multi-source data fusion technology to form a multi-dimensional data fusion set; S2. Dynamically establish an energy consumption baseline model using an adaptive regression algorithm based on the multi-dimensional data fusion set to reflect the normal fluctuation range of equipment energy consumption; S3. Monitor the fluctuation deviation of energy-consuming equipment through fluctuation identification technology, use energy consumption detection mechanism to quickly locate abnormal energy consumption, and diagnose the cause of abnormal energy consumption; S4, transferring energy consumption data and user behavior data to the intelligent energy consumption scheduling system to optimize the task scheduling of IoT devices, solve the causes of abnormal energy consumption, and reduce energy consumption peaks; S5. Build an energy consumption peak optimization mechanism, combine user feedback and system performance data, optimize the weights and thresholds of the Internet of Things, and achieve continuous energy consumption reduction and system performance improvement.

2. According to the method for optimizing energy consumption of IoT devices based on artificial intelligence in claim 1, it is characterized in that: S1 specifically includes: unifying the energy consumption data E, user behavior data U and environmental interaction data A in the IoT devices to the same scale to eliminate the impact of the dimension; The energy consumption data E, user behavior data U and environment interaction data A are given an original data point X with feature j j , which is converted to a standard data point X by the formula j ': Among them, min(X j ) is the minimum value of the jth feature, max(X j ) is the maximum value of the jth feature; According to the standard data point X j 'Select feature j in the IoT device and calculate the absolute correlation coefficient |ρ j ∣, according to ∣ρ j ∣>r threshold Determine the correlation between feature j and energy consumption, when |ρ j ∣>r threshold , then feature j is directly related to energy consumption, otherwise, it is not related; Among them, X j ' .i is the i-th sample of feature j after standardization, Y i is the energy consumption value of the i-th sample, and are the mean of feature j and energy consumption, n is the number of samples, r threshold is the set correlation coefficient threshold; The device E 、w U and w A are the weights of energy consumption data, user behavior data, and environmental interaction data, respectively. The multidimensional fusion data set Z is: Z=(w E E+w U U+w A A) Among them, w E is the weight of energy consumption data, w U is the weight of user behavior data, w A is the weight of the environmental interaction data.

3. According to the method for optimizing energy consumption of IoT devices based on artificial intelligence in claim 1, it is characterized in that: The S2 specifically includes: extracting a feature set F directly related to energy consumption from the multi-dimensional fusion data set Z, and adding a feature j that meets the conditions to the feature set F; Assume that the feature set F contains k features and establish an energy consumption baseline model to predict the energy consumption value k i : Among them, k0 is the intercept term, k j is the weight coefficient of the jth feature, F j is the jth feature in the feature set F, k is the number of features in the feature set F; The optimal energy consumption model is constructed by using a set of parameters so that the energy consumption value predicted by the optimal energy consumption model is close to the actually observed energy consumption value: Set the model's prediction value to k i , calculate the predicted value k i The mean and standard deviation σ p , the normal fluctuation range of energy consumption of IoT devices is Where c is the range constant.

4. According to the method for optimizing energy consumption of IoT devices based on artificial intelligence in claim 1, it is characterized in that: The energy consumption anomaly detection in S3 specifically includes: The energy consumption anomaly is detected by the fluctuation deviation D between the energy consumption value and the predicted value; The formula of the fluctuation deviation D is: D=Y i -k i When |D|>δ, an abnormal situation exists, where δ is a set deviation threshold.

5. The method for optimizing energy consumption of IoT devices based on artificial intelligence according to claim 1 or 4, characterized in that: The energy consumption abnormality locations in S3 specifically include: As mentioned above, the standard deviation σ of the fluctuation deviation D D Confirm the location of abnormal energy consumption; The standard deviation σ D for: Where c is the detection constant, D i is the fluctuation deviation value of the ith one, is the fluctuation deviation D; The mean of D is if |D|>c·σ D , then there is an abnormality in the energy consumption data. If |D|<c·σ D , then there are anomalies in user behavior data.

6. The method for optimizing energy consumption of IoT devices based on artificial intelligence according to claim 1, characterized in that: The intelligent energy consumption scheduling system in S4 includes: an optimization module and an energy consumption peak smoothing module; The optimization module will optimize the execution plan of IoT device tasks and reduce the peak energy consumption of IoT; The energy consumption peak smoothing module dynamically adjusts the task execution time and sequence of the Internet of Things devices to reduce the energy consumption peak of the intelligent energy consumption scheduling system.

7. The method for optimizing energy consumption of IoT devices based on artificial intelligence according to claim 1 or 6, characterized in that: The optimization module of the intelligent energy consumption scheduling system in S4 specifically includes: The energy consumption data E and user behavior data U are used to optimize the execution time and sequence of device tasks through the optimization module, and the device task list T is scheduled using the optimization algorithm: The equipment task list T = {t1, t2, ..., t i }; Among them, t i represents the i-th device task; For each device task t i , using the energy consumption data E to evaluate its expected energy consumption E i Among them, w j is the weight at the jth time point, E(t i , j) represents the expected energy consumption of the i-th device task at the j-th time point; The optimization algorithm is used to sort and allocate the time of the equipment task list T: Among them, T' represents all possible task scheduling schemes, T opt is the optimal scheduling solution, E i (T') is the energy consumption of the ith task under schedule T'.

8. The method for optimizing energy consumption of IoT devices based on artificial intelligence according to claim 1 or 6, characterized in that: The energy consumption peak smoothing module of the intelligent energy consumption scheduling system in S4 specifically includes: The energy consumption peak smoothing module dynamically adjusts the task execution time and sequence of the equipment to reduce the energy consumption peak; For each IoT device d, according to the optimized device scheduling scheme T opt , calculate the energy consumption E of IoT devices d ; Among them, I(t i , t) represents the indicator function, when device d performs task t at time t i Take 1 when it is, otherwise take 0; The optimization algorithm is used to adjust the execution time and sequential calculation of the equipment to minimize the peak energy consumption E p ; The energy consumption peak E obtained after the optimization is minimized p .

9. The method for optimizing energy consumption of IoT devices based on artificial intelligence according to claim 1, characterized in that: The weight optimization of the energy consumption peak optimization mechanism in S5 specifically includes: The weight optimization collects user feedback to evaluate the actual performance of the system and combines it with the system performance indicators. The user feedback is quantitative information about energy consumption perception, equipment availability and satisfaction. The setting defines the user feedback factor f u Among them, s i is the satisfaction score of the i-th user, s avg is the average satisfaction score, N is the number of users who participated in the feedback, and the user feedback factor f is used u Adjust and optimize weight parameters; According to the user feedback factor f u Adjust the weights of energy consumption data E, user behavior data U, and environmental interaction data A, and update the optimization parameters; w E '=a+f u ·λ1 w U '=β+f u ·λ2 w A '=γ+f u ·λ3 Among them, λ1, λ2, and λ3 are the learning rates used to update the weights of energy consumption data E, user behavior data U, and environmental interaction data A, respectively.

10. The method for optimizing energy consumption of IoT devices based on artificial intelligence according to claim 1, characterized in that: The threshold optimization of the energy consumption peak optimization mechanism in S5 specifically includes: The system performance data is used to determine the difference between the actual energy consumption and the predicted energy consumption, and the optimization parameters are adjusted accordingly to define the performance difference factor f p : Among them, E pi is the predicted energy consumption at the i-th time point, E ai is the actual energy consumption at the i-th time point; the updating formula of the anomaly detection threshold is: δ′=δ+f p ·or Where η is the learning rate used to update the anomaly detection threshold; The updating formula of the fluctuation range constant is: c′=c+f p ·ξ Where ξ is the learning rate used to update the fluctuation range constant.

Images