Data-driven illumination energy-saving management and control method and system

Abstract

The invention provides a data-driven illumination energy-saving management and control method and system, and the method comprises the steps: carrying out the deep mining of the historical energy consumption data of illumination equipment, analyzing the energy consumption behavior of a user through an energy consumption state change rule, and further revealing the energy consumption modes of different illumination equipment; then, the method introduces an interval type-2 fuzzy system based on a semi-tensor product, identifies a potential energy-saving moment through fuzzy reasoning and algebraic operation, cooperatively controls the on-off state of lighting equipment, and saves lighting energy consumption while meeting the lighting demand of a user.

Description

Technical Field

[0001] The present invention belongs to the technical field of lighting energy-saving control, and particularly relates to a data-driven lighting energy-saving control method and system. Background Art

[0002] Traditional lighting systems usually adopt manual, sound control, light control or timing control methods, and power supply lines and control lines need to be laid separately. Although these control methods are intuitive and easy to implement, in actual applications, the utilization rate of lamps and the user experience are often poor, and manual assistance for lamp switching is often required, which is not conducive to management.

[0003] The lighting system based on PoE (Power over Ethernet) can largely solve the above problems. Specifically, the lighting system based on PoE uses Ethernet cables to simultaneously achieve data transmission and power supply, eliminating the need for additional power lines, simplifying the system architecture, and improving the convenience of management and maintenance. In addition, the lighting system based on PoE uses Internet of Things technology to integrate multi-source sensing data to real-time sense changes in user activities and environmental conditions. This real-time sensing function provides a scientific basis for analyzing historical lighting sensing data, identifying energy consumption characteristics, and formulating personalized lighting and energy-saving strategies. At the same time, the lighting system can achieve remote status monitoring of lighting devices, further enhancing the intelligent level of the system.

[0004] Effective lighting management not only needs to ensure a comfortable light environment, but also needs to minimize energy consumption to the greatest extent, so as to achieve the dual optimization of energy conservation and intelligence while improving the user experience. In view of this, intelligent control algorithms, with their adaptability and robustness, have been applied to intelligent lighting systems. Researchers have introduced strategies such as dynamic dimming and zonal lighting, enabling the lighting system to automatically adjust the lighting brightness and switching mode according to the ambient light intensity, personnel distribution, and activity type, thereby improving energy efficiency. It should be noted that deeply exploring the energy consumption patterns of lighting devices and accurately identifying the best energy-saving timing are the keys to further optimizing lighting energy consumption management. Summary of the Invention

[0005] Based on this, the embodiments of the present invention provide a data-driven lighting energy-saving control method and system, aiming to explore the energy consumption patterns of lighting devices and accurately identify the best energy-saving timing to meet the lighting needs of users while minimizing energy consumption to the greatest extent.

[0006] The first aspect of the embodiments of the present invention provides a data-driven lighting energy-saving control method, which is applied to a scenario where a lighting device based on PoE is deployed. The lighting device based on PoE consists of several lighting devices, and each lighting device integrates a sensor and a lamp. The method includes: Obtain the historical energy consumption data of all lighting devices, and input the historical energy consumption data into the variational Bayesian Gaussian mixture model to divide the lighting devices into intervals and generate corresponding energy consumption state change sequences; Extract long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequences, and use the self-organizing mapping method and hierarchical clustering method to analyze the long-term and short-term lighting energy consumption characteristics; According to the analysis results, divide the lighting devices belonging to the same energy consumption pattern into one category, and use the GN algorithm to identify associated lighting areas; Obtain the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area, and construct a corresponding interval type-2 fuzzy system, where the matrix semi-tensor product method is used to implement fuzzy inference; Obtain lighting data in real time, and calculate the occupancy rate, energy consumption, and user activity level according to the lighting data; Input the occupancy rate, energy consumption, and user activity level into the interval type-2 fuzzy system, output the evaluation result, and control the corresponding lighting device to turn off according to the evaluation result.

[0007] Furthermore, in the variational Bayesian Gaussian mixture model, the conditional probability distribution of the latent variable is: ; The conditional probability distribution of the observed energy consumption data S is: ; where, is the mean set, is the precision matrix of Gaussian components, is a binary element, s t is the t-th energy consumption sample in the observed energy consumption data. By Bayes' formula, the joint probability distribution of the energy consumption data S and the latent variable is: ; where, and are both prior distributions, is the mixing coefficient, is the number of sampling times, is the total number of Gaussian components, is the mixing coefficient of the corresponding Gaussian component and satisfies , is a Gaussian function with a mean of and a standard deviation of .

[0008] Further, the steps of dividing the intervals of the lighting equipment and generating the corresponding sequence of energy consumption state changes include: Select components from the daily energy consumption data of all lighting equipment, and then select the component number with the highest frequency of occurrence as the target component number; After processing all the energy consumption data with the target component number through the variational Bayesian Gaussian mixture model, the energy consumption distribution of each lighting equipment is fitted by a number of Gaussian component functions, where the number of Gaussian component functions is the same as the target component number.

[0009] Further, in the step of selecting components from the daily energy consumption data of all lighting equipment and then selecting the component number with the highest frequency of occurrence as the target component number, when the number of components C = 1, it means that the distribution of the energy consumption data is fitted by one Gaussian function and no further processing is required; when the number of components C = 2, it indicates that there is an energy consumption situation in the lighting equipment, and the distribution of the energy consumption data is fitted by two Gaussian functions. Then, the low-energy consumption data is screened out, and the remaining data is processed through the variational Bayesian Gaussian mixture model, and then component selection is performed on it to determine the final number of components; when the number of components C ≥ 3, the number with the highest frequency of occurrence of the component number is selected as the basis for subsequent energy consumption classification.

[0010] Further, in the step of analyzing the long-term and short-term lighting energy consumption characteristics by using the self-organizing mapping method and the hierarchical clustering method, the expression for updating the weight vectors of the best matching unit and its neighboring neurons is: ; where, is the weight vector of neuron at time, is the learning rate, is the influence radius of the BMU, is the input data.

[0011] Further, in the step of identifying the associated lighting areas by using the GN algorithm, the modularity is used as an evaluation index to measure the partitioning result of the GN algorithm. The modularity is calculated by the following formula: ; where, is the total number of edges in the lighting network, is the adjacency matrix, indicating whether there is an edge connecting node and node , and are the degrees of node and node respectively, is a function. When the node When it belongs to the same community as the node , otherwise, . .

[0012] Furthermore, the steps of obtaining the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area and constructing a corresponding interval type-2 fuzzy system include: Construct a corresponding type-1 fuzzy model according to the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area; Extract the membership functions of the type-1 fuzzy sets with the same occupancy rate, energy consumption, and user activity level in the type-1 fuzzy model, and then perform integration processing to obtain an interval type-2 fuzzy set.

[0013] Furthermore, in the steps of obtaining the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area and constructing a corresponding interval type-2 fuzzy system, vectorization processing is performed on the input, rule base, and output of the interval type-2 fuzzy system.

[0014] Furthermore, in the process of implementing fuzzy inference using the matrix semi-tensor product method, an algebraic form interval structure matrix is constructed to replace the original rule base for inference operations.

[0015] The second aspect of the embodiments of the present invention provides a data-driven lighting energy-saving control system for implementing the data-driven lighting energy-saving control method described in the first aspect. The system includes: An acquisition module for acquiring the historical energy consumption data of all lighting devices, inputting the historical energy consumption data into a variational Bayesian Gaussian mixture model to perform interval division on the lighting devices, and generating a corresponding energy consumption state change sequence; An analysis module for extracting long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequence, and analyzing the long-term and short-term lighting energy consumption characteristics using a self-organizing mapping method and a hierarchical clustering method; An identification module for classifying the lighting devices belonging to the same energy consumption pattern into one category according to the analysis results, and identifying the associated lighting areas using the GN algorithm; A construction module for obtaining the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area and constructing a corresponding interval type-2 fuzzy system, where the matrix semi-tensor product method is used to implement fuzzy inference; A calculation module for real-time acquiring lighting data and calculating the occupancy rate, energy consumption, and user activity level according to the lighting data; A control module, configured to input the occupancy rate, energy consumption, and user activity level into the interval type-2 fuzzy system, output an evaluation result, and control the turning off of corresponding lighting devices according to the evaluation result.

[0016] The beneficial effects of a data-driven lighting energy-saving control method and system provided by the present invention are as follows: By deeply mining the historical energy consumption data of lighting devices, analyzing the user's energy consumption behavior from the change law of the energy consumption state, and further revealing the energy consumption patterns of different lighting devices; then, the method introduces an interval type-2 fuzzy system based on semi-tensor product, identifies potential energy-saving moments through fuzzy reasoning and algebraic operations, and collaboratively controls the on / off state of lighting devices to save lighting energy while meeting the user's lighting needs. Description of the Drawings

[0017] Figure 1 It is a schematic indoor floor plan of a lighting device based on PoE; Figure 2 It is a flowchart for implementing a data-driven lighting energy-saving control method provided in Embodiment 1 of the present invention; Figure 3 It is a schematic diagram for dividing low and high energy consumption intervals; Figure 4 It is a schematic diagram for dividing medium and high energy consumption intervals; Figure 5 It is a schematic diagram of the structure of an interval type-2 fuzzy system based on semi-tensor product; Figure 6 It is an energy-saving effect diagram; Figure 7 It is a block diagram of the structure of a data-driven lighting energy-saving control system provided in Embodiment 3 of the present invention. Detailed Embodiments

[0018] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0019] It should be noted that when an element is referred to as being "fixedly provided on" another element, it can be directly on the other element or there may also be an intermediate element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intermediate element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this invention belongs. The terms used in the description of the present invention herein are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0021] Embodiment 1 According to an embodiment of the present invention, there is provided an embodiment of a data-driven lighting energy-saving control method. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0022] In Embodiment 1 of the present invention, a data-driven lighting energy-saving control method is provided, which can be used in electronic devices such as computers. It should be noted that this lighting energy-saving control method is applied to the scenario of deploying PoE-based lighting devices. The PoE-based lighting devices are composed of several lighting devices, and each lighting device integrates sensors and lamps. Among them, the sensors include at least a voltage sensor, a current sensor, and a human body infrared detection sensor. In addition, the lighting devices are arranged in parallel on the ceiling of each indoor scene. Please refer to Figure 1 , which is a schematic indoor floor plan of the PoE-based lighting device.

[0023] Specifically, please refer to Figure 2 , Figure 2 which shows a flowchart of the implementation of a data-driven lighting energy-saving control method provided in Embodiment 1 of the present invention, specifically including steps S01 to S06.

[0024] Step S01, obtain the historical energy consumption data of all lighting devices, and input the historical energy consumption data into a variational Bayesian Gaussian mixture model to divide the intervals of the lighting devices and generate corresponding energy consumption state change sequences.

[0025] Specifically, the energy consumption sampling sequence of a certain lighting device within a certain time period with a sampling period of can be described as , where represents the number of samplings. Considering that the actual energy consumption-frequency distribution of lighting devices is usually in the "bimodal" or "multimodal" situation, a linear combination of multiple Gaussian components in the Gaussian mixture model (GMM) is used to describe the probability distribution of the data. The energy consumption sequence can be considered to be independently sampled from a probability density function, and this probability density function can be expressed as: ; wherein, is the total number of Gaussian components, is the mixing coefficient of the corresponding Gaussian component and satisfies , represents a Gaussian function with a mean of and a standard deviation of . It should be noted that due to insufficient prior knowledge of the number of modes during the iterative process of GMM, it is easily affected by local optima.

[0026] Therefore, the present invention adopts a variational Bayesian Gaussian mixture model to automatically determine the optimal number of modes through variational inference, and at the same time improves the generalization ability of the model by introducing a prior distribution. For the observed energy consumption data there is a corresponding latent variable , is a "1-of-K" binary vector, and the vector elements are , that is, are binary elements, wherein, , . Given the mixing coefficient of the component, the conditional probability distribution of the latent variable can be obtained: ; Similarly, given the latent variable and the mixing coefficient , the conditional probability distribution of the observed energy consumption data S can be obtained as: ; wherein, is the mean set, is the precision matrix of the Gaussian components, s t is the t-th energy consumption sample in the observed energy consumption data. Further, from Bayes' formula, the joint probability distribution of the measured energy consumption data S and the latent variable can be obtained as: ; wherein, and are both prior distributions. Further, the optimal model parameter distribution function is determined through an iterative process of maximizing the expectation of the joint probability distribution. During the iterative process, the Gaussian components with the mixing coefficient tending to 0 will be gradually eliminated to determine the optimal number of Gaussian components, and at the same time, the parameters of the Gaussian components are updated according to Bayesian inference, so that the overall model can effectively describe the distribution characteristics of the energy consumption data.

[0027] Since the number of Gaussian components fitted for the daily energy consumption data of each lighting device is different, in order to further study the energy consumption pattern of lighting devices, it is necessary to unify the number of Gaussian components of lighting devices. This process is called component selection. Assume the following three cases: When the number of components C = 1, it means that the distribution of energy consumption data is fitted by one Gaussian function and no further processing is required; when the number of components C = 2, it indicates that there is a certain energy consumption situation in the lighting device, and the distribution of energy consumption data is fitted by two Gaussian functions. Then, the low-energy consumption data is screened out, and the remaining data is processed by the variational Bayesian Gaussian mixture model, and then component selection is performed on it to determine the final number of components; when the number of components C ≥ 3, the number with the highest frequency of occurrence of the number of components is selected as the basis for subsequent energy consumption classification. Exemplarily, please refer to Figure 3 , after the energy consumption data is processed by the variational Bayesian Gaussian mixture model, 2 components are obtained, namely the low and high energy consumption intervals, and the boundary line of the energy consumption interval is determined by the intersection point of the Gaussian component functions. Subsequently, the low-energy consumption data is excluded, and the remaining data is processed by the variational Bayesian Gaussian mixture model again. Please refer to Figure 4 , and 2 new components are obtained, namely the medium energy consumption interval and the high energy consumption interval. Through these two steps, a total of three energy consumption intervals are obtained, and the final number of components obtained is determined to be 3.

[0028] Further, after component selection is performed on the daily energy consumption data of all lighting devices, the number of components with the highest frequency of occurrence is selected as the final number of components, that is, the target number of components. All energy consumption data is reprocessed by the variational Bayesian Gaussian mixture model with the final number of components , and the energy consumption distribution of each lighting device is fitted by Gaussian component functions. It should be noted that due to certain differences in the application scenarios, usage conditions, and rated power consumption of each lighting device, the interval division of its energy consumption data is also different. The daily energy consumption data of the lighting device is mapped to an energy consumption state change sequence , where is the serial number of the lighting device, is the recording date, and this sequence not only characterizes the activity pattern of the user but also reflects the energy consumption pattern of the lighting device itself in the long-term dimension.

[0029] Step S02, extract the long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequence, and use the self-organizing mapping method and the hierarchical clustering method to analyze the long-term and short-term lighting energy consumption characteristics.

[0030] It should be noted that the long-term energy consumption characteristics of lighting devices include: the longest continuous on time per day, the longest continuous off time, the total energy consumption, the number of turn-ons, etc.; the short-term energy consumption characteristics include: the low, medium, and high energy consumption ratios per day, the switching frequencies of low, medium, and high energy consumption per day, etc. Extract the long-term and short-term energy consumption characteristics of lighting devices over several months, and convert them into matrix form as the input data for self-organizing mapping. During the training process of self-organizing mapping, first calculate the shortest distance between the weight vectors of all neurons and each input data, and select the neuron closest to the input data as the Best Matching Unit (BMU). Subsequently, according to the following rules, update the weight vectors of the BMU and its neighboring neurons: ; where, is the weight vector of neuron at moment, is the learning rate, is the influence radius of the BMU, is the input data. As the training progresses, the network will gradually adjust the weights, and gradually decrease, so that similar input data are mapped to adjacent neurons. Then, use agglomerative hierarchical clustering to cluster the neurons containing features in the self-organizing mapping from bottom to top. Thus, lighting devices are finally clustered into categories. It should be noted that although the spatial positions of different lighting devices are different, through the analysis of the energy consumption patterns, it is still possible to reveal groups of devices with similar energy consumption patterns.

[0031] Step S03, according to the analysis results, divide the lighting devices belonging to the same energy consumption pattern into one category, and use the GN algorithm to identify the associated lighting areas.

[0032] In the lighting system, the activity behaviors of users show certain regularities in some areas or at specific times. In order to analyze the dependence relationship formed by the usage rules of spatially adjacent lighting devices during long-term operation, the present invention uses mutual information to measure the correlation between lighting devices, and combines the GN algorithm to refine the division of lighting areas. First, the entire PoE lighting system can be modeled as a large-scale lighting network, where each lighting device is regarded as a node in the network, and the edge weights of the network reflect the correlation between lighting nodes. The edge weights between two lighting nodes are calculated by the mutual information method. The mutual information between lighting node and lighting node is calculated by the following formula: ; Among them, and are the sequences of energy consumption changes of lighting nodes and respectively, is a specific value in is a specific value in is and the joint probability distribution of and are the marginal probability distributions of and respectively. Specifically, taking a week as the time unit, using non - overlapping sliding time windows, pairwise calculating the mutual information values between lighting nodes to construct a mutual information matrix, effectively measuring the long - term energy consumption correlation between lighting devices. It should be noted that for the edges with significantly lower weights in the lighting network, they will be removed to form a preliminary lighting network topology. Then, the GN algorithm is used to partition the lighting network. The network is split by iteratively removing the edges with the maximum edge betweenness value relative to all source nodes in the network, so as to identify the closely related community structure in the lighting network. Since the GN algorithm cannot determine the termination position, therefore, the present invention uses modularity as an evaluation index to measure the partitioning result of the GN algorithm. Modularity is calculated by the following formula: ; Among them, is the total number of edges in the lighting network, is the adjacency matrix, indicating whether there is an edge connecting node and node , and are the degrees of node and node respectively, is a function. When node and node belong to the same community, , otherwise, . Compare the modularity of each partitioning result, and select the partitioning scheme with a larger or the largest modularity as the final partitioning result of the lighting network. By identifying lighting areas with relatively high internal correlation, it is not only beneficial to the subsequent collaborative control of lighting groups, but also provides a reliable basis for the extraction of fuzzy variables in the fuzzy system.

[0033] Step S04: Obtain the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area, and construct a corresponding interval type-2 fuzzy system, where matrix semi-tensor product method is used to implement fuzzy inference.

[0034] In the embodiments of the present invention, the occupancy rate is defined as the ratio of the cumulative on-time of the lighting device within the sampling period to the sampling period divided by the number of lighting devices, so as to characterize the overall usage of the lighting system during this time period; the energy consumption of the lighting device is defined as the sum of the on-energy consumption and standby energy consumption within the sampling period ; the calculation formula for the user activity level is: , where is the number of lighting devices in a certain lighting area divided by the GN algorithm, is the th lighting device, and the number of detected human infrared induction times within the sampling period , and , is corresponding weight coefficient, and the UAL index is used to quantify the activity level of user behavior in the lighting area.

[0035] Please refer to Figure 5 , which is a schematic diagram of the interval type-2 fuzzy system structure based on semi-tensor product. The fuzzy system consists of five parts: fuzzification, fuzzy rule base, fuzzy inference, defuzzifier, and de-fuzzification. It should be noted that the algebraic operation based on semi-tensor product can simplify the fuzzy inference process and make the calculation more efficient.

[0036] Specifically, an interval type-2 fuzzy set is used to perform fuzzification processing on the accurate input variables. The membership degree of each element in its domain of discourse becomes an interval instead of a definite value, so as to more effectively characterize the uncertainty between different lighting devices. The interval type-2 fuzzy set is a special case of the type-2 fuzzy set, and its secondary membership function . Therefore, the interval type-2 fuzzy set on the domain of discourse can be expressed by the following formula: ; where represents the main variable with the domain of discourse , is the secondary variable, is the main membership degree. The closed area formed by the union of all main membership degree values of the interval type-2 fuzzy set is the footprint of uncertainty (FOU), and its upper and lower membership functions are and , thus, the primary membership degree of the interval type-2 fuzzy set can be expressed as . .

[0037] Given that the construction of interval type-2 fuzzy sets usually relies on expert experience, and expert knowledge has problems such as strong subjectivity and difficulty in acquisition. Therefore, the present invention proposes a data-driven method for constructing interval type-2 fuzzy sets, that is, generating type-1 fuzzy sets by statistical analysis of historical lighting data, and then integrating the type-1 fuzzy sets to transform them into interval type-2 fuzzy sets. Specifically, among the types of lighting equipment classified by means of self-organizing mapping method and hierarchical clustering method, occupancy rate, energy consumption and user activity level indicators are extracted for each type of equipment respectively, and the corresponding type-1 fuzzy models are constructed. Then, the membership functions of the type-1 fuzzy sets of the same indicators in each type-1 fuzzy model are extracted, and then integrated to obtain an interval type-2 fuzzy set, so that the constructed interval type-2 fuzzy set is more in line with the actual situation of the lighting system, and the robustness and accuracy of the fuzzy system are improved.

[0038] More specifically, for a specific input variable , represents the number of input variables. After the input variables are fuzzified, an interval membership degree value will be mapped on each interval type-2 fuzzy set , where , is the number of fuzzy sets. For the convenience of subsequent algebraic operations, the interval membership degree value of the input variable is expressed in vector form: ; For the fuzzy rule base, the interval type-2 fuzzy set corresponding to the input variable , and the interval type-2 fuzzy set corresponding to the output variable is , then the fuzzy rule can be expressed as: ; where represents the th fuzzy rule, represents the fuzzy set of the antecedent of the th rule, represents the fuzzy set of the consequent of the th rule, represents the number of fuzzy sets of the consequent of the rule, represents the number of fuzzy rules.

[0039] Generally speaking, the logical relationship in fuzzy rules can be regarded as a mathematical mapping from the input space to the output space. This fuzzy logical relationship can be represented by a structure matrix in the matrix semi-tensor product method. Therefore, constructing an interval structure matrix in algebraic form can replace the original fuzzy rule base for inference operations. Specifically, for the input variable and the output variable corresponding interval type-2 fuzzy sets and , the elements in the fuzzy sets are represented in vector form: ; where is the shorthand form of , represents an interval number between 0 and 1, whose upper bound is and the lower bound is . is the th column of the identity matrix , represents the order of the identity matrix; similarly, the fuzzy set of the antecedent and the fuzzy set of the consequent of the th fuzzy rule can be represented in vector form: ; Furthermore, the structure interval matrix of the fuzzy set of the antecedent and the fuzzy set of the consequent of the th fuzzy rule can be represented as: ; where represents the semi-tensor product operation of matrices. Then, construct the structure interval matrix of the th fuzzy rule: ; Then, obtain the structure interval matrix representing the interval type-2 fuzzy relationship: ; where represents the Boolean addition of interval matrices. Through the fuzzy relation matrix, the input data is mapped to the corresponding vectorized output to obtain the result of fuzzy inference: ; ​​​For the output part of the interval type-2 fuzzy system, the present invention uses the Nie-Tan (NT) algorithm to directly calculate the defuzzified output value of the interval type-2 fuzzy set. It should be noted that this algorithm does not require iterative operations and skips the type-reduction step of the conventional interval type-2 fuzzy system, reducing the computational complexity. The formula for calculating the centroid of the interval type-2 fuzzy set is: ; where , thus obtaining the centroid set corresponding to the consequent fuzzy set , and its vector form is . Then, the output of the final fuzzy system is calculated according to the following formula: ; where is a -dimensional interval vector.

[0040] Step S05: Obtain lighting data in real time, and calculate the occupancy rate, energy consumption, and user activity level according to the lighting data.

[0041] It can be understood that according to the calculation formulas for the occupancy rate, energy consumption, and user activity level disclosed in step S04, after obtaining the lighting data in real time, the corresponding occupancy rate, energy consumption, and user activity level are calculated.

[0042] Step S06: Input the occupancy rate, energy consumption, and user activity level into the interval type-2 fuzzy system, output an evaluation result, and control the corresponding lighting device to turn off according to the evaluation result.

[0043] In view of the possible energy-saving differences of lighting devices in different lighting areas and different time periods, the present invention adopts a dynamic threshold control method to realize the conversion from the output value of the fuzzy system to the actual control action, as follows: ; where represents the serial number of the lighting device, , , represents the threshold value of the lighting device in the lighting area at time , and its value is related to the lighting area and time. When , it means to execute the "turn off" control action on the lighting device, and when

[0044] In summary, for the data-driven lighting energy-saving control method in the above embodiments of the present invention, the method obtains the historical energy consumption data of all lighting devices, and inputs the historical energy consumption data into the variational Bayesian Gaussian mixture model to divide the intervals of the lighting devices and generate corresponding energy consumption state change sequences. Extract long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequence, and use the self-organizing mapping method and hierarchical clustering method to analyze the long-term and short-term lighting energy consumption characteristics; according to the analysis results, divide the lighting devices belonging to the same energy consumption mode into one category, and use the GN algorithm to identify associated lighting areas; obtain the current occupancy rate, energy consumption and user activity level of the lighting devices in each lighting area, construct a corresponding interval type-2 fuzzy system, and use the matrix semi-tensor product method to realize fuzzy inference; according to the output result of the interval type-2 fuzzy system, control the corresponding lighting device to turn off to minimize energy consumption while meeting the user's lighting needs.

[0045] Embodiment 2 To better understand the technical means of a data-driven lighting energy-saving control method in Embodiment 1 of the present invention, a specific example is given in Embodiment 2 of the present invention. Specifically, first deploy a PoE-based lighting system indoors. The indoor area is divided into 32 places according to its usage function, and 1 to 16 lighting devices are arranged in each place. There are 164 lighting devices in all places. Input the energy consumption data S of the 164 lighting devices in two months into the variational Bayesian Gaussian mixture model to divide the intervals of the energy consumption levels of the devices and generate corresponding energy consumption state change sequences. , Subsequently, extract long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequence, use the self-organizing mapping method and hierarchical clustering method to analyze the energy consumption patterns of the lighting devices, and divide the 164 lighting devices into 5 lighting groups. Further, use the GN algorithm to identify 24 lighting areas with high internal correlation, and calculate the occupancy rate, energy consumption and user activity level of the 5 lighting groups in the 24 lighting areas in the past two months with a sampling period minutes. Construct corresponding interval type-2 fuzzy sets according to these index data, vectorize the input, rule base and output processes of the fuzzy system, and use the matrix semi-tensor product method to realize fuzzy inference. Finally, the interval type-2 fuzzy system identifies the lighting devices that cause waste of lighting resources in the lighting system and turns them off.

[0046] Apply the lighting energy-saving control method proposed by the present invention to the PoE-based lighting system. It should be noted that the PoE-based lighting system already has the basic energy-saving function of "turning off the lights when people leave", and the lighting energy-saving control method proposed by the present invention is a further energy-saving optimization on this basis. The energy-saving effect of the lighting system in 2 weeks is as Figure 6As shown, overall, the daily energy-saving rate remains between 2% and 12%, demonstrating a certain lighting energy-saving effect. It is noted that on Saturday and Sunday of the second week, both the energy consumption savings and the energy-saving rate of the lighting system are 0. This is because weekends are rest days, and most lighting equipment is in the standby state of not being used, so the energy-saving space is relatively limited. It should be particularly pointed out that in some lighting areas, the energy-saving rate of individual lighting equipment can exceed 15%, while the overall energy-saving rate of the lighting system remains between 2% and 12%. This is because the energy consumption of lighting equipment consists of two parts: the on energy consumption during operation and the standby energy consumption when not in use. In practice, a considerable number of lighting equipment remains in the standby state every day, thus dragging down the overall energy-saving rate.

[0047] Embodiment 3 Please refer to Figure 7 , Figure 7 which is a structural block diagram of a data-driven lighting energy-saving control system provided by Embodiment 3 of the present invention. The data-driven lighting energy-saving control system 200 is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0048] Specifically, the data-driven lighting energy-saving control system 200 includes: an acquisition module 21, an analysis module 22, an identification module 23, a construction module 24, a calculation module 25, and a control module 26, where: The acquisition module 21 is used to acquire the historical energy consumption data of all lighting equipment, and input the historical energy consumption data into the variational Bayesian Gaussian mixture model to divide the intervals of the lighting equipment and generate the corresponding energy consumption state change sequence. Among them, the conditional probability distribution of the latent variable is: ; The conditional probability distribution of the observed energy consumption data S is: ; Among them, is the mean set, is the precision matrix of the Gaussian components, is the binary element, s t is the t-th energy consumption sample in the observed energy consumption data. By Bayes' formula, the joint probability distribution of the energy consumption data S and the latent variable is: ; Among them, and are both prior distributions, is the mixing coefficient, is the number of sampling times, is the total number of Gaussian components, is the mixing coefficient of the corresponding Gaussian component and satisfies , is a Gaussian function with a mean of and a standard deviation of ; The analysis module 22 is used to extract long - and short - term lighting energy consumption characteristics from the sequence of energy consumption state changes, and analyze the long - and short - term lighting energy consumption characteristics by using the self - organizing mapping method and the hierarchical clustering method. Among them, the expression for updating the weight vectors of the best - matching unit and its neighboring neurons is: ; where, is the weight vector of neuron at time , is the learning rate, is the influence radius of the BMU, is the input data; The recognition module 23 is used to divide the lighting devices belonging to the same energy consumption mode into one category according to the analysis results, and identify the associated lighting areas by using the GN algorithm. Among them, modularity is used as an evaluation index to measure the division results of the GN algorithm. Modularity is calculated by the following formula: ; where, is the total number of edges in the lighting network, is the adjacency matrix, indicating whether there is an edge connecting node and node , and are the degrees of node and node respectively, is a function. When node and node belong to the same community, , otherwise, ; The construction module 24 is used to obtain the historical occupancy rate, historical energy consumption and historical user activity level of various lighting devices in the lighting area, and construct a corresponding interval type - 2 fuzzy system. Among them, matrix semi - tensor product method is used to realize fuzzy inference, and the input, rule base and output of the interval type - 2 fuzzy system are vectorized. In addition, an algebraic form of interval structure matrix is constructed to replace the original rule base for inference operation; A calculation module 25, configured to obtain lighting data in real time, and calculate an occupancy rate, energy consumption, and user activity level according to the lighting data; A control module 26, configured to input the occupancy rate, energy consumption, and user activity level into the interval type-2 fuzzy system, output an evaluation result, and control the corresponding lighting device to turn off according to the evaluation result.

[0049] Further, in some other embodiments of the present invention, the obtaining module 21 includes: A selection unit, configured to perform component selection on the daily energy consumption data of all lighting devices, and then select the component number with the most occurrences as the target component number. Among them, when the component number C = 1, it means that the distribution of the energy consumption data is fitted by a Gaussian function and no further processing is required; when the component number C = 2, it indicates that there is an energy consumption situation in the lighting device, and the distribution of the energy consumption data is fitted by two Gaussian functions. Then, the low-energy consumption data is screened out, and the remaining data is processed by a variational Bayesian Gaussian mixture model, and then component selection is performed on it to determine the final component number; when the component number C ≥ 3, the number with the highest occurrence frequency of the component number is selected as the basis for subsequent energy consumption classification; A fitting unit, configured to process all energy consumption data with the target component number through a variational Bayesian Gaussian mixture model, and the energy consumption distribution of each lighting device is fitted by several Gaussian component functions, where the number of Gaussian component functions is the same as the target component number.

[0050] Further, in some other embodiments of the present invention, the construction module 24 includes: A construction unit, configured to construct a corresponding type-1 fuzzy model according to the historical occupancy rate, historical energy consumption, and historical user activity level of various lighting devices in the lighting area; An extraction unit, configured to extract the membership functions of the type-1 fuzzy sets with the same occupancy rate, energy consumption, and user activity level in the type-1 fuzzy model, and then perform integration processing to obtain an interval type-2 fuzzy set.

[0051] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0052] The above embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

Claims

1. A data-driven lighting energy-saving management and control method, characterized in that: Applied to a scenario where a PoE-based lighting device is deployed, the PoE-based lighting device is composed of a plurality of lighting devices, each of which integrates a sensor and a lamp, the method comprises: Obtain the historical energy consumption data of all lighting equipment, and input the historical energy consumption data into the variational Bayesian Gaussian mixture model to divide the lighting equipment into intervals and generate the corresponding energy consumption state change sequence; Extracting long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequence, and analyzing the long-term and short-term lighting energy consumption characteristics by using a self-organizing mapping method and a hierarchical clustering method; According to the analysis results, lighting devices with the same energy consumption mode are classified into one category, and the GN algorithm is used to identify the associated lighting areas; Obtain the historical occupancy rate, historical energy consumption and historical user activity level of various lighting equipment in the lighting area, and construct the corresponding interval type-2 fuzzy system, in which the matrix semi-tensor product method is used to realize fuzzy reasoning; acquiring lighting data in real time, and calculating occupancy, energy consumption, and user activity levels based on the lighting data; The occupancy rate, energy consumption and user activity level are input into the interval type-II fuzzy system, an evaluation result is output, and according to the evaluation result, the corresponding lighting equipment is controlled to be turned off.

2. The data-driven lighting energy-saving management and control method according to claim 1, characterized in that: In the variational Bayesian Gaussian mixture model, the latent variable The conditional probability distribution of : ； The conditional probability distribution of observed energy consumption data S is: ； in, is the mean set, for The precision matrix of Gaussian components, is a binary element, s t For the tth energy consumption sample in the observed energy consumption data, the energy consumption data S and the latent variable are obtained by the Bayesian formula. The joint probability distribution of is: ； in, and are all prior distributions, is the mixing coefficient, is the sampling number, is the total number of Gaussian components, is the mixing coefficient of the corresponding Gaussian component and satisfies , The mean is And the standard deviation is Gaussian function.

3. The data-driven lighting energy-saving management and control method according to claim 2, characterized in that: The step of dividing the lighting equipment into intervals and generating a corresponding energy consumption state change sequence includes: Component selection is performed on the daily energy consumption data of all lighting equipment, and then the number of components with the highest frequency of occurrence is selected as the target number of components; After all energy consumption data are processed by the variational Bayesian Gaussian mixture model with a target number of components, the energy consumption distribution of each lighting device is fitted by a number of Gaussian component functions, wherein the number of Gaussian component functions is the same as the target number of components.

4. The data-driven lighting energy-saving management and control method according to claim 3 is characterized in that: In the step of performing component selection on the daily energy consumption data of all lighting devices and then selecting the number of components with the highest frequency as the target number of components, when the number of components C=1, it means that the distribution of the energy consumption data is fitted by a Gaussian function and no further processing is performed; when the number of components C=2, it means that the lighting equipment has energy consumption, and the distribution of the energy consumption data is fitted by two Gaussian functions, then the low-energy consumption data is screened out, and the remaining data is processed by a variational Bayesian Gaussian mixture model, and then component selection is performed on it to determine the final number of components; when the number of components C≥3, the number of components with the highest frequency is selected as the basis for subsequent energy consumption classification.

5. The data-driven lighting energy-saving management and control method according to claim 4, characterized in that: In the step of analyzing the long-term and short-term lighting energy consumption characteristics by using the self-organizing mapping method and the hierarchical clustering method, the expression for updating the weight vector of the best matching unit and its neighborhood neurons is: ； in, For neurons exist The weight vector at time, is the learning rate, is the influence radius of the BMU, For input data.

6. The data-driven lighting energy-saving management and control method according to claim 5, characterized in that: In the step of using the GN algorithm to identify the associated lighting areas, modularity is used as an evaluation index to measure the division result of the GN algorithm. Calculated by the following formula: ； in, is the total number of edges in the lighting network, is the adjacency matrix, indicating whether there are edges connecting nodes and nodes , and Node and nodes The degree, As a function, when the node and nodes When they belong to the same community, ,otherwise, .

7. The data-driven lighting energy-saving management and control method according to claim 6, characterized in that: The steps of obtaining the historical occupancy rate, historical energy consumption and historical user activity level of various lighting equipment in the lighting area and constructing the corresponding interval type-2 fuzzy system include: According to the historical occupancy rate, historical energy consumption and historical user activity level of various lighting equipment in the lighting area, a corresponding type I fuzzy model is constructed; The membership functions of the type-one fuzzy sets with the same occupancy rate, energy consumption and user activity level in the type-one fuzzy model are extracted and then integrated to obtain interval type-two fuzzy sets.

8. The data-driven lighting energy-saving management and control method according to claim 7, characterized in that: In the step of obtaining the historical occupancy rate, historical energy consumption and historical user activity level of various lighting equipment in the lighting area and constructing the corresponding interval type-2 fuzzy system, the input, rule base and output of the interval type-2 fuzzy system are vectorized.

9. The data-driven lighting energy-saving management and control method according to claim 8, characterized in that: In the process of realizing fuzzy reasoning by using matrix semi-tensor product method, an interval structure matrix in algebraic form is constructed to replace the original rule base for reasoning operation.

10. A data-driven lighting energy-saving management and control system, characterized in that: A data-driven lighting energy-saving management and control method according to any one of claims 1 to 9, the system comprising: An acquisition module is used to acquire the historical energy consumption data of all lighting devices and input the historical energy consumption data into a variational Bayesian-Gaussian mixture model to divide the lighting devices into intervals and generate a corresponding energy consumption state change sequence; An analysis module, used to extract long-term and short-term lighting energy consumption characteristics from the energy consumption state change sequence, and analyze the long-term and short-term lighting energy consumption characteristics by using a self-organizing mapping method and a hierarchical clustering method; An identification module is used to classify lighting devices belonging to the same energy consumption mode into one category according to the analysis results, and identify the associated lighting areas using the GN algorithm; A construction module is used to obtain the historical occupancy rate, historical energy consumption and historical user activity level of various lighting equipment in the lighting area, and construct the corresponding interval type-2 fuzzy system, in which the matrix semi-tensor product method is used to realize fuzzy reasoning; A calculation module, used to obtain lighting data in real time, and calculate occupancy rate, energy consumption and user activity level based on the lighting data; The control module is used to input the occupancy rate, energy consumption and user activity level into the interval type-II fuzzy system, output an evaluation result, and control the shutdown of the corresponding lighting device according to the evaluation result.

Images