Big data-based street lamp cluster cooperative power saving control method and system

Abstract

This invention discloses a method and system for collaborative energy-saving control of street light clusters based on big data. The method includes acquiring standard data, dividing street light clusters, predicting cluster lighting demand, collaborative energy-saving control of adjacent clusters of street lights, collaborative energy-saving control of street lights within a cluster, and intelligent control of street lights. This invention belongs to the field of data processing technology, specifically a method and system for collaborative energy-saving control of street light clusters based on big data. This solution creatively performs collaborative control of street lights in adjacent clusters and within clusters separately, reducing the energy consumption of street light operation. An improved clustering algorithm is used, employing a multi-level diffused neighbor search strategy for coarse-grained grid block generation, dual-density calculation, and a dual-density progressive comparison mechanism, to improve the accuracy of clustering results and street light cluster division results. An objective function for optimizing street light brightness within a cluster is constructed, and the optimization algorithm is improved based on a dynamic elite mutation position update strategy, reducing the overall energy consumption of the street light cluster and improving the accuracy of the collaborative energy-saving control strategy.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically a method and system for collaborative energy-saving control of street light clusters based on big data. Background Technology

[0002] The collaborative energy-saving control method and system for street light clusters is an intelligent regulation and management system for urban street light clusters. By collecting multi-source heterogeneous data and relying on big data analysis and artificial intelligence algorithms, it can dynamically assess the street lighting demand in different areas and at different times. Under the premise of strictly meeting road lighting safety standards, it can achieve collaborative energy-saving control such as graded dimming, on-demand lighting, and peak-shaving regulation for street light clusters in multiple road sections and areas, significantly reducing the overall energy consumption of street light clusters and providing refined and intelligent energy-saving management support for municipal lighting management.

[0003] However, traditional street light energy-saving control methods suffer from technical problems such as dispersed street light control objects, coarse area division, inaccurate matching of lighting demand, delayed dimming response, insufficient coordination between adjacent areas, and high energy consumption during street light operation. Existing clustering methods suitable for street light cluster division suffer from high computational complexity, strong parameter sensitivity, poor generalization ability, and difficulty in applying to large-scale street light datasets. They are also prone to misidentification of cluster centers and cascading errors in cluster assignment. Traditional energy-saving control strategies often rely on manual experience settings, making it difficult to simultaneously achieve energy consumption reduction, brightness smoothing, and dimming stability. Furthermore, the optimization algorithms suffer from low search efficiency, slow convergence speed, and getting trapped in local optima, resulting in limited energy-saving effects and unstable energy-saving performance. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a big data-based collaborative energy-saving control method and system for streetlight clusters. Addressing the technical problems of traditional streetlight energy-saving control methods, such as dispersed streetlight control objects, coarse area division, inaccurate matching of lighting demand, delayed dimming response, insufficient coordination between adjacent areas, and high energy consumption during streetlight operation, this solution creatively employs a clustering algorithm for streetlight cluster division and cluster lighting demand prediction. Subsequently, combining the basic brightness coordination of adjacent clusters and the optimization of differentiated target brightness within clusters, collaborative energy-saving control is implemented for each streetlight cluster and each streetlight within a cluster, enhancing the brightness connection between adjacent clusters. To ensure the continuity and safety of lighting in key areas, and ultimately reduce the energy consumption of streetlights, this study aims to improve the precision, intelligence, and coordination of urban streetlight energy-saving control while maintaining road lighting safety, brightness continuity, and dimming stability. This will reduce streetlight operating energy consumption and management costs, achieving collaborative energy-saving intelligent control of large-scale streetlight networks. The study also addresses existing clustering methods for streetlight clustering, which suffer from high computational complexity, strong parameter sensitivity, poor generalization ability, and difficulty in applying to large-scale streetlight datasets. Furthermore, these methods are prone to misidentification of cluster centers and chain errors in cluster assignment, leading to inaccurate streetlight cluster grouping. The cluster control system is fundamentally unstable. This solution innovatively improves the clustering algorithm by using a multi-level diffused neighbor search strategy for coarse-grained grid block generation, dual-density calculation, and a dual-density progressive comparison mechanism. This enhances the accuracy of the clustering results, improving the accuracy, stability, and reliability of street light cluster partitioning. It achieves efficient, accurate, and adaptive partitioning of street light clusters in large-scale street light networks, thereby strengthening the overall cluster control and improving the accuracy of subsequent lighting demand prediction and energy-saving control. Furthermore, traditional energy-saving control strategies often rely on manual experience, making it difficult to simultaneously achieve energy reduction, brightness smoothing, and dimming stability. The optimized algorithm also addresses the issue of... The technical problems of low efficiency, slow convergence speed, and getting trapped in local optima, resulting in limited energy-saving effect and unstable energy-saving effect of control strategies, are addressed in this invention. A collaborative optimization method for street light brightness within a cluster is proposed. An objective function for optimizing street light brightness within the cluster is constructed, and the optimization algorithm is improved based on a dynamic elite mutation position update strategy. This improves the convergence efficiency and stability of the optimization search, reduces the overall energy consumption of the street light cluster, reduces the brightness difference between adjacent street lights, suppresses brightness abrupt changes in adjacent control cycles, and improves dimming stability. This achieves efficient optimization allocation of target brightness within the street light cluster, as well as comprehensive optimization of energy saving, comfort, and dimming stability.

[0005] The technical solution adopted by this invention is as follows: The street light cluster collaborative energy-saving control method based on big data provided by this invention includes the following steps:

[0006] Step S1: Obtain standard data;

[0007] Step S2: Streetlight cluster division;

[0008] Step S3: Cluster lighting demand forecasting;

[0009] Step S4: Coordinated power-saving control of adjacent cluster streetlights;

[0010] Step S5: Collaborative power-saving control of streetlights within the cluster;

[0011] Step S6: Intelligent control of streetlights.

[0012] Further, in step S1, the acquisition of standard data specifically involves obtaining raw street light energy-saving control data through data acquisition operations, and performing data optimization processing on the raw street light energy-saving control data to obtain standardized street light energy-saving control data; the data optimization processing specifically involves sequentially performing data cleaning, standardization processing, and feature selection on the raw street light energy-saving control data to obtain standardized street light energy-saving control data.

[0013] The standardized data for street light energy-saving control includes target street light cluster division optimization data, historical lighting demand prediction optimization data, and real-time lighting demand prediction optimization data.

[0014] Furthermore, in step S2, the street light cluster division specifically includes the following steps:

[0015] Step S21: Multidimensional feature space gridding mapping processing, specifically, firstly, the value range of each feature dimension in the target street light cluster partitioning optimization data is statistically analyzed, the maximum and minimum values ​​of each feature dimension are calculated, and according to the preset single-dimensional grid partitioning quantity parameter s, the partitioning step size of each dimension is calculated. Each feature dimension is evenly divided into s intervals according to the partitioning step size to construct an m-dimensional discrete grid space. Finally, the grid index of each data sample in each dimension is calculated, and each data sample is mapped to the corresponding grid cell to obtain the grid cell belonging relationship of all data samples.

[0016] Step S22: Calculate the dual density of the grid cells. Specifically, first, count the number of street light data samples contained in each grid cell and use this as the first-level density value of the grid cell. For grid cells containing two or more data samples, calculate the ratio of its first-level density value to the sum of the Euclidean distances of all data samples in the grid cell and use this as the second-level density value of the grid cell. Finally, sort all grid cells, first in descending order of the first-level density value, and then in descending order of the second-level density value if the first-level density values ​​are the same, to obtain the density sorting sequence of the grid cells. The formula used is as follows:

[0017] ;

[0018] In the formula, This represents the first-level density value of the k-th grid cell. This represents the second-level density value of the k-th grid cell. and These represent the p-th and q-th data samples belonging to the k-th grid cell, respectively. This represents the Euclidean distance between the p-th data sample and the q-th data sample. This represents the k-th grid cell;

[0019] Step S23: Coarse-grained mesh block generation, specifically, based on the density sorting sequence of mesh cells, a multi-level diffusion neighbor search strategy is used to generate coarse-grained mesh blocks, including the following steps:

[0020] Step S231: Initialize the starting grid cell. Specifically, traverse the grid cell density sorting sequence, select the grid cell with the highest density that has not been marked as the starting grid cell of the current coarse-grained grid block, add the grid cell to the queue to be processed, and immediately mark it as processed.

[0021] Step S232: Multi-level diffusion search. Specifically, when the queue to be processed is not empty, the head grid cell is taken out and multi-level diffusion search is performed in all geometrically adjacent directions. The first-level neighbors and second-level neighbors are searched in turn until there are no new unmarked neighbors. For each searched neighbor grid cell, it is determined whether it contains data samples and is not marked. If it contains data samples and is not marked, the neighbor grid cell is added to the queue to be processed and immediately marked as processed. The above process is repeated until the queue to be processed is empty. All marked grid cells are aggregated into a coarse-grained grid block.

[0022] Step S233: Mesh cell traversal, specifically, repeatedly executing the initial mesh cell initialization and multi-level diffusion search until all mesh cells are divided into corresponding coarse-grained mesh blocks, obtaining the complete set of coarse-grained mesh blocks;

[0023] Step S24: Calculate the dual density of the coarse-grained grid block. Specifically, firstly, sum the first-level density values ​​and second-level density values ​​of all grid cells within the coarse-grained grid block to obtain the first-level overall density and second-level overall density of the coarse-grained grid block. Then, based on the first-level density of each grid cell and the first-level overall density of the coarse-grained grid block, calculate the density weighting coefficient of each grid cell within the coarse-grained grid block. Based on the density weighting coefficient, perform a weighted average calculation on the center coordinates of each grid cell within the coarse-grained grid block to obtain the weighted centroid coordinates of the coarse-grained grid block, which serves as the spatial representative point of the coarse-grained grid block.

[0024] Step S25: Cluster center selection. Specifically, based on the weighted centroid coordinates of each coarse-grained grid block, the Euclidean distance between any two coarse-grained grid blocks is calculated, and a coarse-grained grid block distance matrix is ​​constructed. Then, a dual density progressive comparison mechanism is introduced to calculate the relative distance between each coarse-grained grid block. Finally, based on the first-level overall density and relative distance of each coarse-grained grid block, the clustering decision value of each coarse-grained grid block is calculated, and the clustering decision values ​​of all coarse-grained grid blocks are sorted in descending order. The top c coarse-grained grid blocks with the highest clustering decision values ​​are selected as cluster centers, where c represents the number of cluster centers.

[0025] The dual density progressive comparison mechanism specifically compares the densities of any two coarse-grained grid blocks according to a priority-progressive rule. First, it compares the overall density of the first level of the two coarse-grained grid blocks. If there is a difference in the overall density of the first level, the density is directly determined. If the overall density of the first level is equal, then the overall density of the second level of the two coarse-grained grid blocks is compared. The formula used is as follows:

[0026] ;

[0027] ;

[0028] In the formula, This represents the relative distance of the nth coarse-grained grid block. This represents the v-th coarse-grained grid block. This represents the nth coarse-grained grid block. This represents the Euclidean distance between the nth and vth coarse-grained grid blocks. This represents the clustering decision value of the nth coarse-grained grid block. This represents the first-level global density of the v-th coarse-grained grid block. This represents the second-order global density of the v-th coarse-grained grid block. This represents the first-level global density of the nth coarse-grained grid block. This represents the second-order global density of the nth coarse-grained grid block;

[0029] Step S26: Full allocation of coarse-grained grid blocks. Specifically, firstly, the nearest neighbors of each coarse-grained grid block are calculated based on the dual density progressive comparison mechanism to establish the density adjacency relationship between coarse-grained grid blocks. Then, according to the order of the first-level overall density between coarse-grained grid blocks from high to low, all non-cluster center coarse-grained grid blocks are clustered and allocated in sequence. The current coarse-grained grid block is assigned to the cluster to which its nearest neighbor belongs. After all coarse-grained grid blocks have been clustered, the final clustering result is obtained.

[0030] Step S27: Real-time partitioning of street light clusters. Specifically, a street light partitioning clustering algorithm is constructed by first performing multi-dimensional feature space gridding mapping processing, calculating the dual density of grid cells, generating coarse-grained grid blocks, calculating the dual density of coarse-grained grid blocks, selecting cluster centers, and fully distributing coarse-grained grid blocks. The target street light cluster partitioning optimization data is then input into the street light partitioning clustering algorithm to perform real-time clustering of street lights and obtain the street light cluster partitioning results.

[0031] Further, in step S3, the cluster lighting demand prediction specifically involves first constructing a cluster lighting demand prediction model based on a long short-term memory neural network, and using historical lighting demand prediction optimization data as model training data to perform iterative training of the prediction model to obtain the trained cluster lighting demand prediction model. Then, based on the street light cluster division results, the real-time lighting demand prediction optimization data of each street light cluster is input into the trained cluster lighting demand prediction model to obtain the real-time lighting demand prediction results of each street light cluster. Finally, based on the real-time lighting demand prediction results, the initial basic brightness of each street light cluster is obtained.

[0032] Furthermore, in step S4, the coordinated power-saving control of adjacent cluster streetlights specifically includes the following steps:

[0033] Step S41: Calculate the basic brightness difference. Specifically, based on the street light cluster division results, determine the direct adjacency relationship between each street light cluster to obtain directly adjacent street light cluster pairs. Then, based on the initial basic brightness of each street light cluster, calculate the basic brightness difference between directly adjacent street light cluster pairs.

[0034] Step S42: Basic brightness compensation judgment, specifically, compare the basic brightness difference between directly adjacent street light clusters with a preset basic brightness difference threshold. If the basic brightness difference is greater than the preset basic brightness difference threshold, it is determined that the corresponding directly adjacent street light clusters need to be compensated for basic brightness. The street light cluster with high initial basic brightness is determined as the reference street light cluster, and the street light cluster with low initial basic brightness is determined as the street light cluster to be compensated.

[0035] Step S43: Basic brightness compensation. Specifically, when a pair of directly adjacent street light clusters needs basic brightness compensation, the difference between the basic brightness of the reference street light cluster and the preset basic brightness difference threshold is used as the coordinated basic brightness of the street light cluster to be compensated. Finally, the coordinated basic brightness of the street light cluster to be compensated and the initial basic brightness of the uncompensated street light cluster are combined to obtain the basic brightness coordination result of the adjacent clusters.

[0036] Furthermore, in step S5, the coordinated power-saving control of streetlights within the cluster specifically includes the following steps:

[0037] Step S51: Obtain the final base brightness of the street light cluster. Specifically, based on the coordination results of the base brightness of adjacent clusters, obtain the base brightness of each street light cluster and use it as the final base brightness of each street light cluster.

[0038] Step S52: Obtain the street light brightness adjustment range. Specifically, based on the street light location attributes, determine the location attributes of each street light in the same street light cluster. Based on the final base brightness of the street light cluster and the location attributes of each street light, determine the allowable brightness adjustment range of each street light in the street light cluster relative to the final base brightness, and obtain the brightness adjustment range of each street light in the street light cluster.

[0039] Step S53: Construct the objective function for optimizing the brightness of streetlights within the cluster. Specifically, the objective function for optimizing the brightness of streetlights within the same streetlight cluster is constructed by taking the target brightness of each streetlight in the same streetlight cluster as the optimization variable and including the energy consumption optimization term, the brightness smoothing term, and the dimming stability term.

[0040] Step S54: Optimize the target brightness of streetlights within the cluster, specifically including the following steps:

[0041] Step S541: Initialize candidate target brightness combinations. Specifically, the target brightness combination of each street light in the street light cluster is encoded into the search individual position vector in the optimization algorithm. Based on the brightness adjustment range of each street light in the street light cluster, P search individual position vectors are randomly generated to initialize the candidate target brightness combinations and obtain the initial search population.

[0042] Step S542: Calculate the fitness value of the search individual, specifically by calculating the fitness value of each search individual based on the objective function of optimizing the brightness of streetlights within the cluster;

[0043] Step S543: Dynamic elite mutation position update, specifically, at the beginning of each search iteration, first calculate the elite proportion coefficient under the current iteration. Based on this elite ratio coefficient, all individuals in the search population are sorted from best to worst according to their fitness values, and the top-ranked individuals are selected. The search individuals are grouped into an elite set. Then, the adaptive search step size and the average position of the population are calculated. Finally, based on the differential mutation mechanism of elite individuals, the average position of the population, and random individuals in the population, the mutation position of each search individual is updated. A greedy selection mechanism is used to compare the fitness values ​​of the search individuals before and after the update, retaining the search individuals with better fitness values. The formulas used are as follows:

[0044] ;

[0045] ;

[0046] ;

[0047] In the formula, t represents the current iteration number. Indicates the maximum number of iterations. This represents the elite proportion coefficient in the t-th iteration. This represents the adaptive search step size in the t-th iteration. Indicates the initial search step size. Indicates the average position of the population. Indicates the first The initial mutation position of the p-th search individual in the next iteration. and These represent the weighting coefficients for elite individuals and the population average position, respectively. This refers to a search individual randomly selected from a set of elite individuals. , , and This represents the positions of four distinct individuals randomly selected from the search population.

[0048] Step S544: Update the behavior position of the search individual. Specifically, for each search individual in the population, based on the dual conditions of random probability and iterative parity, select one of the following strategies to execute: camouflage search strategy, blood-spraying defense search strategy, and escape movement search strategy, to complete the behavior position update of the search individual.

[0049] Step S545: Search iteration terminates. Specifically, after each iteration, the fitness values ​​of all search individuals in the current population are calculated. When the fitness value of a search individual is higher than the fitness threshold or the maximum number of iterations is reached, the search is terminated and the position of the globally optimal search individual is obtained. The globally optimal search individual position specifically refers to the optimal combination of target brightness of streetlights within the streetlight cluster.

[0050] Step S55: Obtain the street light control strategy within the cluster. Specifically, based on the optimal combination of target brightness of street lights within the street light cluster, extract the target brightness corresponding to each street light in the street light cluster, and match the target brightness of each street light with its respective street light cluster number and street light number to obtain the collaborative control strategy for street lights within the cluster.

[0051] Furthermore, in step S6, the intelligent control of streetlights specifically involves obtaining the intra-cluster streetlight collaborative control strategy of each streetlight cluster based on the basic brightness coordination results of adjacent clusters and through intra-cluster streetlight collaborative power saving control, generating dimming control instructions for streetlights in each streetlight cluster based on the intra-cluster streetlight collaborative control strategy, and causing each streetlight to dim according to the corresponding target brightness according to the dimming control instructions, thereby realizing intelligent control of streetlight cluster collaborative power saving.

[0052] The technical solution adopted by the present invention is as follows: The street light cluster collaborative energy-saving control system based on big data provided by the present invention includes a standard data acquisition module, a street light cluster division module, a cluster lighting demand prediction module, an adjacent cluster street light collaborative energy-saving control module, an intra-cluster street light collaborative energy-saving control module, and a street light intelligent control module.

[0053] The standard data acquisition module specifically obtains standardized street light energy-saving control data through data acquisition and data optimization, and sends the data to the street light cluster division module, the cluster lighting demand prediction module, and the cluster-based street light collaborative energy-saving control module.

[0054] The street light cluster partitioning module receives data sent by the standard data acquisition module, improves the clustering algorithm through coarse-grained grid block generation, dual density calculation, and dual density progressive comparison mechanism using a multi-level diffused neighbor search strategy, and constructs a street light partitioning clustering algorithm. The target street light cluster partitioning optimization data is input into the street light partitioning clustering algorithm to perform real-time street light cluster partitioning, obtain the street light cluster partitioning results, and send the data to the cluster lighting demand prediction module and the adjacent cluster street light collaborative power saving control module.

[0055] The cluster lighting demand prediction module receives data sent by the standard data acquisition module and the street light cluster division module. By constructing and training the cluster lighting demand prediction model, and based on the street light cluster division results and the trained cluster lighting demand prediction model, the initial basic brightness of each street light cluster is finally obtained, and the data is sent to the adjacent cluster street light collaborative energy saving control module.

[0056] The adjacent cluster street light collaborative energy-saving control module receives data sent by the street light cluster division module and the cluster lighting demand prediction module. Based on the street light cluster division result and the initial base brightness of each street light cluster, it calculates the base brightness difference between two directly adjacent street light clusters, and performs base brightness compensation based on the base brightness difference to obtain the base brightness of the street light cluster to be compensated after coordination. Finally, it combines the base brightness corresponding to each street light cluster to obtain the basic brightness coordination result of the adjacent clusters, and sends the data to the street light collaborative energy-saving control module within the cluster.

[0057] The cluster-based street light collaborative energy-saving control module receives data sent by the standard data acquisition module and the adjacent cluster street light collaborative energy-saving control module. Based on the basic brightness coordination results of adjacent clusters, it obtains the final basic brightness of each street light cluster. According to the position attributes of each street light in the same street light cluster, it obtains the brightness adjustment range of each street light in the street light cluster and constructs a cluster-based street light brightness optimization objective function including energy consumption optimization, brightness smoothing, and dimming stability. Based on the dynamic elite mutation position update strategy, the optimization algorithm is improved. The improved optimization algorithm is used to optimize the target brightness of the street lights in the cluster within the street light brightness adjustment range, obtains the optimal combination of target brightness of the street lights in the street light cluster, and finally obtains the cluster-based street light collaborative control strategy based on this combination and sends the data to the street light intelligent control module.

[0058] The street light intelligent control module receives data sent by the street light collaborative energy-saving control module within the cluster, and performs street light dimming control for each street light cluster through the cluster-based street light collaborative control strategy, thereby realizing intelligent collaborative energy-saving control of the street light cluster.

[0059] The beneficial effects achieved by adopting the above solution are as follows:

[0060] (1) Traditional street light energy-saving control methods typically use single lights, fixed road sections, or fixed circuits as control objects, and are fixedly divided according to administrative regions, road numbers, or control box circuits. At the same time, they mainly rely on fixed time periods, ambient light thresholds, or traffic flow thresholds for dimming control, resulting in dispersed street light control objects, coarse area division, inaccurate matching of lighting demand, delayed dimming response, insufficient coordination between adjacent areas, and high energy consumption of street light operation. This solution creatively adopts a clustering algorithm to divide street lights into clusters, grouping street lights with similar lighting needs and related spatial locations and road topologies into the same street light cluster, improving the rationality of street light grouping and enhancing the overall control of the cluster. Then, cluster lighting demand is predicted to predict the future preset time window for each street light cluster. The system determines the lighting demand level within a street light cluster and establishes an initial base brightness accordingly. Then, by coordinating the base brightness of adjacent clusters and optimizing differentiated target brightness within each cluster, it implements collaborative energy-saving control for each street light cluster and individual street lights within each cluster. This improves the advance capability of dimming control, enhances the continuity of brightness between adjacent clusters, and improves the lighting safety of key areas. Ultimately, it reduces the energy consumption of street light operation, transforming street light energy-saving control from decentralized control of individual lights, fixed area control, and passive threshold control to dynamic cluster division, proactive demand prediction, and collaborative optimization of dimming and energy-saving control. While ensuring road lighting safety, brightness continuity, and dimming stability, it improves the refinement, intelligence, and collaboration of urban street light energy-saving control, reduces street light operating energy consumption and management costs, and achieves collaborative energy-saving intelligent control of large-scale street light networks.

[0061] (2) To address the technical problems of existing clustering methods for street light cluster partitioning, which use individual data samples as clustering units, rely on cutoff distance parameters, and employ a single density metric, resulting in high computational complexity, strong parameter sensitivity, poor generalization ability, and difficulty in applying to large-scale street light datasets, and are prone to misidentification of cluster centers and chain errors in cluster assignment, thus causing inaccurate street light cluster grouping and unstable cluster control, this solution innovatively improves the clustering algorithm by using a multi-level diffused neighbor search strategy to generate coarse-grained grid blocks, calculate dual densities, and employ a dual density progressive comparison mechanism. This compresses the clustering objects from the original data samples to... The number of grid blocks, far smaller than the number of samples, significantly reduces the scale of clustering computation, eliminates the dependence of the clustering process on the cutoff distance parameter, enhances the adaptability of the clustering algorithm to street light data of different sizes and distributions, improves the accuracy of cluster center identification through a dual density progressive comparison mechanism, reduces the chain propagation of clustering assignment errors caused by misidentification of centers or unreasonable allocation order, improves the accuracy of clustering results, enhances the accuracy, stability and reliability of street light cluster partitioning results, realizes efficient, accurate and adaptive partitioning of street light clusters in large-scale street light networks, thereby enhancing the overall cluster control and improving the accuracy of subsequent lighting demand prediction and energy-saving control.

[0062] (3) In view of the fact that traditional energy-saving control strategies rely heavily on manual experience and are difficult to simultaneously achieve energy consumption reduction, brightness smoothing and dimming stability, and that optimization algorithms are prone to low search efficiency, slow convergence speed and getting stuck in local optima during the search process of complex street light brightness combinations, resulting in limited energy-saving effect of control strategies, frequent dimming of street lights, obvious brightness fluctuations, unreasonable target brightness combinations and unstable energy-saving effect, this invention proposes a cluster-based street light brightness collaborative optimization method. It constructs a cluster-based street light brightness optimization objective function including energy consumption optimization term, brightness smoothing term and dimming stability term, and transforms the street light target brightness optimization into a method that takes into account energy saving, brightness continuity and dimming stability. This paper addresses a qualitative multi-objective collaborative optimization problem and improves the optimization algorithm based on a dynamic elite mutation position update strategy. During the iteration process, information from elite individuals, the population average position, and random difference mutation is integrated to dynamically update the position of the search individuals. This enhances the global search capability and local exploitation capability of the optimization algorithm, improves the convergence efficiency and stability of the optimization search, obtains the optimal combination of target brightness for each street light within the street light cluster, reduces the overall energy consumption of the street light cluster, minimizes brightness differences between adjacent street lights, suppresses brightness abrupt changes in adjacent control cycles, and improves dimming stability. This achieves efficient optimization of target brightness allocation within the street light cluster, as well as comprehensive optimization of energy saving, comfort, and dimming stability. Attached Figure Description

[0063] Figure 1 A flowchart illustrating the big data-based collaborative power-saving control method for streetlight clusters provided by this invention;

[0064] Figure 2 A schematic diagram of the modules of the big data-based collaborative energy-saving control system for streetlight clusters provided by the present invention;

[0065] Figure 3 A flowchart illustrating the process of dividing the street light cluster in step S2;

[0066] Figure 4 This is a flowchart illustrating the process of coordinated power-saving control of adjacent streetlight clusters in step S4.

[0067] Figure 5 This is a flowchart illustrating the process of coordinated power-saving control of streetlights within the cluster in step S5.

[0068] Figure 6 A flowchart illustrating the process of optimizing the target brightness of streetlights within the cluster in step S54;

[0069] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0070] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0071] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0072] Example 1, see Figure 1 The present invention provides a big data-based collaborative power-saving control method for street light clusters, which includes the following steps:

[0073] Step S1: Obtain standard data, specifically by acquiring and optimizing data to obtain standardized data for street light energy-saving control;

[0074] Step S2: Streetlight cluster partitioning, used to group streetlights with similar lighting needs and spatial and road topological relationships into the same streetlight cluster; specifically, it improves the clustering algorithm by generating coarse-grained grid blocks through a multi-level diffused neighbor search strategy, calculating dual density, and using a dual density progressive comparison mechanism, thereby constructing a streetlight partitioning clustering algorithm, and inputting the target streetlight cluster partitioning optimization data into the streetlight partitioning clustering algorithm to perform real-time streetlight cluster partitioning and obtain the streetlight cluster partitioning results;

[0075] Step S3: Cluster lighting demand prediction, used to predict the lighting demand level of each street light cluster within a future preset time window; specifically, to build and train a cluster lighting demand prediction model, based on the street light cluster division results and the trained cluster lighting demand prediction model, and finally obtain the initial base brightness of each street light cluster.

[0076] Step S4: Collaborative power-saving control of adjacent street light clusters, used to coordinate the base brightness between directly adjacent street light clusters; specifically, based on the street light cluster division results and the initial base brightness of each street light cluster, the base brightness difference between two directly adjacent street light clusters is calculated, and base brightness compensation is performed according to the base brightness difference to obtain the base brightness of the street light cluster to be compensated after coordination. Finally, the base brightness corresponding to each street light cluster is combined to obtain the coordination result of the base brightness of adjacent clusters.

[0077] Step S5: Collaborative energy-saving control of streetlights within the same cluster, used to perform differentiated optimization control of the target brightness of each streetlight within the same cluster; specifically, based on the coordination results of the basic brightness of adjacent clusters, the final basic brightness of each streetlight cluster is obtained; according to the position attributes of each streetlight within the same cluster, the brightness adjustment range of each streetlight within the cluster is obtained; and a cluster-wide streetlight brightness optimization objective function including energy consumption optimization, brightness smoothing, and dimming stability is constructed; the optimization algorithm is improved based on a dynamic elite mutation position update strategy; the improved optimization algorithm is used to optimize the target brightness of streetlights within the cluster within the streetlight brightness adjustment range, obtaining the optimal combination of target brightness of streetlights within the cluster; finally, the collaborative control strategy of streetlights within the cluster is obtained based on this combination.

[0078] Step S6: Intelligent control of streetlights. Through the intra-cluster collaborative control strategy of each streetlight cluster, the dimming control of each streetlight cluster is carried out to achieve intelligent control of streetlight cluster collaborative energy saving.

[0079] By performing the above operations, this solution addresses the technical problems of traditional street light energy-saving control methods, which typically control individual lights, fixed road sections, or fixed circuits, and are fixedly divided according to administrative regions, road numbers, or control box circuits. Furthermore, these methods primarily rely on fixed time periods, ambient light thresholds, or traffic flow thresholds for dimming control, resulting in dispersed street light control objects, coarse area division, inaccurate matching of lighting demand, delayed dimming response, insufficient coordination between adjacent areas, and high energy consumption during street light operation. This solution creatively employs a clustering algorithm to divide street lights into clusters, grouping street lights with similar lighting needs and spatial and road topological relationships into the same cluster. This improves the rationality of street light grouping and enhances the overall control of the cluster. Then, cluster lighting demand is predicted to forecast the future preset lighting needs of each street light cluster. The system determines the lighting demand level within a time window and establishes an initial base brightness accordingly. Then, by coordinating the base brightness of adjacent clusters and optimizing differentiated target brightness within clusters, it implements collaborative energy-saving control for each street light cluster and individual street lights within each cluster. This improves the advance capability of dimming control, enhances the brightness continuity of adjacent clusters, and improves lighting safety in key areas. Ultimately, it reduces street light operating energy consumption, transforming street light energy-saving control from decentralized single-lamp control, fixed-area control, and passive threshold control to dynamic cluster partitioning, proactive demand prediction, and collaborative optimization of dimming and energy-saving control. While ensuring road lighting safety, brightness continuity, and dimming stability, it improves the refinement, intelligence, and collaboration of urban street light energy-saving control, reduces street light operating energy consumption and management costs, and achieves collaborative energy-saving intelligent control of large-scale street light networks.

[0080] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S1, the acquisition of standard data is used to provide a unified data foundation for street light cluster division, cluster lighting demand prediction, collaborative energy-saving control of adjacent cluster street lights, and collaborative energy-saving control of street lights within a cluster. Specifically, it involves obtaining raw street light energy-saving control data through data acquisition operations from the street light operation monitoring system, road traffic detection system, environmental monitoring system, urban road network information system, and street light intelligent control platform, and then performing data optimization processing on the raw street light energy-saving control data to obtain standardized street light energy-saving control data.

[0081] The data optimization process is used to preprocess and optimize the original data of street light energy-saving control to eliminate missing values, outliers, dimensional differences, and low-association redundant features in the original data, thereby obtaining standardized street light energy-saving control data suitable for street light cluster division and cluster lighting demand prediction; specifically, the original data of street light energy-saving control is sequentially cleaned, standardized, and feature-selected to obtain standardized street light energy-saving control data.

[0082] The standardized data for street light energy-saving control includes target street light cluster division optimization data, historical lighting demand prediction optimization data, and real-time lighting demand prediction optimization data.

[0083] The raw data for street light energy-saving control includes target street light cluster division data, historical lighting demand prediction data, and real-time lighting demand prediction data.

[0084] The target street light cluster partitioning data is used to partition street lights into clusters; it includes street light equipment data, street light spatial location data, street light attribute data, and street light pedestrian and vehicle traffic data.

[0085] The street light equipment data includes street light number, lamp type, rated power, energy consumption value, pole height, installation spacing, illumination direction, and dimming method;

[0086] The spatial location data includes the street light geographic coordinates, distances between adjacent street lights, road connectivity, and road topology.

[0087] The street light attribute data includes the road data to which the street light belongs, the street light area function, and the street light location attributes;

[0088] The road data to which the streetlights belong includes road grade, road type, number of lanes, road width, and road speed limit;

[0089] The streetlight areas can be categorized into commercial areas, residential areas, areas surrounding schools and hospitals, bus station areas, industrial areas, and low-activity areas.

[0090] The street light location attributes are used to determine the allowable brightness adjustment range of different street lights relative to the final base brightness; they are divided into intersection lights, zebra crossing lights, bus stop lights, main road lights, ordinary road section lights, and branch street lights.

[0091] The street light pedestrian and vehicle traffic data includes average vehicle traffic volume for each time period, average pedestrian traffic volume for each time period, traffic flow variation patterns, pedestrian flow variation patterns, peak hours, and low traffic volume periods.

[0092] The historical lighting demand forecast data and real-time lighting demand forecast data are used to predict the lighting demand level of each street light cluster within a future preset time window, including traffic flow data of the street light cluster, environmental status data of the street light cluster, time attribute data and regional attribute data of the street light cluster.

[0093] The traffic flow data of the street light cluster includes vehicle volume, pedestrian volume, average vehicle speed, vehicle density, and road congestion status; the environmental status data of the street light cluster includes ambient light intensity, weather conditions, rainfall, haze level, visibility, temperature, and humidity; the time attribute data includes current time, date, weekday or holiday, season, and nighttime.

[0094] The street light cluster area attribute data includes road level, road type, area function type, intersection area, zebra crossing area, bus stop area, school surrounding area, hospital surrounding area, commercial area, residential area, and low activity area;

[0095] The historical lighting demand forecast data also includes historical cluster lighting demand forecast results; the cluster lighting demand forecast results can be divided into D1 low lighting demand level, D2 low-to-medium lighting demand level, D3 medium lighting demand level and D4 high lighting demand level, wherein different lighting demand levels correspond to different basic brightness of streetlights within the cluster.

[0096] The correspondence between the lighting demand level and the basic brightness of the streetlights is as follows: D1, low lighting demand level, basic brightness is 20%; D2, low-to-medium lighting demand level, basic brightness is 40%; D3, medium lighting demand level, basic brightness is 60%; and D4, high lighting demand level, basic brightness is 80%. The basic brightness is expressed as a percentage, representing the ratio of the streetlight's actual output brightness to its rated maximum output brightness. When the basic brightness is 100%, the streetlight operates at its rated maximum brightness; when the basic brightness is between 20% and 80%, it indicates that the streetlight operates at a corresponding proportion of its rated maximum brightness. Expressing the basic brightness as a percentage eliminates differences in rated power, lamp model, and dimming method among different streetlights, allowing for basic brightness setting and coordinated energy-saving control of different streetlights under a unified brightness ratio scale. The basic brightness is used to indicate the current control level of the corresponding streetlight cluster. The brightness benchmark value meets basic safety lighting requirements while also considering energy-saving needs within the cycle; the data cleaning is used to process missing, abnormal, and duplicate data in the original data of street light energy-saving control to improve data integrity and accuracy; specifically, it involves imputing missing values, removing outliers, and deleting duplicate data in the original data; the missing value imputation specifically involves using different methods to imput missing fields in the original data according to the field type, using the mean imputation method for continuous fields and the mode imputation method for categorical fields; the outlier removal specifically involves identifying data that exceeds the preset reasonable value range or deviates from the statistical distribution characteristics based on the preset reasonable value range of each field, and removing the identified outlier data; the duplicate data deletion specifically involves identifying duplicate data based on the street light number, data collection time, data source, and data field content, and retaining one valid data entry.

[0097] The standardization process is used to convert data from different sources, types, and scales into a unified data representation format for subsequent clustering and lighting demand prediction. Specifically, it first uses label encoding technology to map the categorical fields in the original data to unique integer values ​​according to their different categories, thus achieving the quantitative conversion of categorical data. Then, it uses the Z-Score standardization algorithm to standardize all continuous variables, converting each continuous variable into a standardized variable with a mean of 0 and a standard deviation of 1, thereby achieving a unified numerical scale for each variable.

[0098] The feature selection is used to filter out core input features that contribute significantly to street light cluster division and cluster lighting demand prediction, and to remove low-association redundant features in order to reduce computational complexity and improve the efficiency and accuracy of subsequent model processing; specifically, it includes feature selection for street light cluster division and feature selection for lighting demand prediction.

[0099] The street light cluster segmentation feature selection specifically involves calculating the similarity contribution of each candidate input feature to the spatial adjacency of street lights, road topology, road attribute consistency, regional functional consistency, and pedestrian / vehicle flow similarity. A preset threshold for cluster feature contribution is set, and features with a contribution greater than the preset threshold are selected as core input features for target street light cluster segmentation. Low-association redundant features with a contribution less than or equal to the preset threshold are removed. The lighting demand prediction feature selection specifically involves calculating the correlation between each input feature and historical cluster lighting demand prediction results. A preset threshold for lighting demand prediction features is set, and features with an absolute correlation coefficient greater than the preset threshold are selected as core input features for lighting demand prediction. Low-association redundant features with an absolute correlation coefficient less than or equal to the preset threshold are removed.

[0100] Example 3, see Figure 1 , Figure 2 and Figure 3This embodiment is based on the above embodiment. In step S2, the street light cluster division is used to group street lights with similar lighting needs and spatial and road topological relationships into the same street light cluster. This allows subsequent control to no longer focus on individual street lights or fixed road sections, but rather on street light clusters with similar lighting needs. This improves the overall and coordinated nature of energy-saving control, avoids the control dispersion problem caused by independent control of individual lights, and also avoids the inaccurate control problem caused by grouping by fixed administrative regions or fixed road sections. It realizes the transformation of street lights from single-point control to cluster collaborative control. Specifically, it improves the clustering algorithm through coarse-grained grid block generation using a multi-level diffused neighbor search strategy, dual-density calculation, and dual-density progressive comparison mechanism, thereby constructing a street light division clustering algorithm. The target street light cluster division optimization data is input into the street light division clustering algorithm to perform real-time street light cluster division and obtain the street light cluster division result. The steps include:

[0101] Step S21: Multidimensional feature space gridding mapping processing is used to discretize the original continuous feature space into a regular grid structure, realize the first level of data reduction, eliminate the computational overhead of sample-level distance matrix in traditional clustering algorithms, and transform the continuous feature space into an indexable discrete grid space. This lays the foundation for subsequent fast clustering operations from the data structure level and significantly reduces the computational complexity of clustering in large-scale street light networks. Specifically, firstly, the value range of each feature dimension in the target street light cluster partitioning optimization data is statistically analyzed, the maximum and minimum values ​​of each feature dimension are calculated, and the partitioning step size of each dimension is calculated according to the preset single-dimensional grid partitioning quantity parameter s. Each feature dimension is evenly divided into s intervals according to the partitioning step size to construct an m-dimensional discrete grid space. Finally, the grid index of each data sample in each dimension is calculated, and each data sample is mapped to the corresponding grid cell to obtain the grid cell affiliation relationship of all data samples.

[0102] Specifically, the data samples are optimized by dividing the target street light cluster, treating each street light as a data sample unit, and constructing a multi-dimensional data sample feature vector that includes street light equipment data, street light spatial location data, street light attribute data, and street light pedestrian and vehicle traffic data; the formula used is as follows:

[0103] ;

[0104] ;

[0105] In the formula, This represents the value of the i-th data sample in the j-th dimension feature. This represents the minimum value of the j-th feature across all data samples. This represents the maximum value of the j-th dimension feature across all data samples. Let represent the step size of the j-th feature, and s represent the number of grid divisions for each feature dimension. This represents the floor function. This represents the i-th data sample. This represents the grid cell index to which the i-th data sample belongs, used to uniquely identify the grid cell to which the data sample belongs. , and Let represent the values ​​of the i-th data sample in the 1st, 2nd, and m-th dimensions of the feature, respectively. , and Let these represent the minimum values ​​of the 1st, 2nd, and mth features, respectively, across all data samples. , and These represent the maximum values ​​of the 1st, 2nd, and mth dimensions of the feature across all data samples. , and These represent the partitioning step sizes for the 1st, 2nd, and mth features, respectively, where m represents the total number of feature dimensions.

[0106] Step S22: Calculate the dual density of the grid cells to quantify the density and internal cohesion of the data distribution in each grid cell. A completely parameter-free density calculation method is designed to eliminate the dependence of traditional density peak clustering on the cutoff distance parameter. The primary density ensures that densely populated areas of streetlights are prioritized, conforming to the distribution characteristics of urban road streetlights. The secondary density characterizes the homogeneity of streetlight lighting needs within the grid cell. A higher secondary density indicates a more consistent lighting pattern among the streetlights within the grid cell, making it more suitable as a single control unit. The dual density mechanism avoids neglecting illumination while clustering solely based on spatial distance. To address the issue of varying demand, the process involves first counting the number of streetlight data samples within each grid cell, using this as the first-level density value for that grid cell. For grid cells containing two or more data samples, the ratio of their first-level density value to the sum of the Euclidean distances between all data samples within that grid cell is calculated, and this ratio is used as the second-level density value for that grid cell. Finally, all grid cells are sorted, first in descending order of their first-level density values, and then in descending order of their second-level density values ​​if the first-level density values ​​are the same, resulting in a density sorting sequence for the grid cells. The formula used is as follows:

[0107] ;

[0108] In the formula, This represents the first-level density value of the k-th grid cell, i.e., the number of data samples contained in the k-th grid cell. This represents the second-level density value of the k-th grid cell. and These represent the p-th and q-th data samples belonging to the k-th grid cell, respectively. This represents the Euclidean distance between the p-th data sample and the q-th data sample. This represents the k-th grid cell;

[0109] Step S23: Coarse-grained grid block generation, used to aggregate adjacent and connected grid cells into coarse-grained grid blocks, achieving second-level data reduction, significantly reducing the number of objects in subsequent clustering, and fundamentally reducing computational complexity; specifically, based on the density sorting sequence of grid cells, a multi-level diffused neighbor search strategy is used to generate coarse-grained grid blocks, including the following steps:

[0110] Step S231: Initialize the starting grid cell. Specifically, traverse the grid cell density sorting sequence, select the grid cell with the highest density that has not been marked as the starting grid cell of the current coarse-grained grid block, add the grid cell to the queue to be processed, and immediately mark it as processed.

[0111] Step S232: Multi-level diffusion search. Specifically, when the queue to be processed is not empty, the head grid cell is taken out and multi-level diffusion search is performed in all geometrically adjacent directions. The first-level neighbors and second-level neighbors are searched in turn until there are no new unmarked neighbors. For each searched neighbor grid cell, it is determined whether it contains data samples and is not marked. If it contains data samples and is not marked, the neighbor grid cell is added to the queue to be processed and immediately marked as processed. The above process is repeated until the queue to be processed is empty. All marked grid cells are aggregated into a coarse-grained grid block.

[0112] Step S233: Mesh cell traversal, specifically, repeatedly executing the initial mesh cell initialization and multi-level diffusion search until all mesh cells are divided into corresponding coarse-grained mesh blocks, obtaining the complete set of coarse-grained mesh blocks;

[0113] Step S24: Calculate the dual density of the coarse-grained grid block to quantify the overall cluster density of each coarse-grained grid block and determine its spatial representative point. The overall density of the coarse-grained grid block is obtained by aggregating the density of grid cells. The density contribution difference of different grid cells within the coarse-grained grid block is fully considered. The centroid is calculated using a density weighting method, which makes the spatial representative point position more accurate and provides an accurate benchmark for the distance calculation of density peak clustering and the selection of cluster centers. Specifically, firstly, the first-level density value and the second-level density value of all grid cells in the coarse-grained grid block are summed to obtain the first-level overall density and the second-level overall density of the coarse-grained grid block. Then, based on the first-level density of each grid cell and the first-level overall density of the coarse-grained grid block, the density weight coefficient of each grid cell in the coarse-grained grid block is calculated. Based on the density weight coefficient, the center coordinates of each grid cell in the coarse-grained grid block are weighted and averaged to obtain the weighted centroid coordinates of the coarse-grained grid block, which serve as the spatial representative point of the coarse-grained grid block.

[0114] The center coordinates of the grid cell are specifically the average coordinates of all data samples within that grid cell in the multidimensional feature space, i.e., the centroid of the sample in that grid cell. This centroid is calculated by averaging the feature values ​​of each dimension of all data samples contained within the grid cell; the formula used is as follows:

[0115] ;

[0116] ;

[0117] ;

[0118] ;

[0119] In the formula, This represents the first-level global density of the nth coarse-grained grid block. This represents the second-order global density of the nth coarse-grained grid block. This represents the nth coarse-grained grid block. This represents the density weight coefficient of the k-th grid cell in the corresponding coarse-grained grid block. This represents the weighted centroid coordinates of the nth coarse-grained grid block. Indicates the center coordinates of the k-th grid cell;

[0120] Step S25: Cluster center selection. This step automatically identifies the optimal cluster centers for density peak clusters based on the density and spatial distance characteristics of coarse-grained grid blocks. It eliminates the need to pre-specify the number of clusters, avoiding misidentification of cluster centers caused by judging solely by a single density value. Cluster centers can be automatically determined based on data distribution. Specifically, based on the weighted centroid coordinates of each coarse-grained grid block, the Euclidean distance between any two coarse-grained grid blocks is calculated, constructing a coarse-grained grid block distance matrix. Then, a dual density progressive comparison mechanism is introduced to calculate the relative distance between each coarse-grained grid block. Finally, based on the first-level overall density and relative distance of each coarse-grained grid block, the clustering decision value of each coarse-grained grid block is calculated. All clustering decision values ​​are then sorted in descending order, and the top c coarse-grained grid blocks with the highest clustering decision values ​​are selected as cluster centers, where c represents the number of cluster centers.

[0121] The dual density progressive comparison mechanism specifically compares the density of any two coarse-grained grid blocks according to the priority progressive rule. First, it compares the first-level overall density of the two coarse-grained grid blocks. If there is a difference in the first-level overall density, the density is directly determined. If the first-level overall density is equal, the second-level overall density of the two coarse-grained grid blocks is compared.

[0122] The formula used is as follows:

[0123] ;

[0124] ;

[0125] ;

[0126] In the formula, This represents the relative distance of the nth coarse-grained grid block. This represents the v-th coarse-grained grid block. This represents the weighted centroid coordinates of the v-th coarse-grained grid block. This represents the function for calculating Euclidean distance. This represents the Euclidean distance between the nth and vth coarse-grained grid blocks. This represents the clustering decision value of the nth coarse-grained grid block. This represents the first-level global density of the v-th coarse-grained grid block. This represents the second-order global density of the v-th coarse-grained grid block;

[0127] Step S26: Full allocation of coarse-grained grid blocks, used to complete the clustering of all coarse-grained grid blocks, realizing complete cluster allocation from the cluster center to all grid blocks, ensuring the accuracy of clustering allocation; according to the chain allocation strategy from high to low density, the result deviation caused by the traditional allocation order is avoided; specifically, firstly, the nearest neighbor of each coarse-grained grid block is calculated based on the dual density progressive comparison mechanism to establish the density adjacency relationship between coarse-grained grid blocks, and then, according to the order of the first-level overall density between coarse-grained grid blocks from high to low, all non-cluster center coarse-grained grid blocks are clustered and allocated in sequence, and the current coarse-grained grid block is assigned to the cluster of its nearest neighbor. After all coarse-grained grid blocks have been clustered, the final clustering result is obtained.

[0128] The clustering result is specifically a set of c clusters, each cluster corresponding to a group of streetlights that are spatially close and have similar lighting requirements, forming an independent streetlight control cluster. Different clusters represent streetlight clusters with different lighting requirements, realizing adaptive cluster partitioning of the urban streetlight network; the formula used is as follows:

[0129] ;

[0130] In the formula, This represents the nearest neighbor of the nth coarse-grained grid block;

[0131] Step S27: Real-time street light cluster partitioning. Specifically, the street light partitioning clustering algorithm is first constructed by the multi-dimensional feature space gridding mapping process, calculating the dual density of grid cells, generating coarse-grained grid blocks, calculating the dual density of coarse-grained grid blocks, selecting cluster centers, and fully distributing coarse-grained grid blocks. The target street light cluster partitioning optimization data is then input into the street light partitioning clustering algorithm to perform real-time clustering partitioning of the street lights and obtain the street light cluster partitioning results.

[0132] The street light cluster division result includes the street light cluster number and the street light numbers contained in each street light cluster.

[0133] By performing the above operations, this solution addresses the technical problems of existing clustering methods for street light cluster partitioning. These methods use individual data samples as clustering units, rely on truncation distance parameters, and employ a single density metric, resulting in high computational complexity, strong parameter sensitivity, poor generalization ability, and difficulty in applying to large-scale street light datasets. Furthermore, they are prone to misidentification of cluster centers and cascading errors in cluster assignment, leading to inaccurate street light cluster grouping and unstable cluster control. This innovative approach improves the clustering algorithm by using a multi-level diffused neighbor search strategy for coarse-grained grid block generation, dual density calculation, and a dual density progressive comparison mechanism. This improves the clustering algorithm by transforming the clustering objects from the original data samples... By compressing the data into grid blocks far smaller than the number of samples, the scale of clustering computation is significantly reduced, eliminating the dependence of the clustering process on the cutoff distance parameter. This enhances the adaptability of the clustering algorithm to street light data of different sizes and distributions. The dual-density progressive comparison mechanism improves the accuracy of cluster center identification and reduces the chain propagation of clustering assignment errors caused by misidentification of centers or unreasonable allocation order. This improves the accuracy of clustering results, enhances the accuracy, stability, and reliability of street light cluster partitioning results, and enables efficient, accurate, and adaptive partitioning of street light clusters in large-scale street light networks. Consequently, it enhances the overall control of the clusters and improves the accuracy of subsequent lighting demand prediction and energy-saving control.

[0134] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, the cluster lighting demand prediction is used to predict the lighting demand level of each street light cluster within a future preset time window, and to determine the initial basic brightness of the corresponding street light cluster based on the lighting demand level. Specifically, firstly, a cluster lighting demand prediction model is constructed based on a long short-term memory neural network, and historical lighting demand prediction optimization data is used as model training data to perform iterative training of the prediction model to obtain the trained cluster lighting demand prediction model. Then, based on the street light cluster division results, the real-time lighting demand prediction optimization data of each street light cluster is input into the trained cluster lighting demand prediction model to obtain the real-time lighting demand prediction results of each street light cluster. Finally, based on the real-time lighting demand prediction results, the initial basic brightness of each street light cluster is obtained.

[0135] The preset future time window can be 15 minutes in the future;

[0136] The iterative training of the prediction model specifically involves using the cross-entropy loss function and iteratively updating the weight matrix and bias parameters of the cluster lighting demand prediction model through backpropagation algorithm and gradient descent optimization method. The model parameters are continuously optimized through multiple rounds of iteration. Iterative training stops when the preset maximum number of training times is reached or the loss function value converges to a set threshold.

[0137] The cluster lighting demand prediction model uses a long short-term memory neural network to learn the historical traffic flow, environmental conditions, time attributes, and regional attributes of the street light clusters in a time series. This allows it to capture the changing patterns of cluster lighting demand at different times, thereby improving the accuracy of predicting lighting demand levels within a preset time window in the future.

[0138] Example 5, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S4, the adjacent street light cluster coordinated energy-saving control is used to coordinate the basic brightness between directly adjacent street light clusters, avoiding abrupt changes in overall brightness due to large differences in lighting demand levels between two adjacent street light clusters, and forming a smoother brightness transition between adjacent street light clusters. Specifically, based on the street light cluster division results and the initial basic brightness of each street light cluster, the basic brightness difference between two directly adjacent street light clusters is calculated, and basic brightness compensation is performed according to the basic brightness difference to obtain the basic brightness of the street light cluster to be compensated after coordination. Finally, the basic brightness corresponding to each street light cluster is combined to obtain the basic brightness coordination result of adjacent clusters. The steps include:

[0139] Step S41: Calculate the basic brightness difference. Specifically, based on the street light cluster division results, determine the direct adjacency relationship between each street light cluster to obtain directly adjacent street light cluster pairs. Then, based on the initial basic brightness of each street light cluster, calculate the basic brightness difference between directly adjacent street light cluster pairs.

[0140] The directly adjacent street light clusters represent two street light clusters that are spatially adjacent or that have a road topology connection between them;

[0141] Step S42: Basic brightness compensation judgment, specifically, comparing the basic brightness difference between directly adjacent street light clusters with a preset basic brightness difference threshold. If the basic brightness difference is greater than the preset basic brightness difference threshold, it is determined that the corresponding directly adjacent street light clusters need basic brightness compensation. The street light cluster with higher initial basic brightness is identified as the reference street light cluster, and the street light cluster with lower initial basic brightness is identified as the street light cluster to be compensated. If the basic brightness difference is less than or equal to the preset basic brightness difference threshold, it is determined that the corresponding directly adjacent street light clusters do not need basic brightness compensation, and the initial basic brightness of the two street light clusters remains unchanged. The preset basic brightness difference threshold is set to 25%.

[0142] Step S43: Basic brightness compensation. Specifically, when a pair of directly adjacent street light clusters needs to be compensated for basic brightness, the difference between the basic brightness of the reference street light cluster and the preset basic brightness difference threshold is used as the coordinated basic brightness of the street light cluster to be compensated. Finally, the coordinated basic brightness of the street light cluster to be compensated and the initial basic brightness of the uncompensated street light cluster are combined to obtain the basic brightness coordination result of the adjacent clusters.

[0143] When the street light cluster to be compensated is identified as the object to be compensated in multiple directly adjacent street light cluster pairs, the multiple coordinated basic brightness corresponding to the street light cluster to be compensated is calculated respectively, and the maximum value among the multiple coordinated basic brightness is determined as the final coordinated basic brightness of the street light cluster to be compensated in the current control cycle.

[0144] The coordinated baseline brightness is only used as the baseline brightness for the street light cluster to be compensated to perform collaborative power saving control within the cluster during the current control cycle. It does not participate in the baseline brightness compensation of other directly adjacent street light clusters, so as to avoid the continuous propagation of compensation results among multiple adjacent street light clusters.

[0145] Example 6, see Figure 1 , Figure 2 , Figure 5 and Figure 6 This embodiment is based on the above embodiment. In step S5, the coordinated energy-saving control of streetlights within the same cluster is used to perform differentiated optimization control on the target brightness of each streetlight within the same cluster, so that each streetlight reduces energy consumption while meeting the lighting needs of its location, and maintains the smoothness of brightness changes and dimming stability within the same cluster; specifically, it includes the following steps:

[0146] Step S51: Obtain the final base brightness of the street light cluster, which is used to determine the brightness benchmark for coordinated power-saving control of street lights within the cluster during the current control cycle; specifically, based on the coordination results of the base brightness of adjacent clusters, obtain the base brightness of each street light cluster and use it as the final base brightness of each street light cluster.

[0147] The final base brightness serves as the benchmark value for adjusting the brightness of each street light within the corresponding street light cluster;

[0148] Step S52: Obtaining the street light brightness adjustment range, used to determine the adjustable brightness range of each street light relative to the final base brightness of its respective street light cluster; specifically, based on the street light location attributes, determine the location attributes of each street light within the same street light cluster, and based on the final base brightness of the street light cluster and the location attributes of each street light, determine the allowable brightness adjustment range of each street light within the street light cluster relative to the final base brightness, thus obtaining the brightness adjustment range of each street light within the street light cluster;

[0149] The allowable brightness adjustment range is used to limit the brightness value range of each street light in subsequent optimization calculations, so that the target brightness of each street light meets the lighting needs of its location and does not deviate from the overall basic brightness level of the street light cluster to achieve energy-saving optimization. It includes the lower limit and upper limit of brightness adjustment of each street light relative to the final basic brightness of the street light cluster. Street lights with different location attributes correspond to different allowable brightness adjustment ranges. Specifically, the allowable brightness adjustment range of intersection lights is the final basic brightness plus 10% to 20%, the allowable brightness adjustment range of zebra crossing lights is the final basic brightness plus 10% to 20%, the allowable brightness adjustment range of bus stop lights is the final basic brightness plus 5% to 15%, the allowable brightness adjustment range of main road lights is the final basic brightness to the final basic brightness plus 15%, the allowable brightness adjustment range of ordinary road section lights is the final basic brightness minus 10% to the final basic brightness plus 10%, and the allowable brightness adjustment range of branch street lights is the final basic brightness minus 20% to the final basic brightness.

[0150] Step S53: Construct an objective function for optimizing street light brightness within the cluster, used to reduce the overall energy consumption of the street light cluster while maintaining smooth brightness and stable dimming within the cluster; specifically, using the target brightness of each street light within the same street light cluster as optimization variables, construct an objective function for optimizing street light brightness within the cluster that includes energy consumption optimization terms, brightness smoothing terms, and dimming stability terms; the formula used is as follows:

[0151] ;

[0152] ;

[0153] ;

[0154] ;

[0155] In the formula, This represents the objective function value for optimizing the brightness of streetlights within the cluster. This represents an energy consumption optimization item used to reduce the overall energy consumption of the street light cluster. This represents a brightness smoothing term, used to reduce the brightness difference between adjacent streetlights within the same streetlight cluster. This indicates a dimming stabilization term, used to reduce the magnitude of brightness changes between adjacent control cycles. , and These represent the weighting coefficients for the energy consumption optimization term, brightness smoothing term, and dimming stability term, respectively. This indicates the number of streetlights contained within a streetlight cluster. This represents the rated power of the r-th street light. This represents the target brightness of the r-th street light. Indicates the duration of the current control cycle. This represents the target brightness of the e-th street light. This indicates that the r-th street light and the e-th street light are adjacent street lights in the C-th street light cluster. This represents the target brightness of the r-th street light in the previous control cycle;

[0156] Step S54: Optimize the target brightness of streetlights within the cluster. This step searches for the target brightness combination of streetlights within the cluster that meets the requirements of energy saving, brightness smoothness, and dimming stability, within the allowable brightness adjustment range of each streetlight. Specifically, it uses an improved optimization algorithm to dynamically optimize and search for the target brightness combination of streetlights within the cluster, obtaining the optimal combination of target brightness for streetlights within the cluster. This includes the following steps:

[0157] Step S541: Initialize candidate target brightness combinations to generate the initial search population for optimizing the target brightness of streetlights within the cluster; specifically, encode the target brightness combination of each streetlight in the streetlight cluster into the position vector of the search individual in the optimization algorithm, and randomly generate P position vectors of search individuals based on the brightness adjustment range of each streetlight in the streetlight cluster, thereby initializing the candidate target brightness combinations and obtaining the initial search population.

[0158] Step S542: Calculate the fitness value of the search individual to evaluate the overall performance of the candidate target brightness combination corresponding to each search individual in terms of energy saving, brightness smoothing and dimming stability; specifically, calculate the fitness value of each search individual based on the objective function for optimizing the brightness of streetlights within the cluster.

[0159] Step S543: Dynamic elite mutation position update, used to adaptively balance global exploration and local development capabilities during iterative search, dynamically adjust the search range and step size, fully integrate global population information to guide the optimization direction, accelerate convergence while ensuring the diversity of brightness combinations, effectively avoid getting trapped in local optima, and significantly improve the accuracy, stability and convergence efficiency of street light brightness optimization. A three-layer adaptive optimization framework is constructed, consisting of dynamic shrinkage of the elite range, adaptive adjustment of mutation intensity, and guidance by global information fusion. Specifically, at the beginning of each search iteration, the elite proportion coefficient under the current iteration is first calculated. Based on this elite ratio coefficient, all individuals in the search population are sorted from best to worst according to their fitness values, and the top-ranked individuals are selected. The search individuals are grouped into an elite set. Then, the adaptive search step size and the average position of the population are calculated. Finally, based on the differential mutation mechanism of elite individuals, the average position of the population, and random individuals in the population, the mutation position of each search individual is updated. A greedy selection mechanism is used to compare the fitness values ​​of the search individuals before and after the update, retaining the search individuals with better fitness values. The formulas used are as follows:

[0160] ;

[0161] ;

[0162] ;

[0163] ;

[0164] ;

[0165] In the formula, t represents the current iteration number. Indicates the maximum number of iterations. This represents the elite proportion coefficient in the t-th iteration. This represents the adaptive search step size in the t-th iteration. This represents the initial search step size, with a value range of [value missing]. , This represents the average position of the population, used to integrate global search information. This represents the position of the p-th searched individual. Indicates the first The initial mutation position of the p-th search individual in the next iteration. Indicates the first The final position of the mutation of the p-th search individual in the next iteration. and These represent the weighting coefficients for elite individuals and the average position of the population, respectively, with values ​​ranging from [value range missing]. , This refers to a search individual randomly selected from a set of elite individuals. , , and This represents the positions of four distinct individuals randomly selected from the search population. This represents the objective function for optimizing streetlight brightness within the cluster, i.e., the fitness function of the optimization algorithm. Indicates the first The position of the p-th search individual in the next iteration;

[0166] Step S544: Update the behavior position of the search individual. This step is used to diversify the exploration of the solution space, further enhancing the diversity and globality of the population search and avoiding the local optimum trap caused by a single search mode. Specifically, for each search individual in the population, based on the dual conditions of random probability and iterative parity, one of the following strategies is selected: camouflage search strategy, blood-spraying defense search strategy, or escape movement search strategy. This completes the update of the search individual's behavior position. The formula used is as follows:

[0167] ;

[0168] ;

[0169] In the formula, The random perturbation term representing the camouflage search strategy is used to simulate the random search perturbation of the camouflage behavior of the search entity. and All indicate The random normalization coefficients of a uniformly distributed interval, and , Represents a random binary switch variable. , , , and This represents the positions of four distinct search individuals randomly selected from the search population after the dynamic elite mutation position update. Indicates the first The optimal search individual position in the next iteration. Let represent a random variable that follows a Cauchy distribution, with a mean of 0 and a standard deviation of 1. Represents the random step size coefficient. , Represents the gravitational acceleration constant. This represents a small perturbation term. express Randomness parameters within a range Indicates the first The position of the p-th search individual in the next iteration;

[0170] Step S545: Search iteration terminates. Specifically, after each iteration, the fitness values ​​of all search individuals in the current population are calculated. When the fitness value of a search individual is higher than the fitness threshold or the maximum number of iterations is reached, the search is terminated and the position of the globally optimal search individual is obtained. The globally optimal search individual position specifically refers to the optimal combination of target brightness of streetlights within the streetlight cluster.

[0171] Step S55: Obtain the street light control strategy within the cluster. Specifically, based on the optimal combination of target brightness of street lights within the street light cluster, extract the target brightness corresponding to each street light in the street light cluster, and match the target brightness of each street light with its respective street light cluster number and street light number to obtain the collaborative control strategy for street lights within the cluster.

[0172] The intra-cluster street light collaborative control strategy includes the street light cluster number, street light number, and target brightness.

[0173] By performing the above operations, this invention addresses the technical problems of traditional energy-saving control strategies relying heavily on manual experience, making it difficult to simultaneously achieve energy reduction, brightness smoothness, and dimming stability. Furthermore, optimization algorithms often suffer from low search efficiency, slow convergence, and getting trapped in local optima during complex streetlight brightness combination searches, resulting in limited energy-saving effects, frequent streetlight dimming, significant brightness fluctuations, unreasonable target brightness combinations, and unstable energy-saving performance. This invention proposes a cluster-based streetlight brightness collaborative optimization method. It constructs a cluster-based streetlight brightness optimization objective function including energy consumption optimization, brightness smoothness, and dimming stability terms, transforming the streetlight target brightness optimization into a method that considers energy saving, brightness continuity, and dimming stability. This paper addresses the multi-objective collaborative optimization problem of dimming stability and improves the optimization algorithm based on a dynamic elite mutation position update strategy. During the iteration process, information from elite individuals, the population average position, and random difference mutation is integrated to dynamically update the position of the search individuals. This enhances the global search capability and local exploitation capability of the optimization algorithm, improves the convergence efficiency and stability of the optimization search, obtains the optimal combination of target brightness for each street light within the street light cluster, reduces the overall energy consumption of the street light cluster, minimizes brightness differences between adjacent street lights, suppresses brightness abrupt changes in adjacent control cycles, and improves dimming stability. This achieves efficient optimization of target brightness allocation within the street light cluster and comprehensive optimization of energy saving, comfort, and dimming stability.

[0174] Example 7, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S6, the intelligent street light control is used to perform dimming control on each street light according to the street light collaborative control strategy within the cluster. Specifically, based on the basic brightness coordination results of adjacent clusters and through the street light collaborative power saving control within the cluster, the street light collaborative control strategy within each cluster is obtained. Based on the street light collaborative control strategy within the cluster, dimming control instructions are generated for the street lights in each street light cluster. According to the dimming control instructions, each street light is dimmed according to the corresponding target brightness, thereby realizing intelligent control of street light cluster collaborative power saving.

[0175] Example 8, see Figure 1 and Figure 2 Based on the above embodiments, the big data-based street light cluster collaborative energy-saving control system provided by the present invention includes a standard data acquisition module, a street light cluster division module, a cluster lighting demand prediction module, an adjacent cluster street light collaborative energy-saving control module, an intra-cluster street light collaborative energy-saving control module, and a street light intelligent control module.

[0176] The standard data acquisition module specifically obtains standardized street light energy-saving control data through data acquisition and data optimization, and sends the data to the street light cluster division module, the cluster lighting demand prediction module, and the cluster-based street light collaborative energy-saving control module.

[0177] The street light cluster partitioning module receives data sent by the standard data acquisition module and is used to partition street lights with similar lighting needs and spatial and road topological relationships into the same street light cluster. Specifically, it improves the clustering algorithm by generating coarse-grained grid blocks through a multi-level diffused neighbor search strategy, calculating dual density, and using a dual density progressive comparison mechanism to construct a street light partitioning clustering algorithm. The target street light cluster partitioning optimization data is input into the street light partitioning clustering algorithm to perform real-time street light cluster partitioning, obtain the street light cluster partitioning results, and send the data to the cluster lighting demand prediction module and the adjacent cluster street light collaborative energy saving control module.

[0178] The cluster lighting demand prediction module receives data sent by the standard data acquisition module and the street light cluster division module, and is used to predict the lighting demand level of each street light cluster within a future preset time window. Specifically, it constructs and trains a cluster lighting demand prediction model. Based on the street light cluster division results and the trained cluster lighting demand prediction model, it finally obtains the initial basic brightness of each street light cluster and sends the data to the adjacent cluster street light collaborative energy saving control module.

[0179] The adjacent cluster street light collaborative energy-saving control module receives data sent by the street light cluster division module and the cluster lighting demand prediction module, and is used to coordinate the basic brightness between directly adjacent street light clusters. Specifically, based on the street light cluster division result and the initial basic brightness of each street light cluster, it calculates the basic brightness difference between two directly adjacent street light clusters, and performs basic brightness compensation based on the basic brightness difference to obtain the basic brightness of the street light cluster to be compensated after coordination. Finally, it combines the basic brightness corresponding to each street light cluster to obtain the basic brightness coordination result of adjacent clusters, and sends the data to the street light collaborative energy-saving control module within the cluster.

[0180] The cluster-based street light collaborative energy-saving control module receives data from the standard data acquisition module and the adjacent cluster street light collaborative energy-saving control module. This data is used to perform differentiated optimization control on the target brightness of each street light within the same street light cluster. Specifically, based on the basic brightness coordination results of adjacent clusters, the final basic brightness of each street light cluster is obtained. According to the positional attributes of each street light within the same cluster, the brightness adjustment range of each street light within the cluster is obtained. A cluster-based street light brightness optimization objective function is constructed, including energy consumption optimization, brightness smoothing, and dimming stability terms. The optimization algorithm is improved based on a dynamic elite mutation position update strategy. The improved optimization algorithm is used to optimize the target brightness of street lights within the cluster within the street light brightness adjustment range, obtaining the optimal combination of target brightness for street lights within the cluster. Finally, based on this combination, a cluster-based street light collaborative control strategy is obtained, and the data is sent to the street light intelligent control module.

[0181] The street light intelligent control module receives data sent by the street light collaborative energy-saving control module within the cluster, and performs street light dimming control for each street light cluster through the cluster-based street light collaborative control strategy, thereby realizing intelligent collaborative energy-saving control of the street light cluster.

[0182] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0183] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0184] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for collaborative energy-saving control of street light clusters based on big data, characterized in that: The method includes the following steps: Step S1: Obtain standard data. Through data acquisition and optimization, obtain standardized data for street light energy-saving control. Step S2: Streetlight cluster partitioning. The clustering algorithm is improved by generating coarse-grained grid blocks through a multi-level diffused neighbor search strategy, calculating dual density, and using a dual density progressive comparison mechanism. A streetlight partitioning clustering algorithm is constructed. The target streetlight cluster partitioning optimization data is input into the streetlight partitioning clustering algorithm to obtain the streetlight cluster partitioning results. Step S3: Cluster lighting demand prediction. By constructing and training a cluster lighting demand prediction model, the initial basic brightness of each street light cluster is obtained based on the street light cluster division results and the trained cluster lighting demand prediction model. Step S4: Coordinated power saving control of adjacent street light clusters. Based on the street light cluster division results and the initial base brightness of each street light cluster, the base brightness difference between two directly adjacent street light clusters is calculated, and base brightness compensation is performed according to the base brightness difference. Finally, the coordinated base brightness result of adjacent clusters is obtained. Step S5: Collaborative power-saving control of streetlights within the cluster. Based on the coordination results of the basic brightness of adjacent clusters, the final basic brightness of each streetlight cluster is obtained, and an objective function for optimizing the brightness of streetlights within the cluster is constructed. Then, the optimization algorithm is improved based on the dynamic elite mutation position update strategy. Within the range of streetlight brightness adjustment, the target brightness of streetlights within the cluster is optimized to obtain the optimal combination of target brightness of streetlights within the streetlight cluster. Finally, the collaborative control strategy of streetlights within the cluster is obtained based on this combination. Step S6: Intelligent control of streetlights. Through the intra-cluster collaborative control strategy of each streetlight cluster, the dimming control of each streetlight cluster is carried out to achieve intelligent control of streetlight cluster collaborative power saving.

2. The street light cluster collaborative energy-saving control method based on big data according to claim 1, characterized in that: The street light cluster division specifically includes the following steps: Step S21: Multidimensional feature space gridding mapping processing, specifically, firstly, the value range of each feature dimension in the target street light cluster partitioning optimization data is statistically analyzed, the maximum and minimum values ​​of each feature dimension are calculated, and according to the preset single-dimensional grid partitioning quantity parameter s, the partitioning step size of each dimension is calculated. Each feature dimension is evenly divided into s intervals according to the partitioning step size to construct an m-dimensional discrete grid space. Finally, the grid index of each data sample in each dimension is calculated, and each data sample is mapped to the corresponding grid cell to obtain the grid cell belonging relationship of all data samples. Step S22: Calculate the dual density of the mesh cells; Step S23: Generate coarse-grained mesh blocks; Step S24: Calculate the dual density of the coarse-grained mesh blocks; Step S25: Cluster center selection; Step S26: Full allocation of coarse-grained grid blocks. Specifically, firstly, the nearest neighbors of each coarse-grained grid block are calculated based on the dual density progressive comparison mechanism to establish the density adjacency relationship between coarse-grained grid blocks. Then, according to the order of the first-level overall density between coarse-grained grid blocks from high to low, all non-cluster center coarse-grained grid blocks are clustered and allocated in sequence. The current coarse-grained grid block is assigned to the cluster to which its nearest neighbor belongs. After all coarse-grained grid blocks have been clustered, the final clustering result is obtained. Step S27: Real-time partitioning of street light clusters. Specifically, a street light partitioning clustering algorithm is constructed by first performing multi-dimensional feature space gridding mapping processing, calculating the dual density of grid cells, generating coarse-grained grid blocks, calculating the dual density of coarse-grained grid blocks, selecting cluster centers, and fully distributing coarse-grained grid blocks. The target street light cluster partitioning optimization data is then input into the street light partitioning clustering algorithm to perform real-time clustering of street lights and obtain the street light cluster partitioning results.

3. The street light cluster collaborative energy-saving control method based on big data according to claim 2, characterized in that: The calculation of the dual density of the grid cells involves first counting the number of streetlight data samples contained in each grid cell, using this as the first-level density value for that grid cell. For grid cells containing two or more data samples, the ratio of their first-level density value to the sum of the Euclidean distances between all data samples within that grid cell is calculated, and this ratio is used as the second-level density value for that grid cell. Finally, all grid cells are sorted, first in descending order of their first-level density values, and then in descending order of their second-level density values ​​if the first-level density values ​​are the same, resulting in a density sorting sequence for the grid cells. The formula used is as follows: ； In the formula, This represents the first-level density value of the k-th grid cell. This represents the second-level density value of the k-th grid cell. and These represent the p-th and q-th data samples belonging to the k-th grid cell, respectively. This represents the Euclidean distance between the p-th and q-th data samples. This represents the k-th grid cell.

4. The street light cluster collaborative energy-saving control method based on big data according to claim 2, characterized in that: The generation of coarse-grained grid blocks specifically involves generating coarse-grained grid blocks based on the density sorting sequence of grid cells, using a multi-level diffused neighbor search strategy, and includes the following steps: Step S231: Initialize the starting grid cell. Specifically, traverse the grid cell density sorting sequence, select the grid cell with the highest density that has not been marked as the starting grid cell of the current coarse-grained grid block, add the grid cell to the queue to be processed, and immediately mark it as processed. Step S232: Multi-level diffusion search. Specifically, when the queue to be processed is not empty, the head grid cell is taken out and multi-level diffusion search is performed in all geometrically adjacent directions. The first-level neighbors and second-level neighbors are searched in turn until there are no new unmarked neighbors. For each searched neighbor grid cell, it is determined whether it contains data samples and is not marked. If it contains data samples and is not marked, the neighbor grid cell is added to the queue to be processed and immediately marked as processed. The above process is repeated until the queue to be processed is empty. All marked grid cells are aggregated into a coarse-grained grid block. Step S233: Mesh cell traversal, specifically, repeatedly executing the initial mesh cell initialization and multi-level diffusion search until all mesh cells are divided into corresponding coarse-grained mesh blocks, obtaining the complete set of coarse-grained mesh blocks; The calculation of the dual density of the coarse-grained grid block specifically involves first summing the first-level density values ​​and second-level density values ​​of all grid cells within the coarse-grained grid block to obtain the first-level overall density and second-level overall density of the coarse-grained grid block. Then, based on the first-level density of each grid cell and the first-level overall density of the coarse-grained grid block, the density weighting coefficient of each grid cell within the coarse-grained grid block is calculated. Based on the density weighting coefficient, the center coordinates of each grid cell within the coarse-grained grid block are weighted and averaged to obtain the weighted centroid coordinates of the coarse-grained grid block, which serve as the spatial representative point of the coarse-grained grid block.

5. The street light cluster collaborative energy-saving control method based on big data according to claim 2, characterized in that: The selection of cluster centers specifically involves calculating the Euclidean distance between any two coarse-grained grid blocks based on the weighted centroid coordinates of each coarse-grained grid block, constructing a coarse-grained grid block distance matrix, then introducing a dual density progressive comparison mechanism to calculate the relative distance between each coarse-grained grid block, and finally calculating the clustering decision value of each coarse-grained grid block based on the first-level overall density and relative distance of each coarse-grained grid block. The clustering decision values ​​of all coarse-grained grid blocks are then sorted in descending order, and the top c coarse-grained grid blocks with the highest clustering decision values ​​are selected as cluster centers, where c represents the number of cluster centers. The dual density progressive comparison mechanism specifically compares the densities of any two coarse-grained grid blocks according to a priority-progressive rule. First, it compares the overall density of the first level of the two coarse-grained grid blocks. If there is a difference in the overall density of the first level, the density is directly determined. If the overall density of the first level is equal, then the overall density of the second level of the two coarse-grained grid blocks is compared. The formula used is as follows: ； ； In the formula, This represents the relative distance of the nth coarse-grained grid block. This represents the v-th coarse-grained grid block. This represents the nth coarse-grained grid block. This represents the Euclidean distance between the nth and vth coarse-grained grid blocks. This represents the clustering decision value of the nth coarse-grained grid block. This represents the first-level global density of the v-th coarse-grained grid block. This represents the second-order global density of the v-th coarse-grained grid block. This represents the first-level global density of the nth coarse-grained grid block. This represents the second-level overall density of the nth coarse-grained grid block.

6. The street light cluster collaborative energy-saving control method based on big data according to claim 1, characterized in that: The cluster lighting demand prediction specifically involves first constructing a cluster lighting demand prediction model based on a long short-term memory neural network, and using historical lighting demand prediction optimization data as model training data to iteratively train the prediction model to obtain the trained cluster lighting demand prediction model. Then, based on the street light cluster division results, the real-time lighting demand prediction optimization data of each street light cluster is input into the trained cluster lighting demand prediction model to obtain the real-time lighting demand prediction results of each street light cluster. Finally, based on the real-time lighting demand prediction results, the initial base brightness of each street light cluster is obtained. The coordinated power-saving control of adjacent cluster streetlights specifically includes the following steps: Step S41: Calculate the basic brightness difference. Specifically, based on the street light cluster division results, determine the direct adjacency relationship between each street light cluster to obtain directly adjacent street light cluster pairs. Then, based on the initial basic brightness of each street light cluster, calculate the basic brightness difference between directly adjacent street light cluster pairs. Step S42: Basic brightness compensation judgment, specifically, compare the basic brightness difference between directly adjacent street light clusters with a preset basic brightness difference threshold. If the basic brightness difference is greater than the preset basic brightness difference threshold, it is determined that the corresponding directly adjacent street light clusters need to be compensated for basic brightness. The street light cluster with high initial basic brightness is determined as the reference street light cluster, and the street light cluster with low initial basic brightness is determined as the street light cluster to be compensated. Step S43: Basic brightness compensation. Specifically, when a pair of directly adjacent street light clusters needs basic brightness compensation, the difference between the basic brightness of the reference street light cluster and the preset basic brightness difference threshold is used as the coordinated basic brightness of the street light cluster to be compensated. Finally, the coordinated basic brightness of the street light cluster to be compensated and the initial basic brightness of the uncompensated street light cluster are combined to obtain the basic brightness coordination result of the adjacent clusters.

7. The street light cluster collaborative energy-saving control method based on big data according to claim 1, characterized in that: The coordinated power-saving control of streetlights within the cluster specifically includes the following steps: Step S51: Obtain the final base brightness of the street light cluster. Specifically, based on the coordination results of the base brightness of adjacent clusters, obtain the base brightness of each street light cluster and use it as the final base brightness of each street light cluster. Step S52: Obtain the street light brightness adjustment range. Specifically, based on the street light location attributes, determine the location attributes of each street light in the same street light cluster. Based on the final base brightness of the street light cluster and the location attributes of each street light, determine the allowable brightness adjustment range of each street light in the street light cluster relative to the final base brightness, and obtain the brightness adjustment range of each street light in the street light cluster. Step S53: Construct the objective function for optimizing the brightness of streetlights within the cluster. Specifically, the objective function for optimizing the brightness of streetlights within the same streetlight cluster is constructed by taking the target brightness of each streetlight in the same streetlight cluster as the optimization variable and including the energy consumption optimization term, the brightness smoothing term, and the dimming stability term. Step S54: Optimize the target brightness of streetlights within the cluster; Step S55: Obtain the street light control strategy within the cluster. Specifically, based on the optimal combination of target brightness of street lights within the street light cluster, extract the target brightness corresponding to each street light in the street light cluster, and match the target brightness of each street light with its respective street light cluster number and street light number to obtain the collaborative control strategy for street lights within the cluster.

8. The street light cluster collaborative energy-saving control method based on big data according to claim 7, characterized in that: The optimization of target brightness for streetlights within the cluster specifically includes the following steps: Step S541: Initialize candidate target brightness combinations. Specifically, the target brightness combination of each street light in the street light cluster is encoded into the search individual position vector in the optimization algorithm. Based on the brightness adjustment range of each street light in the street light cluster, P search individual position vectors are randomly generated to initialize the candidate target brightness combinations and obtain the initial search population. Step S542: Calculate the fitness value of the search individual, specifically by calculating the fitness value of each search individual based on the objective function of optimizing the brightness of streetlights within the cluster; Step S543: Dynamic elite mutation position update, specifically, at the beginning of each search iteration, first calculate the elite proportion coefficient under the current iteration. Based on this elite ratio coefficient, all individuals in the search population are sorted from best to worst according to their fitness values, and the top-ranked individuals are selected. The search individuals are grouped into an elite set. Then, the adaptive search step size and the average position of the population are calculated. Finally, based on the differential mutation mechanism of elite individuals, the average position of the population, and random individuals in the population, the mutation position of each search individual is updated. A greedy selection mechanism is used to compare the fitness values ​​of the search individuals before and after the update, retaining the search individuals with better fitness values. The formulas used are as follows: ； ； ； In the formula, t represents the current iteration number. Indicates the maximum number of iterations. This represents the elite proportion coefficient in the t-th iteration. This represents the adaptive search step size in the t-th iteration. Indicates the initial search step size. Indicates the average position of the population. Indicates the first The initial mutation position of the p-th search individual in the next iteration. and These represent the weighting coefficients for elite individuals and the population average position, respectively. This refers to a search individual randomly selected from a set of elite individuals. , , and This represents the positions of four distinct individuals randomly selected from the search population. Step S544: Update the behavior position of the search individual. Specifically, for each search individual in the population, based on the dual conditions of random probability and iterative parity, select one of the following strategies to execute: camouflage search strategy, blood-spraying defense search strategy, and escape movement search strategy, to complete the behavior position update of the search individual. Step S545: Search iteration terminates. Specifically, after each iteration, the fitness values ​​of all search individuals in the current population are calculated. When the fitness value of a search individual is higher than the fitness threshold or the maximum number of iterations is reached, the search is terminated and the globally optimal search individual position is obtained. The globally optimal search individual position specifically refers to the optimal combination of target brightness of streetlights within the streetlight cluster.

9. The street light cluster collaborative energy-saving control method based on big data according to claim 1, characterized in that: The acquisition of standard data specifically involves obtaining raw street light energy-saving control data through data acquisition operations, and then performing data optimization processing on the raw street light energy-saving control data to obtain standardized street light energy-saving control data. The data optimization processing specifically involves sequentially performing data cleaning, standardization processing, and feature selection on the raw street light energy-saving control data to obtain standardized street light energy-saving control data. The standardized data for street light energy-saving control includes target street light cluster division optimization data, historical lighting demand prediction optimization data, and real-time lighting demand prediction optimization data. The intelligent control of streetlights specifically involves obtaining the intra-cluster streetlight collaborative control strategy of each streetlight cluster based on the basic brightness coordination results of adjacent clusters and through intra-cluster streetlight collaborative power saving control. Based on the intra-cluster streetlight collaborative control strategy, dimming control instructions are generated for the streetlights in each streetlight cluster. According to the dimming control instructions, each streetlight is dimmed according to the corresponding target brightness, thereby realizing the intelligent control of streetlight cluster collaborative power saving.

10. A big data-based collaborative energy-saving control system for streetlight clusters, used to implement the big data-based collaborative energy-saving control method for streetlight clusters as described in any one of claims 1-9, characterized in that: This includes a standard data acquisition module, a street light cluster division module, a cluster lighting demand prediction module, an adjacent cluster street light collaborative energy-saving control module, an intra-cluster street light collaborative energy-saving control module, and a street light intelligent control module; The standard data acquisition module specifically obtains standardized street light energy-saving control data through data acquisition and data optimization, and sends the data to the street light cluster division module, the cluster lighting demand prediction module, and the cluster-based street light collaborative energy-saving control module. The street light cluster partitioning module receives data sent by the standard data acquisition module, improves the clustering algorithm through coarse-grained grid block generation, dual density calculation, and dual density progressive comparison mechanism using a multi-level diffused neighbor search strategy, and constructs a street light partitioning clustering algorithm. The target street light cluster partitioning optimization data is input into the street light partitioning clustering algorithm to perform real-time street light cluster partitioning, obtain the street light cluster partitioning results, and send the data to the cluster lighting demand prediction module and the adjacent cluster street light collaborative power saving control module. The cluster lighting demand prediction module receives data sent by the standard data acquisition module and the street light cluster division module. By constructing and training the cluster lighting demand prediction model, and based on the street light cluster division results and the trained cluster lighting demand prediction model, the initial basic brightness of each street light cluster is finally obtained, and the data is sent to the adjacent cluster street light collaborative energy saving control module. The adjacent cluster street light collaborative energy-saving control module receives data sent by the street light cluster division module and the cluster lighting demand prediction module. Based on the street light cluster division result and the initial base brightness of each street light cluster, it calculates the base brightness difference between two directly adjacent street light clusters, and performs base brightness compensation based on the base brightness difference to obtain the base brightness of the street light cluster to be compensated after coordination. Finally, it combines the base brightness corresponding to each street light cluster to obtain the basic brightness coordination result of the adjacent clusters, and sends the data to the street light collaborative energy-saving control module within the cluster. The cluster-based street light collaborative energy-saving control module receives data sent by the standard data acquisition module and the adjacent cluster street light collaborative energy-saving control module. Based on the basic brightness coordination results of adjacent clusters, it obtains the final basic brightness of each street light cluster. According to the position attributes of each street light in the same street light cluster, it obtains the brightness adjustment range of each street light in the street light cluster and constructs a cluster-based street light brightness optimization objective function including energy consumption optimization, brightness smoothing, and dimming stability. Based on the dynamic elite mutation position update strategy, the optimization algorithm is improved. The improved optimization algorithm is used to optimize the target brightness of the street lights in the cluster within the street light brightness adjustment range, obtains the optimal combination of target brightness of the street lights in the street light cluster, and finally obtains the cluster-based street light collaborative control strategy based on this combination and sends the data to the street light intelligent control module. The street light intelligent control module receives data sent by the street light collaborative energy-saving control module within the cluster, and performs street light dimming control for each street light cluster through the cluster-based street light collaborative control strategy, thereby realizing intelligent collaborative energy-saving control of the street light cluster.

Images