Method for establishing intelligent control model of solar street lamp and solar street lamp
By establishing an intelligent control model for solar street lights, and using historical power generation data and real-time fluctuation sequence clustering and prediction mechanisms, the shortcomings of solar street lights in intelligent control and energy management are solved, precise power generation prediction and street light control optimization are achieved, and the stability of the system and energy utilization efficiency are improved.
Patent Information
- Application Number
- CN202510433457.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-08
- Publication Date
- 2025-07-22
AI Technical Summary
Solar street lights have shortcomings in intelligent control and energy management, which leads to energy waste and system stability problems, especially when the lighting conditions are poor, reducing power generation efficiency, affecting the reliability and use effect of street lights.
By establishing an intelligent control model for solar street lights, historical power generation data are obtained for cluster analysis, predictive models are constructed, and weighted integration is used to utilize the similarity between real-time fluctuation sequences and historical cluster clusters to optimize street light control strategies.
It realizes accurate prediction of solar power generation and street light control optimization, improves energy usage efficiency and system stability, reduces maintenance costs, and enhances the intelligence level of the model.
Smart Images
Figure CN120354076A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of solar street lamp control, and particularly to a method for establishing an intelligent control model of a solar street lamp and a solar street lamp. Background Art
[0002] As a green, renewable and environmentally friendly energy source, solar energy has been gradually fully applied in various industries. In urban infrastructure construction, as a public facility driven by clean energy, solar street lamps have become lighting solutions for more and more urban streets, roads, squares and other places. Compared with traditional electric street lamps, solar street lamps have the advantages of no electricity cost, simple installation and long service life.
[0003] However, although solar street lamps have many advantages, they also face some challenges in practical applications, especially in intelligent control, energy management and system stability. Traditional solar street lamps usually adjust their working states through simple time switches or photocells, lacking intelligent control and system adaptability. This often leads to overuse of batteries or failure to effectively utilize the charging time during the day, thus affecting the use efficiency and long-term operation reliability of the street lamps. When the solar panels are under poor lighting conditions such as cloudy days, rainy days or winter, their power generation efficiency decreases, resulting in the street lamps may not work properly, affecting their reliability and use effect. Summary of the Invention
[0004] The object of the present invention is to solve the problem of energy waste caused by inaccurate control of solar lamps.
[0005] To achieve the above object, the present invention provides a method for establishing an intelligent control model of a solar street lamp and a solar street lamp, including: obtaining the historical power generation of solar energy to construct a power generation sequence; performing a first-order difference on the power generation sequence to obtain a fluctuation sequence, performing initial clustering on the fluctuation sequence to obtain initial clustering clusters, and merging the initial clustering clusters to obtain several clustering clusters, constructing a prediction model for each clustering cluster, with one clustering cluster corresponding to one prediction model, and one clustering cluster containing one fluctuation subsequence; obtaining a real-time power generation sequence and calculating a real-time fluctuation sequence, respectively calculating the similarity between the real-time fluctuation sequence and each clustering cluster, inputting the real-time fluctuation sequence into each prediction model to obtain several power generation prediction values; setting a control model, using the similarity as a weight in the control model to integrate the power generation prediction values to obtain a final prediction value, and controlling the solar street lamp according to the control model.
[0006] Preferably, the obtaining of the initial clustering clusters includes: performing initial clustering on all elements in the fluctuation sequence to obtain initial clustering clusters, retaining the initial clustering clusters that meet the preset conditions, and iteratively dividing the initial clustering clusters that do not meet the preset conditions until all initial clustering clusters that meet the preset conditions are obtained, where the preset condition is that the elements in the initial clustering cluster are continuously adjacent in the fluctuation sequence.
[0007] Preferably, the obtaining of the initial clustering clusters includes: calculating the standard deviation of the fluctuation sequence, calculating the element difference between any two adjacent elements in the fluctuation sequence, and when the element difference is not greater than one standard deviation, merging any two adjacent elements into one initial clustering cluster.
[0008] Preferably, the merging of the initial clustering clusters to obtain several clustering clusters includes: calculating the difference between any two adjacent initial clustering clusters, and merging any two adjacent initial clustering clusters with a difference less than the preset difference threshold to obtain several clustering clusters.
[0009] Preferably, the difference between any two adjacent initial clustering clusters includes: taking the value of the dynamic time warping between any two adjacent initial clustering clusters as the difference.
[0010] Preferably, it is characterized in that the difference between any two adjacent initial clustering clusters satisfies the relational expression:
[0011] s i,i+1 represents the difference between the i-th initial clustering cluster and the (i + 1)-th initial clustering cluster, represents the value of the m-th element in the i-th initial clustering cluster, represents the value of the n-th element in the (i + 1)-th initial clustering cluster, M represents the total number of elements in the i-th initial clustering cluster, and N represents the total number of elements in the (i + 1)-th initial clustering cluster.
[0012] Preferably, the prediction model is a long short-term memory recurrent neural network or an ARIMA model.
[0013] Preferably, the final predicted value satisfies the relational expression:
[0014] CQ represents the final predicted value, ρ t represents the similarity between the real-time fluctuation sequence and the t-th clustering cluster, cq t represents the power generation prediction value of the prediction model corresponding to the t-th clustering cluster.
[0015] Preferably, the similarity satisfies the relational expression:
[0016] ρ represents the similarity, X kDenote the k-th fluctuation subsequence, Y denote the real-time fluctuation sequence, and COV denote the covariance. Denote the standard deviation of the k-th fluctuation subsequence, σ Y Denote the standard deviation of the real-time fluctuation sequence, and K denote the total number of fluctuation subsequences.
[0017] In a second aspect, the present invention discloses a solar street lamp, which is characterized in that it is controlled according to the prediction model established in the method for establishing an intelligent control model of a solar street lamp.
[0018] Advantages of the present invention:
[0019] By establishing a clustering and prediction mechanism based on historical power generation data and real-time fluctuation sequences, the present invention realizes accurate prediction of solar power generation and optimization of street lamp control. The model analyzes the power generation fluctuations through clustering, and weights and integrates multiple prediction results according to the similarity between the real-time fluctuation sequence and the historical clustering clusters, thereby improving the accuracy and robustness of the prediction. It can effectively cope with power generation fluctuations caused by factors such as weather and seasons, provides an intelligent control scheme for solar street lamps based on actual power generation, ensures the stability of energy supply under different environmental conditions, optimizes the energy use efficiency, reduces the maintenance cost, and improves the intelligence level of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 is a flowchart of the method for establishing an intelligent control model of a solar street lamp according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.
[0022] Next, the specific embodiments of the present invention will be described in detail in conjunction with the accompanying drawings.
[0023] Refer to Figure 1 , the method for establishing an intelligent control model of a solar street lamp includes steps S1 - S4, which will be specifically described below.
[0024] S1: Obtain the historical power generation of solar energy and construct a power generation sequence.
[0025] In one embodiment, the charging amount data, i.e., the power generation amount, at each sampling moment of the solar street lamp in the collection history is collected, and a power generation amount sequence is constructed based on the power generation amounts in consecutive preset time periods. There may be missing values or outliers in the power generation amount sequence, and it needs to be preprocessed. The steps of preprocessing include filling in the missing values to ensure the integrity of the data, and deleting the outliers to remove possible measurement errors or extreme data points, so as to ensure the accuracy and reliability of the power generation amount sequence.
[0026] S2: Perform a first-order difference on the power generation amount sequence to obtain a fluctuation sequence, perform initial clustering on the fluctuation sequence to obtain initial clustering clusters, merge the initial clustering clusters to obtain several clustering clusters, and construct a prediction model for each clustering cluster. The clustering clusters and the prediction models are in one-to-one correspondence, and each clustering cluster contains a fluctuation subsequence.
[0027] It should be noted that when predicting the future charging amount using the historical charging amount sequence, since the solar power generation is affected by factors such as weather and temperature, the power generation amount usually shows volatility, which in turn causes large fluctuations in the charging amount sequence itself. If all historical data is directly used to train the prediction model, the model may not be able to fully identify the fluctuation characteristics in different time periods of the sequence, because the influence degrees of factors such as weather on different time periods are inconsistent. For example, the difference in the influence of sunny days and cloudy days on the power generation amount is relatively large. If the model fails to distinguish these differences, it may lead to inaccurate prediction results. Therefore, simply relying on global historical data for training is likely to cause the model to fail to capture the seasonal fluctuations and weather change characteristics in the time series, thereby reducing the prediction accuracy.
[0028] In one embodiment, a first-order difference is performed on the power generation amount sequence to obtain a fluctuation sequence, and initial clustering is performed on the fluctuation sequence to obtain initial clustering clusters. Since the fluctuation sequence is a continuous time series, it is necessary to ensure that the elements within the clustering clusters are also continuously adjacent in time order during the clustering process.
[0029] Initial clustering is performed on all elements in the fluctuation sequence to obtain initial clustering clusters. The initial clustering clusters that meet the preset conditions are retained, and the initial clustering clusters that do not meet the preset conditions are iteratively segmented until all initial clustering clusters that meet the preset conditions are obtained. The preset condition is that the elements within the initial clustering cluster are continuously adjacent in the fluctuation sequence.
[0030] Among them, obtaining the initial clustering clusters includes: calculating the standard deviation of the fluctuation sequence, calculating the element difference between any two adjacent elements in the fluctuation sequence. When the element difference is not greater than one standard deviation, any two adjacent elements are merged into one initial clustering cluster.
[0031] Exemplarily, the fluctuation sequence is {l1, l2, l3, l4, l5, l6, l7, l8, l9, l 10}, the clustering clusters obtained according to the clustering method are {l1, l2, l3}, {l4, l6}, {l5, l7, l8, l9, l 10}, the first clustering cluster meets the preset conditions, but the second and third clustering clusters do not meet the preset conditions. The second and third clustering clusters need to be iteratively segmented to ensure that there are no discontinuous elements within the clustering clusters. The initial clustering clusters after segmentation are {l1, l2, l3}, {l4}, {l5}, {l6}, {l7, l8, l9, l 10}.
[0032] Calculate the differences between any two adjacent initial clustering clusters. The differences include: taking the dynamic time warping value between any two adjacent initial clustering clusters as the difference. Merge any two adjacent initial clustering clusters with differences less than the preset difference threshold to obtain several clustering clusters.
[0033] Construct a prediction model for the clustering clusters. The prediction model is a long short-term memory recurrent neural network or an ARIMA model. The clustering clusters and the prediction models are in one-to-one correspondence, and one clustering cluster contains one fluctuating subsequence.
[0034] In another embodiment, obtaining the initial clustering clusters includes: calculating the standard deviation of the fluctuating sequence, calculating the element differences between any two adjacent elements in the fluctuating sequence. When the element difference is not greater than one standard deviation, merge any two adjacent elements into one initial clustering cluster.
[0035] In another embodiment, the differences between any two adjacent initial clustering clusters satisfy the relational expression: s i,i+1 represents the difference between the i-th initial clustering cluster and the (i + 1)-th initial clustering cluster, represents the value of the m-th element in the i-th initial clustering cluster, represents the value of the n-th element in the (i + 1)-th initial clustering cluster, M represents the total number of elements in the i-th initial clustering cluster, and N represents the total number of elements in the (i + 1)-th initial clustering cluster.
[0036] S3: Obtain the real-time power generation sequence and calculate the real-time fluctuating sequence, calculate the similarity between the real-time fluctuating sequence and each clustering cluster respectively, and input the real-time fluctuating sequence into each prediction model to obtain several power generation prediction values.
[0037] In one embodiment, obtain the real-time power generation sequence and calculate the real-time fluctuating sequence, calculate the similarity between the real-time fluctuating sequence and each clustering cluster respectively. The similarity satisfies the relational expression:
[0038] ρ represents the similarity, X kDenote the k-th fluctuating subsequence, Y denote the real-time fluctuation sequence, COV denote the covariance, and σ Xk denote the standard deviation of the k-th fluctuating subsequence, and σ Y denote the standard deviation of the real-time fluctuation sequence, and K denote the total number of fluctuating subsequences.
[0039] It can accurately capture the potential laws in historical data and compare them with the current real-time fluctuation situation, thus providing a strong basis for prediction and adjustment.
[0040] S4: Set up a control model. In the control model, use the similarity as a weight to integrate the predicted power generation values to obtain the final predicted value, and control the solar street lamp according to the control model.
[0041] It should be noted that the control model contains a preset control method.
[0042] In one embodiment, input the real-time fluctuation sequence into each prediction model respectively to obtain several predicted power generation values, and the number of prediction models is equal to the number of predicted power generation values. Use the similarity as a weight to integrate the predicted power generation values to obtain the final predicted value, and the final predicted value satisfies the relationship:
[0043] CQ denote the final predicted value, and ρ t denote the similarity between the real-time fluctuation sequence and the t-th clustering cluster, and cq t denote the predicted power generation value of the prediction model corresponding to the t-th clustering cluster.
[0044] By inputting the real-time fluctuation sequence into different prediction models respectively and performing weighted integration according to the similarity of each model, it can effectively utilize the advantages of different models, synthesize multiple prediction results, and thus obtain a more accurate final predicted value. By introducing the similarity as a weight, it can ensure that the prediction model with a higher similarity to the current real-time fluctuation sequence has a greater impact on the final result. In this way, it can not only improve the prediction accuracy in various different situations, but also enhance the adaptability of the model to different environmental changes. Finally, by integrating the prediction results of each model, it can reduce the deviation and error of a single model and improve the stability and reliability of the prediction.
[0045] Obtain the final predicted value and the remaining power of the solar street lamp, and control the solar street lamp according to the preset control method in the control model.
[0046] The preset control method is as follows: when the final predicted value is large and the remaining power is large, it indicates that the system has sufficient energy support in the future. Therefore, the brightness of the solar street lamp can be set relatively high; when the final predicted value is large but the remaining power is small, considering that the current energy is relatively tight but the future charging amount is sufficient, the brightness of the street lamp is set to medium; when the final predicted value is small but the remaining power is large, it means that there is a large amount of remaining power currently but the future charging amount is insufficient, so the brightness of the street lamp remains medium; when both the final predicted value and the remaining power are small, it indicates that the energy is tight. To save power, the brightness of the street lamp is set to be small.
[0047] A solar street lamp is controlled according to the control model in the method for establishing an intelligent control model of a solar street lamp.
[0048] It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all fall within the protection scope of this application. Therefore, the protection scope of the patent of this application shall be subject to the appended claims.
Claims
1. A method for establishing an intelligent control model of a solar street lamp, characterized in that Including: Obtain the power generation of solar energy in history and construct a power generation sequence; Perform a first-order difference on the power generation sequence to obtain a fluctuation sequence, perform initial clustering on the fluctuation sequence to obtain initial clustering clusters, merge the initial clustering clusters to obtain several clustering clusters, and construct a prediction model for each clustering cluster. The clustering clusters and the prediction models are in one-to-one correspondence, and each clustering cluster contains a fluctuation subsequence; Obtain the real-time power generation sequence and calculate the real-time fluctuation sequence, calculate the similarity between the real-time fluctuation sequence and each clustering cluster respectively, and input the real-time fluctuation sequence into each prediction model to obtain several power generation prediction values; Set up a control model, use the similarity as a weight in the control model to integrate the power generation prediction values, obtain the final prediction value, and control the solar street lamp according to the control model.
2. The method for establishing an intelligent control model of a solar street lamp according to claim 1, characterized in that The obtaining of the initial clustering clusters includes: Perform initial clustering on all elements in the fluctuation sequence to obtain initial clustering clusters, retain the initial clustering clusters that meet the preset conditions, and iteratively divide the initial clustering clusters that do not meet the preset conditions until all initial clustering clusters that meet the preset conditions are obtained. The preset condition is that the elements in the initial clustering cluster are continuously adjacent in the fluctuation sequence.
3. The method for establishing an intelligent control model of a solar street lamp according to claim 1, characterized in that The obtaining of the initial clustering clusters includes: Calculate the standard deviation of the fluctuation sequence, calculate the element difference between any two adjacent elements in the fluctuation sequence. When the element difference is not greater than one standard deviation, merge any two adjacent elements into one initial clustering cluster.
4. The method for establishing an intelligent control model of a solar street lamp according to claim 1, characterized in that The merging of the initial clustering clusters to obtain several clustering clusters includes: Calculate the difference between any two adjacent initial clustering clusters, and merge any two adjacent initial clustering clusters with a difference less than the preset difference threshold to obtain several clustering clusters.
5. The method for establishing an intelligent control model of a solar street lamp according to claim 4, characterized in that, The difference between any two adjacent initial clustering clusters includes: Use the value of dynamic time warping between any two adjacent initial clustering clusters as the difference.
6. The method for establishing an intelligent control model of a solar street lamp according to claim 4, characterized in that, The difference between any two adjacent initial clustering clusters satisfies the relational expression: s i,i+1 represents the difference between the \(i\)-th initial clustering cluster and the \((i + 1)\)-th initial clustering cluster, represents the value of the \(m\)-th element in the \(i\)-th initial clustering cluster, represents the value of the \(n\)-th element in the \((i + 1)\)-th initial clustering cluster, \(M\) represents the total number of elements in the \(i\)-th initial clustering cluster, and \(N\) represents the total number of elements in the \((i + 1)\)-th initial clustering cluster.
7. The method for establishing an intelligent control model of a solar street lamp according to claim 1, characterized in that The prediction model is a long short-term memory recurrent neural network or an ARIMA model.
8. The method for establishing an intelligent control model of a solar street lamp according to claim 1, characterized in that, The final prediction value satisfies the relational expression: CQ represents the final predicted value, ρ t represents the similarity between the real-time fluctuation sequence and the t-th clustering cluster, cq t represents the predicted power generation value of the prediction model corresponding to the t-th clustering cluster.
9. The method for establishing an intelligent control model of a solar street lamp according to claim 1, characterized in that The similarity satisfies the relational expression: ρ represents the similarity, X k represents the k-th fluctuating subsequence, Y represents the real-time fluctuation sequence, and COV represents the covariance. represents the standard deviation of the k-th fluctuating subsequence, σ Y represents the standard deviation of the real-time fluctuation sequence, and K represents the total number of fluctuating subsequences.
10. A solar street lamp, characterized in that, Control according to the control model in the method for establishing the intelligent control model of the solar street lamp according to any one of claims 1 to 9.