An apparatus of a data analysis algorithm based on intelligent road poles and a method of using the same

Abstract

The present application belongs to the technical field of intelligent road pole control, and particularly relates to a device based on a data analysis algorithm of an intelligent road pole, comprising: an algorithm platform, a network communication protocol and an operating system, the algorithm platform is provided with a plurality of network communication interfaces and is used for data interaction between field devices; the algorithm platform comprises two redundant platforms, the two redundant platforms are respectively in real-time operation state and tracking operation state, and the two redundant platforms are switched between the operation state and the tracking state through a serial communication interface; the algorithm platform communicates with an upper computer through the network communication protocol and is used for real-time data transmission, and the redundant platforms realize data backup through network communication mode. The present application adopts a multi-model algorithm, so that the controlled object has the advantages of small overshoot and short adjustment time in the control process, and the upper computer software displayed externally can be used to more intuitively and conveniently build a logic configuration, thereby improving work efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent road pole control, and in particular to a device based on a data analysis algorithm of an intelligent road pole and a use method thereof. BACKGROUND

[0002] Building a smart city has become a cross-cutting hotspot of information technology and urban development in today's world, such as using a wireless sensor network to build a smart environment monitoring system has gradually been recognized and popularized, and at present, street lamp management in the city is orderly and widely distributed, which is more suitable for building a wireless sensor network in the construction of a smart city.

[0003] The intelligent road pole control process promotes the development of complex process control with its unique complexity, which mainly manifests in nonlinearity and large time delay, and the traditional proportional integral derivative control algorithm cannot meet the control quality of the nonlinear and large delay of the controlled object in the intelligent road pole control process.

[0004] Therefore, we propose a device based on a data analysis algorithm of an intelligent road pole and a use method thereof to solve the above problems. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and propose a device based on a data analysis algorithm of an intelligent road pole and a use method thereof.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0007] A device based on a data analysis algorithm of an intelligent road pole, comprising: an algorithm platform, a network communication protocol and an operating system, the algorithm platform is provided with a plurality of network communication interfaces and is used for data interaction with field devices; the algorithm platform comprises two redundant platforms, the two redundant platforms are respectively in real-time operation state and tracking operation state, and the two redundant platforms realize switching between operation state and tracking state through a serial communication interface; the algorithm platform communicates with an upper computer through a network communication protocol and is used for real-time data transmission, the redundant platform realizes data backup through a network communication mode; the algorithm platform adopts an MQTT communication protocol to transmit between field sensors and actuators; the operating system is used for task switching, scheduling and management, a multi-model algorithm module is arranged in the algorithm platform, and the process of the multi-model algorithm module is realized by means of the operating system.

[0008] In the above-mentioned device based on a data analysis algorithm of an intelligent road pole, a plurality of multi-model algorithms can be operated in the algorithm platform, a plurality of sub-models can be stored in the algorithm, and each sub-model involves one model file.

[0009] In the aforementioned device based on the data analysis algorithm of smart road poles, the folder naming rule for each of the multi-model algorithm modules is named using the unique module identifier downloaded from the host computer software, and the folder can store 5 sub-model files, with sub-models 1 to 5 using the naming convention A1 to A5.

[0010] In the aforementioned data analysis algorithm device based on smart poles, the multi-model algorithm module also requires an additional file to record the file information of all multi-model algorithm modules in the current algorithm platform. This file information can add or remove sub-model files and distinguish the operating status and backup status between redundant devices.

[0011] A method for using a device based on a data analysis algorithm for smart road poles includes the following steps:

[0012] S1. Decompose a complex controlled object into several sub-objects through an algorithm platform, and design a controller for each sub-object.

[0013] S2. The module properties dialog box allows for setting the initial state of the multi-model algorithm, managing sub-model files, and setting parameter properties. It also allows for multiple sub-model sampling of the multi-model algorithm module with a sampling time interval of 1 second.

[0014] S3. After the sub-model sampling is completed, the controller parameters are set. The sampling time domain N should cover the entire process of the controlled object's step response; the prediction time domain P should be close to the sampling time domain; and the control time domain M should be selected within 8.

[0015] S4. Then, the outputs of each controller are weighted and summed through a weighter to obtain the control quantity acting on the actual system. Specifically, the dynamic matrix control algorithm is used to calculate the actual output and the predicted output, and the output error is fed back to the original prediction model in a weighted manner for correction.

[0016] In the above-described method of using the data analysis algorithm based on smart poles, in the multi-model algorithm, for a linear controlled object, under the condition that the initial state is stable, the sampled value A'=[a1',a2'…,a...] of its step response is obtained by applying a control increment at one end. N '] T N is called the modeling time domain. The choice of N should make the process response value close to its steady-state value, i.e., a N '≈a ∞ From the proportionality property of a linear system, its dynamic vector for a unit step response is:

[0017]

[0018] The predicted model output of the future N time instants is

[0019]

[0020] where i = 1,..., N; The predicted output of the future P time instants under the assumption that the control action remains unchanged is Y(k+i) = Y0(k+i) + AΔU (k) (4) where ΔU(k) = [Δu(k),..., Δu(k+M-1)], i = 1,..., P, at time instant k under the action of M successive control increments Δu(k),..., Δu(k+M-1)

[0021] Y(k) = Y0(k) + AΔU M (k) (3)

[0022] where

[0023] ΔU M = [Δu(k),..., Δu(k+M-1)] T

[0024] At time instant k, the desired value y r (k+i), i = 1,..., P, is given, and the quadratic performance index is taken

[0025]

[0026] where Y r (k) = [y r (k+1),..., y r (k+P)] T , Q, λ are diagonal matrices of weight coefficients;

[0027] The optimal solution of the quadratic performance index is

[0028] ΔU(k) = (A T QA+λ) -1 A T Q(Y r (k) - Y(k)) (5)

[0029] At time instant k, only the current control increment Δu(k) is implemented on the controlled object, and the actual output y(k+1) is compared with the predicted output The output error e(k+1) is fed back to the original prediction model in a weighted manner to make corrections

[0030]

[0031] Y r (k+i) = Y0(k+i) + He(k+1) (7)

[0032] In the formula, H is a correction matrix.

[0033] In the use method of the data analysis algorithm based on the intelligent road pole, the multi-model calculation method adopts a recursive Bayesian probability weight algorithm as a weighter of the multi-model algorithm, and the weight of the algorithm is calculated through the matching error of each sub-model and the object and the historical matching probability, and the conditional probability calculation formula is:

[0034]

[0035] In the formula, P j,k is the conditional probability of the matching degree of the sub-model j at the k moment; is the matching error of the sub-model j and the controlled object at the k moment; K is a recursive calculation convergence coefficient; z is the number of sub-models, and the weight calculation formula is:

[0036]

[0037] In the formula, w j,k is the weight coefficient of the sub-model j at the k moment; δ is a minimum value, when P j,k > δ, w j,k = 0; c is the value of the model number j when P j,k ≤ δ, that is, in order to ensure that the total sum of all weights is 1, when the weight when P j,k > δ is calculated, the probability value when P j,k ≤ δ is not included, and the weight coefficient and the output of the control quantity of each sub-model are calculated in a weighted sum to obtain the control quantity actually acting on the controlled object:

[0038]

[0039] Compared with the prior art, the device and the use method of the data analysis algorithm based on the intelligent road pole have the advantages that:

[0040] The multi-model algorithm adopted by the present application enables the controlled object to have the advantages of small overshoot and short adjustment time in the control process, and the upper computer software displayed externally can be more intuitive and convenient to build a logic configuration, thereby improving work efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0041] Fig. 1 The device architecture block diagram of the device and the use method of the data analysis algorithm based on the intelligent road pole are provided by the present application;

[0042] Fig. 2 The dynamic matrix block diagram of the device and the use method of the data analysis algorithm based on the intelligent road pole are provided by the present application;

[0043] Fig. 3A method step diagram of a device and a method for using the same based on a data analysis algorithm of a smart pole is provided. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all.

[0045] EMBODIMENT

[0046] REFERENCE Figs. 1-3 A device based on a data analysis algorithm of a smart pole, comprising: an algorithm platform, a network communication protocol and an operating system, the algorithm platform is provided with a plurality of network communication interfaces and is used for data interaction between field devices;

[0047] The algorithm platform comprises two redundant platforms, the two redundant platforms are in real-time operation state and tracking operation state respectively, the two redundant platforms are switched between operation state and tracking state through a serial communication interface, a plurality of multi-model algorithms can be operated in the algorithm platform, a plurality of sub-models can be stored in the algorithm, each sub-model involves one model file, specifically, the folder naming rule of each multi-model algorithm module is named by a unique module identifier downloaded by the upper computer software, and five sub-model files can be stored in the folder, sub-models 1-5 adopt the naming mode of A1-A5,

[0048] Specifically, an additional file is needed in the multi-model algorithm module to record the file information of all multi-model algorithm modules in the current algorithm platform, and the file information can perform addition and subtraction operations on the sub-model files and distinguish the operation state and backup state between the redundant devices.

[0049] Further, the algorithm platform communicates with the upper computer through the network communication protocol and is used for real-time data transmission, the redundant platform adopts a network communication mode to realize data backup; the algorithm platform and the field sensors and actuators adopt the MQTT communication protocol for transmission.

[0050] The operating system is used for switching, scheduling and managing tasks, the algorithm platform is provided with a multi-model algorithm module, and the process of the multi-model algorithm module is realized by means of the operating system, the use of the multi-model algorithm makes the controlled object have the advantages of small overshoot and short adjustment time in the control process, and the upper computer software displayed externally can be used to more intuitively and conveniently build a logic configuration, thereby improving work efficiency.

[0051] A method for using a device based on a data analysis algorithm of a smart pole, comprising the following steps:

[0052] S1, a complex controlled object is decomposed into several sub-objects through an algorithm platform, and a controller is designed for each sub-object;

[0053] S2, initial state setting, sub-model file management and parameter attribute setting are performed on the multi-model algorithm through a module attribute dialog box, and the multi-model algorithm module is sampled for multiple times, and the sampling time interval is 1s;

[0054] S3, after the sub-model sampling is completed, the controller parameters are set, the sampling time domain N should cover the entire process of the step response of the controlled object; the prediction time domain P should be close to the sampling time domain; the control time domain M is selected to be within 8;

[0055] S4, then the outputs of the controllers are weighted and summed through a weighter to obtain the control amount acting on the actual system, and the actual output and the predicted output are calculated by using the dynamic matrix control algorithm, and the output error is fed back to the original prediction model in a weighted manner for correction.

[0056] Among them, for the linear controlled object in the multi-model algorithm, under the condition that the initial state is a stable state, the sampling value A' = [a1', a2',..., aN'] of the step response is obtained by applying a control increment at one end of the object. N ' T N is called the modeling time domain, and the selection of N should make the process response value close to its steady-state value, that is, a N ' ≈a ∞ According to the proportional property of the linear system, the unit step response dynamic vector is

[0057]

[0058] When a control increment Δu(k) is applied at one end of the system at time k, the future N-time output of the prediction model is

[0059]

[0060] In the formula: i = 1,..., N; is the future output prediction value when the control action is assumed to remain unchanged, under the action of M consecutive control increments Δu(k),..., Δu(k+M-1) at time k, the future P-time output prediction value is

[0061] Y(k) = Y0(k) + AΔU M (k) (3)

[0062] In the formula:

[0063] ΔU M = [Δu(k),..., Δu(k+M-1)]T

[0064] At k moment, given the desired value y r (k+i), i=1, …, P, take quadratic form optimization performance index

[0065] In the formula: Y r (k) = [y r (k+1), …, y r (k+P)] T , Q, λ is the weight coefficient of the diagonal matrix;

[0066] Quadratic form optimization performance index optimal solution is

[0067] ΔU(k) = (A T QA+λ) -1 A T Q(Y r (k)-Y(k)) (5)

[0068] At k moment, only the current control increment Δu(k) is implemented in the controlled object, and the actual output y(k+1) is compared with the predicted output The output error e(k+1) is fed back to the original prediction model in a weighted manner to correct it

[0069]

[0070] Y r (k+i) = Y0(k+i) + He(k+1) (7)

[0071] In the formula: H is the correction matrix.

[0072] Further, the multi-model calculation method uses the recursive Bayesian probability weight algorithm as the weighter of the multi-model algorithm, and the weight of the algorithm is calculated through the matching error of the current each sub-model and the object and the historical matching probability, and the conditional probability calculation formula is:

[0073]

[0074] In the formula: P j,k is the conditional probability of the matching degree of the sub-model j at k moment; is the matching error of the sub-model j and the controlled object at k moment; K is the convergence coefficient of recursive calculation; z is the number of sub-models, and the weight calculation formula is:

[0075]

[0076] In the formula: w j,k is the weight coefficient of the sub-model j at k moment; δ is the minimum value, and P j,k> δ, make w j,k = 0; c is P j,k ≤ δ, the value of model number j, that is, in order to ensure that the total ownership value is 1, in the calculation of P j,k > δ, the weight is not included in the calculation of P j,k ≤ δ, the probability value, the weight coefficient and the output of each sub-model control quantity are calculated by a weighted sum to obtain the control quantity actually acting on the controlled object:

[0077]

[0078] In the algorithm implementation process, a large number of matrix operations are used, so the general matrix operation is written in the form of a function library for convenient calling. In addition, the storage space of the embedded platform is limited compared with large-scale matrix operation, so special attention should be paid to the dynamic opening and release of the memory space during a large number of matrix operations.

[0079] The above is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can make equivalent replacements or changes according to the technical scheme and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A device based on a data analysis algorithm for smart road poles, characterized in that, include: An algorithm platform, which is equipped with multiple network communication interfaces for data interaction with field devices; The algorithm platform includes two redundant platforms, which are respectively in real-time computing state and tracking running state. The two redundant platforms switch between running state and tracking state through a serial communication interface. The algorithm platform communicates with the host computer via a network communication protocol for real-time data transmission. The redundant platform uses network communication to back up data. The algorithm platform communicates with field sensors and actuators using the MQTT communication protocol. An operating system is provided for switching, scheduling and managing tasks. The algorithm platform is equipped with a multi-model algorithm module, and the process of the multi-model algorithm module is implemented with the help of the operating system. In the multi-model algorithm, for a linear accused object, under the condition that the initial state is stable, the sampled value A'=[a1',a2'…,a...] of its step response is obtained by applying a control increment at one end. N '] T N is called the modeling time domain. The choice of N should make the process response value close to its steady-state value, i.e., a N '≈a ∞ From the proportionality property of a linear system, its dynamic vector for a unit step response is: When the time control increment Δu(k) is applied at one end of the system at time k, the output of the prediction model for the next N times is: In the formula: i = 1, ..., N; Assuming the control action remains constant, what are the predicted output values ​​for the next P times under the action of M consecutive control increments Δu(k), ..., Δu(k+M-1) at time k? Y(k)=Y0(k)+AΔU M (k) (3) In the formula: ΔU M =[Δu(k),…,Δu(k+M-1)] T At time k, given the expected value y r (k+i), i=1,…,P,take the quadratic optimization performance index In the formula: Y r (k)=[y r (k+1), ..., y r (k+P)] T Q and λ are diagonal matrices composed of weight coefficients; The optimal solution for the quadratic optimization performance index is ΔU(k)=(A T QA+λ) -1 AND T Q(Y r (k)-Y(k)) (5) At time k, only the current control increment Δu(k) is applied to the controlled object, and the actual output y(k+1) is compared with the predicted output. The output error e(k+1) is fed back to the original prediction model in a weighted manner for correction. Y r (k+i)=Y0(k+i)+He(k+1) (7) In the formula: H is the correction matrix; The multi-model calculation method employs a recursive Bayesian probability weighting algorithm as the weighter for the multi-model algorithm. The weights are calculated using the matching errors between each sub-model and the object, as well as historical matching probabilities. The conditional probability calculation formula is as follows: In the formula: P j,k Let be the conditional probability of the matching degree of submodel j at time k; The value of z is the matching error between sub-model j and the controlled object at time k; K is the convergence coefficient calculated recursively; z is the number of sub-models, and the weight calculation formula is: In the formula: w j,k δ is the weight coefficient of submodel j at time k; δ is the minimum value, when P j,k When >δ, make w j,k =0; c is P j,k The value of model number j when ≤δ, that is, in order to ensure that the sum of all values ​​is 1, when calculating P j,k When the weight is greater than δ, P is not included. j,k The probability value ≤δ, the weighting coefficients, and the outputs of the control quantities of each sub-model are calculated by weighted summation to obtain the actual control quantity acting on the controlled object:

2. The device based on a data analysis algorithm for smart road poles according to claim 1, characterized in that, The algorithm platform can run multiple multi-model algorithms, and the algorithm can store multiple sub-models, each of which involves one model file.

3. The device based on a data analysis algorithm for smart road poles according to claim 2, characterized in that, Each of the multi-model algorithm modules is named using a unique module identifier downloaded from the host computer software, and each folder can contain 5 sub-model files, with sub-models 1 to 5 named A1 to A5.

4. The device based on a data analysis algorithm for smart road poles according to claim 3, characterized in that, The multi-model algorithm module also requires an additional file to record the file information of all multi-model algorithm modules in the current algorithm platform. This file information can add or remove sub-model files and distinguish the running status and backup status between redundant devices.

5. A method of using a device based on a data analysis algorithm for smart road poles according to any one of claims 1-4, characterized in that, Includes the following steps: S1. Decompose a complex controlled object into several sub-objects through an algorithm platform, and design a controller for each sub-object. S2. The module properties dialog box allows for setting the initial state of the multi-model algorithm, managing sub-model files, and setting parameter properties. It also allows for multiple sub-model sampling of the multi-model algorithm module with a sampling time interval of 1 second. S3. After the sub-model sampling is completed, the controller parameters are set. The sampling time domain N should cover the entire process of the step response of the controlled object. The prediction time domain P should be close to the sampling time domain; the control time domain M should be selected within 8. S4. Then, the outputs of each controller are weighted and summed through a weighter to obtain the control quantity acting on the actual system. Specifically, the dynamic matrix control algorithm is used to calculate the actual output and the predicted output, and the output error is fed back to the original prediction model in a weighted manner for correction.

6. The method of using a device based on a data analysis algorithm for smart road poles according to claim 5, characterized in that, In the multi-model algorithm, for a linear accused object, under the condition that the initial state is stable, the sampled value A'=[a1',a2'…,a...] of its step response is obtained by applying a control increment at one end. N '] T N is called the modeling time domain. The choice of N should make the process response value close to its steady-state value, i.e., a N '≈a ∞ From the proportionality property of a linear system, its dynamic vector for a unit step response is: When the time control increment Δu(k) is applied at one end of the system at time k, the output of the prediction model for the next N times is: In the formula: i = 1, ..., N; Assuming the control action remains constant, what are the predicted output values ​​for the next P times under the action of M consecutive control increments Δu(k), ..., Δu(k+M-1) at time k? Y(k)=Y0(k)+AΔU M (k) (3) In the formula: ΔU M =[Δu(k),…,Δu(k+M-1)] T At time k, given the expected value y r (k+i), i=1,…,P,take the quadratic optimization performance index In the formula: Y r (k)=[y r (k+1), ..., y r (k+P)] T Q and λ are diagonal matrices composed of weight coefficients; The optimal solution for the quadratic optimization performance index is ΔU(k)=(A T QA+λ) -1 AND T Q(Y r (k)-Y(k)) (5) At time k, only the current control increment Δu(k) is applied to the controlled object, and the actual output y(k+1) is compared with the predicted output. The output error e(k+1) is fed back to the original prediction model in a weighted manner for correction. Y r (k+i)=Y0(k+i)+He(k+1) (7) In the formula: H is the correction matrix.

7. The method of using a device based on a data analysis algorithm for smart road poles according to claim 5, characterized in that, The multi-model computation method employs a recursive Bayesian probability weighting algorithm as the weighter for the multi-model algorithm. The weights are calculated using the matching errors between each sub-model and the object, as well as historical matching probabilities. The conditional probability calculation formula is as follows: In the formula: P j,k Let be the conditional probability of the matching degree of submodel j at time k; The value of z is the matching error between sub-model j and the controlled object at time k; K is the convergence coefficient calculated recursively; z is the number of sub-models, and the weight calculation formula is: In the formula: w j,k δ is the weight coefficient of submodel j at time k; δ is the minimum value, when P j,k When >δ, make w j,k =0; c is P j,k The value of model number j when ≤δ, that is, in order to ensure that the sum of all values ​​is 1, when calculating P j,k When the weight is greater than δ, P is not included. j,k The probability value ≤δ, the weighting coefficients, and the outputs of the control quantities of each sub-model are calculated by weighted summation to obtain the actual control quantity acting on the controlled object:

Images