CN113099408A - Simulation-based data mechanism dual-drive sensor node deployment method and system - Google Patents

Abstract

The invention discloses a simulation-based data mechanism dual-drive sensor node deployment method and system, which take the sensor deployment problem of an actual building as a research background, consider multiple factors influencing building environment sensing, such as uncertainty of personnel behaviors in the building, incommunity of space interaction, variability of outdoor weather and the like, and efficiently search an optimal deployment strategy while effectively reducing the size of a feasible region by using a sequence optimization method in combination with data mechanism dual-drive. The invention fully considers the cost and the sensing precision of the sensor at the same time, and finally achieves the optimal balance of the precision and the cost to meet the problems of sensor deployment in practical buildings.

Description

Simulation-based data mechanism dual-driven sensor node deployment method and system

technical field

The invention belongs to the technical field of sensor network node deployment and building energy saving, and in particular relates to a method and system for deploying sensor nodes based on a simulation-based data mechanism and dual drives.

Background technique

In recent years, with the continuous development of society, the overall energy consumption in the world has been increasing year by year, which not only brings about the problem of energy shortage, but also causes a series of environmental problems. The focus and innovation frontier of foreign industry and academia. The World Council for Sustainable Development pointed out that in many countries, building energy consumption accounts for at least 40% of the total social energy consumption, and nearly 40% of the building energy consumption is related to heating, ventilation and air conditioning systems (HVAC). Therefore, designing intelligent building systems that optimize HVAC operation to reduce building energy consumption is of great significance to alleviating the world's energy and environmental problems.

In order to realize the optimization of building operation, it is necessary to use sensor network to collect building environment information. And the reliability and effectiveness of the sensor is the guarantee for the stable operation of the intelligent building system. However, the sensing accuracy of the sensor is affected by multiple factors, and low-precision sensing will affect the control and decision-making of the building system, which will lead to problems such as instability and high energy consumption of the building system.

Sensor networks are widely used in various fields. However, in the architectural field, traditional sensor deployment methods are difficult to generate strategies to improve the observation performance of building environmental information due to the high uncertainty of human behavior and the low interaction between scattered rooms. At the same time, traditional sensor deployment schemes also fail to fully consider the need to protect personnel privacy and reduce economic costs. Therefore, there is a need to improve the node deployment method for building sensor networks.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present invention provides a method and system for the deployment of building sensor nodes based on simulation and data mechanism, which can take into account the requirements of sensing accuracy and economy at the same time, and find an optimal sensor deployment solution that achieves a balance between the two.

In order to achieve the above object, the present invention adopts the following technical scheme: a sensor node deployment method based on a simulated data mechanism dual drive, comprising the following steps:

S1. Divide the building into M sub-regions according to the distribution of the actual building space, and determine the number of sensors k _i to be deployed, the value range k _i ∈[a,b] and the value step size according to the economic cost budget c;

S2. Under the condition that the number of sensors is k _i , generate the feasible region of the sensor node deployment strategy of size N from the feasible region Ω of all deployment schemes

S3. Determine an objective function according to the number k _i of the initialized sensors determined in S1, and the objective function minimizes the sensing error of the sensor network node deployment strategy;

S4. According to the environmental information and meteorological information obtained from the actual building, use the building energy consumption simulation engine EnergyPlus to perform mechanism-driven simulation to obtain simulation-based building environment data. The simulation-based building environment data includes the real value of environmental distribution and k environmental data under _i sensor conditions;

中的最优解；S5. Input the simulation-based building environment data obtained in S4 into the neural network for training and prediction, and combine the sequential optimization algorithm S3 to solve the objective function, and output the feasible region of the sensor node deployment strategy

the optimal solution in ;

S6, store the optimal solution output in S5, let k _i =k _i +c;

S7, let i=i+1, and judge whether k _i >b is established;

S8. If k _i >b does not hold, repeat steps S2 to S7 until k _i >b in S7, make decisions through data analysis, and output a sensor network node deployment strategy that integrates the optimal economic cost and sensing accuracy.

Further, in S4, the process of utilizing the building energy consumption simulation engine EnergyPlus to obtain simulation-based building environment data includes the following steps:

S401. Acquire building surrounding environment, weather conditions, building envelope information, building load information, and HVAC control information, and input the building energy consumption simulation engine EnergyPlus;

S402. Define the output as temperature and relative humidity through the building energy simulation engine EnergyPlus output module, and obtain simulation-based building environment data, where the simulation-based building environment data includes the real value of the building environment distribution and the buildings collected by k _i sensors environmental data.

Further, S5 includes the following steps:

S501. Process the real value of the simulation-based building environment distribution obtained in S4 into a matrix of size M×x according to the divided M building sub-regions, where x represents the dimension, and the matrix is input into the fully connected BP neural network as a training set , to train to get the building environment distribution model;

S502, processing the building environment data obtained in S4 based on the simulation-based k _i sensor conditions into a matrix of size M×x, and inputting the matrix as a test set into the building environment distribution model obtained in S501 for prediction, Obtain the environmental distribution output under the condition of k _i sensors, that is, the predicted value of the building environment distribution;

S503, according to the predicted value of the building environment distribution output by the neural network in S502 and the actual value of the environment distribution based on simulation under the corresponding deployment strategy, draw a sequential optimization performance curve;

S504, according to the type of the sequence optimization performance curve obtained in S503, determine the feasible solution set S of the sequence optimization rough selection model;

S505. Based on the feasible solution set S of the rough selection model obtained in S504, calculate the root mean square error of each predicted value of the environmental distribution in the feasible solution set S and the simulation-based building environment data obtained in S4, and calculate the error from the smallest To the largest order, the feasible solution with the smallest error is determined as the optimal sensor network stage deployment scheme under the condition of k _i sensors.

Further, in S502, the predicted value of the environment distribution represented by each node of the output layer of the neural network is a P-dimensional vector.

Further, in S503, the step of drawing the sequence optimization performance curve includes:

S5031, calculate respectively the root mean square error between the environment distribution predicted value of neural network output in S502 and the actual value of environment distribution based on simulation under the deployment strategy corresponding to it;

S5032: Sort the errors obtained in S5031 in ascending order, take the serial number of the feasible solution as the abscissa, take the error corresponding to the feasible solution as the ordinate, and draw a curve graph is the sequential optimization performance curve.

Further, the step of obtaining the feasible solution set S of the sequential optimization and rough selection described in S504 includes:

S5041. Determine the optimization problem type according to the type of the sequential optimization performance curve;

从而得到可 行解集S即由可行域中误差最小的前s项构成，其中k和g是能以比较高的概率p包含不少于 k个较优解的自定义的参数，Z₀、ρ、γ和η是根据该文献中的非线性回归表得到的参数，e是 自然常数。S5042. Determine the feasible solution set size s from the optimization problem type obtained in S5041,

Therefore, the feasible solution set S is formed by the first s items with the smallest error in the feasible domain, where k and g are self-defined parameters that can contain no less than k better solutions with a relatively high probability p, Z ₀ , ρ , γ and η are parameters obtained according to the nonlinear regression table in the literature, and e is a natural constant.

Further, in S3, the objective function is:

是对于给定k_i个传感器条件下的传感器部署方案的可行域，j表示其中第j个 子区域(j∈[1,M])，

表示k_i个传感器条件下第j个子区域的环境估计误差的评估函数，

in,

is the feasible region of the sensor deployment scheme for a given k _i sensor condition, where j denotes the jth sub-region (j∈[1,M]),

represents the evaluation function of the environmental estimation error of the jth sub-region under k _i sensor conditions,

是第j个区域的R_j个房间下的基于机理仿真的整体环境值矩阵，

表示第j 个区域的R_j个房间下的基于数据驱动的环境预测值矩阵，

是基于机理仿真的整体 环境值矩阵和基于数据驱动的环境预测值矩阵之间的欧氏距离。in,

is the overall environment value matrix based on the mechanism simulation under the R _j rooms in the j th area,

represents the matrix of data-driven environmental predictions under R _j rooms in the j th area,

is the Euclidean distance between the overall environmental value matrix based on the mechanism simulation and the data-driven environmental prediction value matrix.

Further, the simulation-based building environment data in S4 is 2-dimensional, which are ambient temperature and relative humidity.

A simulation-based data mechanism dual-driven sensor node deployment system includes a strategy generation module, a mechanism-driven module, a data-driven module, and an analysis and decision-making module;

与Ω；The strategy generation module inputs the initial number of sensors k _i and the initial feasible region size N according to the actual building structure distribution and sensor cost budget, and is used to generate the size of k _i sensors under budget conditions and without considering the sensor budget cost. is the feasible region of N

with Ω;

The mechanism-driven module collects various relevant parameters of the actual building, and inputs it into the building simulation engine _EnergyPlus to perform mechanism-based modeling and simulation. building environment data, and delivering the building environment data to a data-driven module;

的最优解，并输入到分析决策模块；The data-driven module includes an optimal sensor node deployment strategy for generating k _i sensor budgets through sequential optimization combined with a neural network, that is,

The optimal solution is input to the analysis decision module;

的最优解，并进行综合分 析，最终输出经济成本和传感精度综合最优的传感器部署策略。The analysis and decision-making module is used to store the data generated by the optimal solution generation sub-module every time.

The optimal solution is obtained, and comprehensive analysis is carried out to finally output the comprehensive optimal sensor deployment strategy for economic cost and sensing accuracy.

A dual-driven sensor node deployment system based on a simulation data mechanism, comprising a memory and a processor, the memory stores a computer program that can run on the processor, and when the processor executes the computer program, the right The steps of the above method are required.

Compared with the prior art, the present invention has at least the following beneficial technical effects:

The sensor deployment method of the present invention takes the sensor deployment problem of the actual building as the research background, and takes into account the uncertainty of the behavior of people in the building, the disconnection of spatial interaction, the variability of outdoor weather, and other factors that affect the sensing of the building environment. , using the sequential optimization method combined with the data mechanism dual drive to effectively reduce the size of the feasible region and efficiently find the optimal deployment strategy. On the one hand, the building simulation engine is used to simulate the mechanism-driven building environment. The experiment cost is greatly reduced; on the other hand, the data-driven order optimization based rough selection feasible solution set is generated through the neural network, and the efficiency of the optimal solution search is improved. The size of the original sample space is too large, and other optimization algorithms will lead to long solution time and dimensional disaster, while sequential optimization can avoid the above shortcomings, thereby improving efficiency and greatly improving the efficiency of generating optimal sensor deployment strategies.

Further, the present invention aims to minimize the sensing error of the sensor network node deployment strategy, calculates the optimal deployment strategy under different sensor numbers, makes decisions through data analysis, and finally obtains the comprehensive optimal economic cost and sensing accuracy. At the same time, the sensor cost and sensing accuracy are fully considered, and finally the optimal trade-off between accuracy and cost is achieved to meet the problems faced by sensor deployment in actual buildings, and then the energy saving and emission reduction and optimal control of building systems are achieved. produce a positive effect.

The system of the present invention includes a strategy generation module, a mechanism driving module, a data driving module and an analysis decision module, and can obtain an optimal sensor deployment scheme considering both sensing accuracy and economic requirements.

Description of drawings

1 is a schematic flowchart of a method for deploying sensor nodes based on a simulation-based dual-driven sensor node provided by the present invention;

2 is a structural block diagram of a sensor node deployment system based on a simulation-based dual-drive sensor node provided by the present invention;

Fig. 3 is a kind of structure diagram of the neural network sub-module of the sensor node deployment system based on the simulation data mechanism dual drive provided by the present invention;

Figure 4 shows the relationship curve between the sensing accuracy and sensing cost of an office building in Xi'an based on the objective function value and constraints of the above scheme;

Figure 5 is a schematic diagram of the optimal deployment strategy of sensors based on the above scheme in an office building in Xi'an, (a) is a schematic diagram of the optimal deployment strategy of sensors on the first floor, (b) is a schematic diagram of the optimal deployment strategy of sensors on the first floor;

FIG. 6 is a schematic diagram of a sensor node deployment system based on a simulation-based data mechanism dual-driven.

Detailed ways

The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Fig. 1 is the schematic flow chart of the method of the present invention, as shown in the figure, a kind of sensor node deployment method based on simulation data mechanism dual drive provided by the present invention comprises the following steps:

S1. Divide the building into M sub-regions according to the distribution of the actual building space, and determine the value range k _i ∈ [a, b] and the value step c of the number of sensors to be deployed according to the economic cost budget.

In this embodiment, the actual building space location distribution includes floor distribution, orientation distribution, solar radiation, and ventilation of rooms in the building. For example, all rooms in an office building on the second floor with windows facing north can be divided into one Zone, all rooms on the second floor with windows facing south can be divided into another zone; similarly, all rooms on the first floor with windows facing north can also be divided into one zone, and so on. In addition, the sensors installed in each sub-zone The number is less than or equal to the number of points that can be installed with sensors in this area. The economic cost budget refers to the cost of purchasing and installing the sensor. The value range of the number of sensors k _i ∈[a,b] and the value step c, where a<b, c≤ba and a∈N ^* ,b∈N ^* ,c∈N ^* , _ki ∈N ^* , N ^* is the set of positive integers.

其中可行域Ω是所有可能的部 署方案的全集，即可行解全集，理论上Ω包含无穷个可行的部署方案。S2. Under the condition that the number of sensors is k _i , the feasible region of the sensor node deployment strategy of size N, N is a positive integer, is generated from the feasible region Ω of all deployment schemes by random sampling.

The feasible region Ω is the complete set of all possible deployment schemes, that is, the complete set of feasible solutions. In theory, Ω contains infinite feasible deployment schemes.

There are many methods for reducing the feasible region in this step, and a random sampling method is adopted in this embodiment.

S3. Determine an objective function and constraint conditions according to the number of sensors k _i , and the objective function minimizes the sensing error of the node deployment strategy of the sensor network.

In this step, the objective function is:

表示k_i个传感器条件下第j个子区域的环境估计误差的评估函数，

Among them, M is the division of the building into M sub-areas in S1 according to the spatial distribution of the building, and j represents the j-th sub-area (j∈[1,M]),

represents the evaluation function of the environmental estimation error of the jth sub-region under k _i sensor conditions,

表示第j个子 区域中的R_j个房间下的基于机理仿真的整体环境值矩阵，通过EnergyPlus仿真得到的，

表 示第j个子区域的R_j个房间下的基于数据驱动的环境预测值矩阵，

是两者之间的 欧氏距离。Among them, j represents the j-th sub-area, and R _j is the number of rooms contained in the j-th sub-area, then

represents the overall environment value matrix based on the mechanism simulation under the R _j rooms in the jth sub-region, obtained through EnergyPlus simulation,

represents the matrix of data-driven environmental predictions under R _j rooms in the j th subregion,

is the Euclidean distance between the two.

其中Ω是不限制传感器个数的情况下所有传感器部署方案的可 行域，

是k_i个传感器条件下的传感器部署方案的可行域。The constraints are:

where Ω is the feasible region of all sensor deployment schemes without limiting the number of sensors,

is the feasible region of the sensor deployment scheme under the condition of k _i sensors.

S4. Input EnergyPlus according to the parameters obtained through experiments and investigations in the actual building, use the building energy consumption simulation engine EnergyPlus to perform mechanism-driven simulation, and obtain simulation-based building environment data. The simulation-based building environment data includes the real value of environmental distribution and the building environment data collected by k _i sensors, the real value of the environment distribution is the real environment data of the building itself, including temperature and relative humidity.

In this embodiment, the parameters input to EnergyPlus include: building surrounding environment information, weather information, building envelope, building geometry, cooling and heating load information, HVAC information and control flow information;

Information on the surrounding environment of the building includes: the distribution of surrounding buildings, the location of the building, etc.;

Meteorological information includes: outdoor temperature, outdoor relative humidity, outdoor wind speed, etc. where the building is located;

HVAC information includes: the distribution of HVAC and fresh air equipment in the building;

Control flow information includes building loads and control operation strategies for air conditioning equipment.

In this embodiment, indoor temperature and outdoor relative humidity are mainly considered, so the generated building environment data based on simulation is 2-dimensional, which are ambient temperature and relative humidity respectively; in addition, as long as the corresponding parameters can be input into EnergyPlus, the environment distribution is real Values and environmental data under k _i sensor conditions can be simulated by EnergyPlus.

中的最优策略解，具体包括以下 步骤：S5. Process the simulation-based building environment data obtained in S4, input the neural network for training and prediction, and combine the sequential optimization method to perform data-driven optimization selection of feasible solutions, and output

The optimal policy solution in , specifically includes the following steps:

S501. Process the M building sub-regions divided according to S1 to the real value of the simulation-based building environment distribution obtained in S4 into a matrix with a size of M×2 using the pandas library of python, where 2 represents two characteristics of temperature and humidity, and Input it as a training set into a fully connected BP neural network for training to obtain a building environment distribution model;

In this embodiment, the number of nodes in the input layer of the neural network is the same as the number of sub-areas divided by the building, which is M, and each input node is 2-dimensional, representing the environmental distribution of each sub-area, including ambient temperature and relative humidity; The number of nodes in the output layer of the neural network is the same as that of the input layer, which is also M. Each node represents the environmental distribution of each sub-region. Specifically, the temperature and humidity are used as a two-dimensional coordinate system. The temperature and humidity information of the node position of the sensor is drawn in this coordinate system, and the coordinate system is divided into P equal parts according to the value range of temperature and humidity, and the percentage of nodes in each equal part to all nodes is calculated, that is, each output layer. The nodes are all P-dimensional vectors.

In this embodiment, various parameters of the neural network are trained and adjusted through the training set, and each parameter includes the number of layers of the neural network, the number of neurons in each layer, the selection of the loss function, and the selection of the gradient descent method. , regularization parameters, choice of excitation function, weights and biases, learning rate, mini-batch, etc. Finally, a neural network model with good performance is obtained by training, and the indicators for evaluating the performance of neural network include recall rate, accuracy rate, ROC and so on. The neural network model belongs to the prediction regression model, which reflects the mapping relationship between the temperature and humidity values of each sub-region in the building and the corresponding environmental distribution.

S502: Process the building environment data obtained in S4 based on the simulation-based k _i sensors into a matrix of size M×2, where 2 represents two features of temperature and humidity; and input the neural network trained in S501 to predict regression The model is used for prediction, and the environmental distribution output under the condition of k _i sensors is obtained, that is, the predicted value of the building environment distribution.

In this actual step, the environmental data obtained in S4 under the condition of k _i sensors based on the simulation is processed into the same structure as the input of the training set in S501 by using the same processing method, that is, the test set is also M input nodes, and each Each node includes two dimensions of temperature and humidity; the output of the neural network model is the predicted value of the environmental distribution of M sub-regions under the conditions of k _i sensors corresponding to the input.

S503, according to the environment distribution predicted value of the neural network prediction regression output in S502 and the real value of the environment distribution based on simulation under the corresponding sensor deployment strategy, draw the sequential optimization performance curve;

In this step, the step of drawing the sequence optimization performance curve includes:

S5031, calculate respectively the root mean square error between the environmental distribution predicted value obtained by the neural network model training in S502 and the actual value of the environmental distribution obtained by EnergyPlus simulation under the corresponding sensor deployment strategy;

中的N个可行解的序号作为横坐标(序号从1到N)，将与该 可行解对应的误差作为纵坐标，绘制曲线图得到序优化性能曲线；S5032 , sort the root mean square errors obtained in S5031 in ascending order, and randomly sample the feasible regions obtained from the complete set of feasible regions Ω in S2

The sequence number of the N feasible solutions in the abscissa is used as the abscissa (the sequence number is from 1 to N), and the error corresponding to the feasible solution is used as the ordinate, and a graph is drawn to obtain the sequence optimization performance curve;

S504, consult relevant data and literature, how to determine the type of the order optimization performance curve obtained in S503 by Professor Yu Qi's paper "Universal alignmentprobabilities and subset selection for ordinal optimization", thereby determining the feasible solution set S of the order optimization rough selection model;

In this step, the types of the sequence optimization performance curve include: steep type, general type, neutral type, and gentle type.

In this embodiment, the step of obtaining the feasible solution set S of the sequence optimization and rough selection includes:

S5041. Determine the type of the sequential optimization performance curve to determine the optimization problem type;

其中k和g 是能以比较高的概率p包含不少于k个较优解的自定义的参数，Z₀、ρ、γ和η是根据该文献中 的非线性回归表得到的参数，e是自然常数，从而得到可行解集S即由可行域中误差最小的前 s项构成。S5042. Determine the feasible solution set size s from the optimization problem type obtained in S5041,

where k and g are self-defined parameters that can contain no less than k better solutions with a relatively high probability p, Z ₀ , ρ, γ and η are parameters obtained according to the nonlinear regression table in this document, e is a natural constant, so the feasible solution set S is composed of the first s items with the smallest error in the feasible region.

S505. Based on the rough-selected feasible solution set S obtained in S504, calculate the difference between each predicted value of the building environment distribution output by the neural network model in the rough-selected feasible solution set S and the simulation-based overall environment distribution data obtained in S4. The root mean square error is calculated, and the errors are sorted from small to large, and the feasible solution with the smallest error is determined as the optimal sensor network stage deployment scheme under the condition of k _i sensors.

S6. Store the optimal policy solution output in S5, and increase the number of sensors with a step c, that is, k _i = _ki +c;

S7, set i=i+1 and judge whether the current _ki is greater than the upper limit of the budget cost, that is, judge that _ki >b;

S8. If the judgment result of S7 is "No", repeat steps S2-S7. Until the judgment result of S7 is "Yes", the decision is made through data analysis, and the comprehensive optimal sensor network node deployment strategy of economic cost and sensing accuracy is output. The data analysis process is as follows: draw the relationship curve between cost _ki and error, and use the Pareto analysis method in data analysis to obtain the optimal deployment strategy.

Referring to FIG. 2 , it is a schematic diagram of a system provided by the present invention: a dual-driven sensor node deployment system based on simulation data mechanism includes a strategy generation module 101 , a mechanism driving module 102 , a data driving module 103 and an analysis decision module 104 ;

和不考虑传感器预算成本情况下 的大小为N的可行域Ωⁱ；The strategy generation module 101 initially inputs the initial number of sensors k ₁ according to the actual building structure distribution and sensor cost budget, that is, according to the given feasible domain size N and the sensor value range k _i ∈ [a, b] and the initial feasible region size N for generating a feasible region of size N under _ki sensor budgets

and the feasible region Ω ⁱ of size N without considering the sensor budget cost;

The mechanism driving module 102 collects various related parameters of the actual building (including building surrounding environment information, weather information, building envelope, building geometry, cooling and heating load information, HVAC information and control flow information), and input it into the building simulation engine EnergyPlus Carry out mechanism-based modeling and simulation, and obtain the building environment data under the condition of simulation-based k _i sensor budget and without considering the sensor budget cost, and send them to the data-driven module;

的最优解，并输入到分析决策模块；The data-driven module 103 includes a data processing sub-module, a neural network sub-module and an optimal solution generation sub-module. The simulation-based building environment data is processed through the data processing sub-module, and the building environment data without considering the sensor budget cost is processed into a training set. The input is an M×2-dimensional matrix, where M represents the sub-area of the building, 2 represents the temperature and Humidity two-dimensional environmental index; the output of the training set is an M×P-dimensional matrix, where M represents the sub-area divided by the building, and P represents the temperature and humidity as a two-dimensional coordinate system, and each deployable sensor in this sub-area based on the simulation The temperature and humidity information of the position of the node is drawn in this coordinate system, and the coordinate system is divided into P equal parts according to the value range, and the percentage of nodes in each equal part to all nodes is calculated. Similarly, the building environment data under the simulation-based _ki sensor budget conditions are processed into the input of the test set, which is also an M×2-dimensional matrix, where M represents the sub-regions of the building, and 2 represents the two-dimensional environment of temperature and humidity. index. Input the test set into the trained neural network sub-module, and output the predicted value of the building environment distribution under the condition of _ki sensor budget of M×P dimension. Input the obtained predicted value into the optimal solution generation sub-module, and generate the optimal sensor node deployment strategy under the condition of k _i sensor budgets through the sequential optimization algorithm, namely

The optimal solution is input to the analysis decision module;

3 is a schematic diagram of the neural network structure of the neural network sub-module. where {T,H} ₁ ,{T,H} ₂ ,...,{T,H} _M indicates that the number of input layer nodes of the neural network is M, and each input layer node is 2-dimensional, respectively represent the ambient temperature T and relative humidity H of each sub-region; Y ₁ , Y ₂ ,..., Y _M represent the predicted value of the neural network model, the number of nodes in the output layer of the neural network is M, and each output Layer nodes are all P-dimensional vectors;

的最优解，将存储内容输入到数据分析子模块进行综合数 理统计分析，将分析结果输入到优化决策子模块，分析的目的是最终输出经济成本和传感精度 综合最优的传感器部署策略。The analysis and decision-making module 104 includes a storage sub-module, a data analysis sub-module and an optimization decision-making sub-module, and the storage sub-module stores the data generated by the optimal solution generation sub-module each time.

The storage content is input into the data analysis sub-module for comprehensive mathematical statistical analysis, and the analysis results are input into the optimization decision-making sub-module. The purpose of the analysis is to finally output the comprehensive optimal sensor deployment strategy for economic cost and sensing accuracy.

Figure 4 shows the relationship curve between the sensing accuracy and sensing cost of an office building in Xi'an based on the objective function value and constraints of the above scheme, where the number of sensors ranges from k _i ∈ [30, 80], and the step size c = 5. It can be seen from this relationship diagram that the optimal trade-off between sensing accuracy and cost is achieved when k _i =50. Figure 5 shows the comprehensive optimal sensor deployment strategy under the above conditions, in which four different patterns represent different building orientations, and different wireframes represent different building sub-areas. As can be seen from the figure, in this example, the building is divided into 7 sub-regions, that is, M=7. Places filled with different patterns represent room nodes where sensors are deployed in the corresponding sub-area, and the numbers in them represent the number of sensors deployed in that room node.

The embodiment of the present invention provides a sensor node deployment system based on a simulation-based dual-drive of data mechanism, which is used for implementing the above-mentioned method for deploying a sensor node based on the simulation-based dual-drive of the data mechanism. The data driving module and the analysis decision module can be divided into functional modules according to the above method examples. For example, each functional module can be divided corresponding to each function, or the two functions can be integrated into one processing module. The above-mentioned integrated modules can be realized in the form of hardware, or can be realized in the form of software function modules, such as a composition shown in FIG. 6, including a memory and a processor, and the memory is stored with a computer program that can run on the processor. , when the processor executes the computer program, it implements the steps of the above-mentioned method for deploying sensor nodes based on a simulation-based dual-driven data mechanism.

It should be noted that, the division of modules in the embodiment of the present invention is schematic, and is only a logical function division, and another division manner may be used in actual implementation.

In the above-mentioned embodiments, it may be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using a software program, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the procedures or functions according to the embodiments of the present invention result in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable system. Computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server, or data center over a wire (e.g. Coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (eg infrared, wireless, microwave, etc.) means to transmit to another website site, computer, server or data center. Computer-readable storage media can be any available media that can be accessed by a computer or data storage devices including one or more servers, data centers, etc., that can be integrated with the media. Useful media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVDs), or semiconductor media (eg, solid state disks (SSDs)), and the like.

Although the invention is described herein in conjunction with various embodiments, those skilled in the art can understand and implement the disclosure by reviewing the drawings, the disclosure, and the appended claims in practicing the claimed invention. Other variations of the embodiment. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that these measures cannot be combined to advantage.

Although the invention has been described in conjunction with specific features and embodiments thereof, it will be apparent that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention as defined by the appended claims, and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the invention. It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention. Thus, provided that these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims (10)

1. A simulation-based dual-drive sensor node deployment method based on a data mechanism is characterized by comprising the following steps:

s1, dividing the building into M sub-areas according to the distribution situation of the space position of the actual building, and determining the number k of the sensors to be deployed according to the economic cost budget_iAnd a value range k_i∈[a,b]And a value step length c;

s2, when the number of sensors is k_iUnder the condition of (2), generating a feasible domain of a sensor node deployment strategy with the size of N from the feasible domains omega of all deployment schemes

S3, number k of initialized sensors determined according to S1_iDetermining an objective function, wherein the objective function enables the sensing error of the sensor network node deployment strategy to be minimum;

s4, according to the environmental information and weather information obtained from the actual building, utilizing the building energy consumption simulation engine EnergyPlus to perform simulation based on mechanism driving, and obtaining the building environment data based on simulation, wherein the building environment data based on simulation comprises the real value of environment distribution and k_iEnvironmental data under individual sensor conditions;

s5, inputting the simulation-based building environment data obtained in S4 into a neural network for training and prediction, solving an objective function by combining an order optimization algorithm S3, and outputting a feasible domain of a sensor node deployment strategy

The optimal solution of (1);

s6, storing the optimal solution output in S5 and making k_i＝k_i+c；

S7, let i equal i +1, and judge k_i>b is true or not;

s8, if k_i>If b is not true, repeating the steps S2-S7 until k of S7_i>b, making decisions by data analysisAnd outputting a sensor network node deployment strategy with comprehensive and optimal economic cost and sensing precision.

2. The simulation-based data mechanism dual-drive sensor node deployment method of claim 1, wherein in S4, the process of obtaining simulation-based building environment data by using a building energy consumption simulation engine energy pyplus comprises the following steps:

s401, acquiring building surrounding environment, weather conditions, building envelope information, building load information and HVAC control information, and inputting a building energy consumption simulation engine EnergyPlus;

s402, defining and outputting temperature and relative humidity through an energy plus output module of a building energy consumption simulation engine to obtain simulation-based building environment data, wherein the simulation-based building environment data comprises a real value and k of building environment distribution_iBuilding environment data collected by each sensor.

3. The method for deploying the sensor nodes based on the simulation of the double driving of the data mechanism according to claim 1, wherein the step of S5 is as follows:

s501, processing the real values of the simulation-based building environment distribution obtained in the S4 into an M x matrix according to the divided M building sub-regions, wherein x represents the dimension, and the matrix is used as a training set and input into a full-connection BP neural network for training to obtain a building environment distribution model;

s502, simulating k obtained in S4_iProcessing the building environment data under the condition of each sensor into a matrix with the size of M x, inputting the matrix into the building environment distribution model obtained in the step S501 as a test set for prediction to obtain k_iOutputting environment distribution under the condition of each sensor, namely predicting values of building environment distribution;

s503, drawing a sequence optimization performance curve according to the predicted value of the building environment distribution output by the neural network in the S502 and the real value of the environment distribution based on simulation under the deployment strategy corresponding to the predicted value;

s504, determining a feasible solution set S of the sequence optimization roughing model according to the type of the sequence optimization performance curve obtained in the S503;

s505, respectively calculating the root mean square error of each environment distribution predicted value in the feasible solution set S and the simulation-based building environment data obtained in the S4 based on the feasible solution set S of the rough selection model obtained in the S504, sequencing the errors from small to large, and determining the feasible solution with the minimum error as k_iAnd (3) an optimal sensor network stage deployment scheme under the condition of each sensor.

4. The method for deploying sensor nodes according to claim 3, wherein in step S502, the predicted value of the environment distribution represented by each node in the output layer of the neural network is a P-dimensional vector.

5. The simulation-based data mechanism dual-drive sensor node deployment method of claim 3, wherein in S503, the step of drawing the sequence optimization performance curve includes:

s5031, respectively calculating a root mean square error between the environment distribution predicted value output by the neural network in the S502 and the simulation-based environment distribution true value under the deployment strategy corresponding to the environment distribution predicted value;

s5032, sequencing the errors obtained in the S5031 from small to large, taking the serial number of the feasible solution as an abscissa, taking the error corresponding to the feasible solution as an ordinate, and drawing a curve graph, namely a sequence optimization performance curve.

6. The simulation-based data mechanism dual-drive sensor node deployment method of claim 3, wherein the step of obtaining the order-optimized rough-selected feasible solution set S in S504 comprises:

s5041, determining the type of an optimization problem according to the type of the sequence optimization performance curve;

s5042, determining the size S of the feasible solution set according to the optimization problem type obtained in S5041,

so as to obtain a feasible solution set S, namely the feasible solution set S is composed of the first S items with the minimum error in a feasible domain, wherein k and g are self-defined parameters which can contain not less than k better solutions with a higher probability p, and Z₀ρ, γ, and η are parameters obtained from the nonlinear regression table in this document, and e is a natural constant.

7. The simulation-based data mechanism dual-drive sensor node deployment method of claim 1, wherein in S3, the objective function is:

wherein,

is for a given k_iThe feasible domain of the sensor deployment scenario under sensor conditions, j denotes the jth sub-domain (j ∈ [1, M) therein])，

Represents k_iAn evaluation function of the environmental estimation error of the jth sub-region under the sensor condition,

wherein,

is R of the jth region_jAn overall environment value matrix based on mechanism simulation under each room,

r representing the jth region_jA data-driven environment predictor matrix under each room,

the Euclidean distance between an overall environment value matrix based on mechanism simulation and an environment prediction value matrix based on data driving.

8. The dual-drive sensor node deployment method based on simulation data mechanism according to claim 1, wherein the simulation-based building environment data in S4 is 2-dimensional, and is respectively ambient temperature and relative humidity.

9. A simulation-based dual-drive sensor node deployment system for a data mechanism is characterized by comprising a strategy generation module, a mechanism driving module, a data driving module and an analysis decision module;

the strategy generation module inputs the initial sensor number k according to the actual building structure distribution condition and the sensor cost budget_iAnd an initial feasible domain size N for generating k_iFeasible domain of size N under individual sensor budget conditions and without consideration of sensor budget cost

And Ω;

the mechanism driving module collects various relevant parameters of an actual building, inputs the parameters into a building simulation engine EnergyPlus to carry out mechanism-based modeling simulation, and respectively obtains k based on simulation_iBuilding environment data under the condition of sensor budget and without considering sensor budget cost and transmitting the building environment data to a data driving module;

the data driven module includes logic for generating k by combining order optimization with neural network_iOptimal sensor node deployment strategy under individual sensor budget conditions, i.e.

And inputting the optimal solution to an analysis decision module;

the analysis decision module is used for storing the optimal solution generated by the optimal solution generation submodule each time

And performing comprehensive analysis, and finally outputting a sensor deployment strategy with comprehensive optimal economic cost and sensing precision.

10. A dual drive sensor node deployment system based on simulated data mechanisms, comprising a memory and a processor, wherein the memory has stored thereon a computer program operable on the processor, and wherein the processor, when executing the computer program, performs the steps of the method of any one of claims 1 to 8.

Images