Intelligent power grid parameter Internet of Things monitoring method, management system and device
By constructing a multi-layer neural network model and a power grid consumption prediction model optimized by the bat algorithm, the problems of error and interference in smart grid parameter measurement are solved, and higher accuracy and reliability of power grid parameter monitoring and management are achieved.
Patent Information
- Application Number
- CN202411107792.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-22
AI Technical Summary
The measurement of smart grid parameters is subject to problems such as large sensor accuracy errors, interference, and abnormal measured parameter values, which lead to measurement uncertainty and randomness, affecting the accuracy of IoT monitoring of grid parameters.
A multi-layer neural network model combined with the Bat algorithm for optimization is adopted, including CNN convolutional neural network, DRNN neural network, ARIMA power grid parameter model, BiLSTM neural network and wavelet decomposition model, to construct a power grid electricity consumption prediction model. By fusing data from multiple sensors and connecting models in series, the prediction accuracy and reliability are improved.
It effectively addresses the ambiguity and uncertainty of sensor measurements, improves the objectivity and reliability of power grid parameter prediction, and enables more accurate power grid parameter monitoring and optimization management.
Smart Images

Figure CN119010346A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of automated equipment for collecting and monitoring power grid parameters, and in particular to a method, management system and device for monitoring smart grid parameters in the Internet of Things. Background Art
[0002] Smart grid is the mainstream form of development of current power companies. Smart grid integrates a variety of advanced technologies such as grid information communication, grid parameter sensing and measurement, and grid intelligent control technology. In addition to ensuring the daily operation of the grid, it can also achieve tasks such as automated detection, maintenance and management of grid parameters to meet the needs of modern social development.
[0003] Smart grids now realize automated supervision and control of power grids, achieve real-time monitoring and accurate judgment of faults, solve the drawbacks of traditional models, improve the reliability of power grid operation, and optimize management levels. The integration of the Internet of Things, big data processing, and intelligent control technologies into smart grids can improve the level of power grid operation management and daily operation efficiency, conduct real-time dynamic supervision of the operation of smart grids, implement scientific adjustment and management of power consumption devices, and ensure the quality of power transmission.
[0004] Smart grid parameters are mostly measured using parameter sensors. However, during the measurement of smart grid parameters, there are often problems such as large sensor accuracy errors, interference, and abnormal measured parameter values, as well as uncertainty and randomness. How to improve the accuracy of smart grid parameter IoT monitoring has a huge impact on the realization of smart grid automation supervision and control.
[0005] The present invention can scientifically manage electricity consumption through smart grid parameter collection, big data processing and electricity consumption prediction, achieve real-time allocation and scientific supply of electric energy, save electricity while meeting daily production and living electricity consumption, and promote the sustainable development of the power industry. Summary of the invention
[0006] Purpose of the invention: In response to the problems pointed out in the background technology, the present invention discloses a smart grid parameter Internet of Things monitoring method, management system and device, which collect, analyze and predict power consumption information of the power grid, and adjust and optimize the power supply in real time according to the power demand of users in different regions, so as to ensure the optimal use of energy and realize scientific management and control of electric energy.
[0007] Technical solution: The present invention discloses a method for monitoring parameters of a smart grid through the Internet of Things, comprising the following steps:
[0008] Collect output values of multiple current sensors, multiple electric meters, and multiple groups of environmental parameter sensor groups;
[0009] Construct a power grid power consumption prediction model, including power supply detection module 1-C, FCM power grid parameter classifier, BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2-5, BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 2;
[0010] Multiple current sensor outputs are used as inputs of power supply detection module 1, multiple meter output values are used as inputs of power supply detection module 2, multiple sets of environmental parameter sensor group values are used as inputs of FCM power grid parameter classifier, and multiple types of environmental parameter sensor group values output by FCM power grid parameter classifier are used as inputs of corresponding power supply detection modules 3-C respectively. Power supply detection modules 1~C predict the output values of corresponding current sensors, meters, and multiple sets of environmental parameter sensors and output corresponding sensor prediction values;
[0011] The output of power supply detection module 1 is used as the corresponding input of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2. The time series difference between the output of power supply detection module 1 and the output of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2 and the rate of change of the time series difference are used as the corresponding input of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2. The output of power supply detection module 3-C is used as the corresponding input of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 4. The outputs of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2, power supply detection module 2 and BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 4 are used as the corresponding input of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 5. The outputs of the grid parameter model-ARIMA grid parameter model 3 and the BiLSTM neural network grid parameter model-ARIMA grid parameter model 5 are used as the corresponding inputs of the CNN convolutional neural network grid parameter model-BA's DRNN neural network grid parameter model 2 of BA. The time series difference and the rate of change of the difference between the output of the power supply detection module 2 and the output of the BiLSTM neural network grid parameter model-ARIMA grid parameter model 5, and the time series difference and the rate of change of the difference between the output of the CNN convolutional neural network grid parameter model-BA's DRNN neural network grid parameter model 2 and the output of the power supply detection module 2 are used as the corresponding inputs of the BiLSTM neural network grid parameter model-ARIMA grid parameter model 3, respectively. The three parameter interval numbers output by the CNN convolutional neural network grid parameter model-BA's DRNN neural network grid parameter model 2 of BA are used as the power consumption prediction values of the detected grid. The structure and function of the grid power consumption prediction model are shown in Figure 2 shown.
[0012] Furthermore, the power supply detection modules 1 to C have the same structure, and all include BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1, ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model, BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1, BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1-2, wavelet decomposition power grid parameter model, BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter model 1-2;
[0013] The output values of multiple current sensors or multiple electric meters or multiple groups of environmental parameter sensor groups are respectively used as the corresponding inputs of BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1, the corresponding inputs of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1 and the corresponding inputs of BA's CNN convolutional neural network power grid parameter model-BiLSTM neural network power grid parameter model 1-2, the outputs of BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1 are respectively used as the inputs of ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model, BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1 and the input of wavelet decomposition power grid parameter model, the outputs of ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model are used as the corresponding inputs of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1, BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1 and The outputs of BiLSTM neural network power parameter model-BA's DRNN neural network power parameter model 1 are used as the corresponding inputs of BA's CNN convolutional neural network power parameter model-BiLSTM neural network power parameter model 1, and the outputs of wavelet decomposition power parameter model are used as the inputs of BiLSTM neural network power parameter model-BA's DRNN neural network power parameter model 2. The outputs of BiLSTM neural network power parameter model-BA's DRNN neural network power parameter model 2 and BA's CNN convolutional neural network power parameter model-BiLSTM neural network power parameter model 1 are used as the corresponding inputs of BA's CNN neural network power parameter model-BiLSTM neural network power parameter model 2, and the outputs of BiLSTM neural network power parameter model-ARIMA power parameter model 1 and BA's CNN neural network power parameter model-BiLSTM neural network power parameter model 1-2 are three parameter intervals consisting of a, b and c, and the number is [a, b, c] as the output prediction values of multiple current sensors or multiple electric meters or multiple groups of environmental parameter sensor groups. See the structure and function of the power supply detection module. Figure 1 shown.
[0014] Further, the CNN convolutional neural network power grid parameter model of BA-the DRNN neural network power grid parameter model of BA 1-2 is the CNN neural network power grid parameter model of BA and the DRNN neural network power grid parameter model of BA connected in series;
[0015] The CNN neural network power grid parameter model of the BA replaces the connection weights and thresholds between the CNN neural network power grid parameter models with the individual position parameters of the bat population, encodes each individual in a real number encoding manner, and obtains the initial bat population. The prediction error of the CNN neural network power grid parameter model is used as the fitness function of the BA, and the fitness value is calculated and returned to the BA. The BA updates the position parameters to obtain new bat individuals, calculates new fitness function values, and repeats the update until the fitness function value drops to the minimum. At this time, the bat is the best selected individual, and the position parameter value of the bat is the initialization weight and bias of the optimal CNN neural network power grid parameter model to be found;
[0016] The BA's DRNN neural network power grid parameter model replaces the connection weights and thresholds between the DRNN neural network power grid parameter models with the individual position parameters of the bat population, encodes each individual using real number coding, and obtains the initial bat population. The DRNN neural network power grid parameter model prediction error is used as the fitness function of the BA, and the fitness value is calculated and returned to the BA. The BA updates the position parameters to obtain new bat individuals, calculates new fitness function values, and repeats the update until the fitness function value drops to the minimum. At this time, the bat is the best selected individual, and the bat's position parameter value is the initialization weight and bias of the optimal DRNN neural network power grid parameter model to be found.
[0017] Furthermore, the environmental parameter sensor group includes a temperature sensor, a humidity sensor and a wind speed sensor; they are installed at multiple points in the surrounding environment of the smart grid to be monitored; multiple current sensors and multiple electric meters are installed at multiple points in the lines of the smart grid to be detected.
[0018] The present invention also discloses a smart grid parameter Internet of Things management system, including grid parameter measurement terminals 1~n, grid control terminals 1~m, a grid cloud server, and a grid remote monitoring terminal; the grid cloud server is connected to the grid parameter measurement terminals 1~n and the grid control terminals 1~m via a grid information transmission terminal, and the grid cloud server is connected to the grid remote monitoring terminal via a 5G network; the grid parameter measurement terminals 1~n collect grid parameters and upload them to the grid cloud server via the grid information transmission terminal, and the grid cloud server is used to receive data collected by the grid parameter measurement terminals 1~n, and the grid remote monitoring terminal implements grid communication parameter setting, grid data analysis and data management, and utilizes a grid power consumption prediction model to implement the smart grid parameter Internet of Things monitoring method as described in any one of claims 1 to 4, and the grid cloud server issues control instructions based on the data processing results of the grid remote monitoring terminal; the grid control terminal receives the control instructions of the grid cloud server and controls the grid equipment to execute the instructions. See the structure and function of the smart grid parameter Internet of Things management system for details. Figure 3 shown.
[0019] Preferably, it also includes a power grid monitoring APP, and the power grid cloud server is connected to the power grid monitoring APP via a 5G network.
[0020] Preferably, the power grid parameter measuring terminals 1 to n include a temperature sensor, a humidity sensor, a wind speed sensor, a current sensor, a voltage sensor, an electric meter and corresponding signal conditioning circuits, an STM32 microprocessor, a video image sensor and a wireless communication module; the temperature sensor, the humidity sensor, the wind speed sensor, the current sensor, the voltage sensor, the electric meter and the video image sensor are all connected to the STM32 microprocessor through a signal conditioning circuit; an ad hoc network is built between the power grid parameter measuring terminals 1 to n through a wireless communication module to realize information interaction with the power grid cloud server.
[0021] Preferably, the power grid control terminals 1 to m include an STM32 microprocessor, a wireless communication module, a load controller, a conversion controller and a switch controller. The load controller, the conversion controller and the switch controller are connected to the STM32 microprocessor to control the load, the power generation device and the power transmission equipment respectively; the STM32 microprocessor receives signals from the load controller, the conversion controller and the switch controller and sends control instructions to the load controller, the conversion controller and the switch controller; each power grid control terminal communicates through the wireless communication module and realizes information interaction with the power grid cloud server.
[0022] The present invention also discloses a smart grid parameter Internet of Things monitoring device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the steps of the above-mentioned smart grid parameter Internet of Things monitoring method are implemented.
[0023] Beneficial effects:
[0024] 1. Aiming at the uncertainty and randomness of problems such as sensor accuracy error, interference and abnormal measurement parameter values in the process of smart grid parameter measurement, the present invention expresses the measurement values of smart grid parameter sensors in the form of three-parameter interval numerical values, effectively dealing with the fuzziness and uncertainty of the measurement values of smart grid parameter sensors, and improving the objectivity and reliability of the predicted values of smart grid sensor parameters.
[0025] 2. The DRNN neural network power grid parameter model adopted in the present invention is a dynamic recurrent neural network with feedback and the ability to adapt to time-varying characteristics. The network can more directly and vividly reflect the dynamic changes in power grid parameters and can more accurately predict power grid parameters. The hidden layer of each DRNN neural network power grid parameter model is a regression layer with a self-feedback function, which improves the dynamic performance of the DRNN neural network power grid parameter model and provides an effective method for predicting smart grid parameters.
[0026] 3. The present invention uses wavelet decomposition of the power grid parameter model to decompose the time series signal of the power supply parameters into different resolution spaces. The effect after such processing is that the power supply parameter time series decomposed into each resolution space is predicted by using the corresponding model respectively, which is more accurate than directly predicting the original data sequence of the power supply parameters.
[0027] Fourth, the present invention adopts the ARIMA power grid parameter model based on the original data of smart grid parameters obeying the time series distribution, using the principle that the changes of smart grid parameters have a certain inertial trend, integrating the original time series variables of smart grid parameters with trend factors, periodic factors and random errors, etc., transforming the non-stationary sequence into a stationary random sequence with zero mean through differential data conversion and other methods, and performing numerical fitting and prediction of smart grid parameters through repeated identification and model diagnosis comparison and selection of ideal models. This method combines the advantages of autoregression and moving average methods, has the characteristics of being unconstrained by data types and strong applicability, and is a model with good effect in short-term prediction of smart grid parameters.
[0028] 5. The BiLSTM neural network power grid parameter model of the present invention can be composed of two LSTM neural network power grid parameter models with opposite directions. The smart grid parameter information input to the BiLSTM neural network power grid parameter model will always correspond to two LSTM neural network power grid parameter models with opposite directions, and the output is also affected by the two LSTM neural network power grid parameter models with opposite directions. The hidden state of the bidirectional BiLSTM neural network power grid parameter model mainly depends on the current smart grid parameter signal input, the output of the hidden state corresponding to the detection signal at the previous moment, and the output of the hidden state of the detection signal in different directions, which improves the accuracy, reliability and robustness of the prediction of smart grid parameter information. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 A power supply detection module for the present invention;
[0030] Figure 2 The power grid power consumption prediction model of the present invention;
[0031] Figure 3 It is the smart grid parameter Internet of Things information management system of the present invention;
[0032] Figure 4 The power grid parameter measuring terminal of the present invention;
[0033] Figure 5 The power grid control terminal of the present invention;
[0034] Figure 6 It is the power grid information transmission terminal of the present invention;
[0035] Figure 7 This is a functional diagram of the power grid remote monitoring terminal software of the present invention. DETAILED DESCRIPTION
[0036] Combination Figure 1-7 The technical solution of the present invention is further described as follows:
[0037] 1. Design of power supply detection module
[0038] The power supply detection module includes BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1, ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model, BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1, BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1-2, wavelet decomposition power grid parameter model, BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter model 1-2. The structure and function of the power supply detection module are shown in Figure 1 shown.
[0039] 1. BA's CNN convolutional neural network power grid parameter model - BA's DRNN neural network power grid parameter model design
[0040] The CNN convolutional neural network power grid parameter model of BA-the DRNN neural network power grid parameter model of BA is the CNN neural network power grid parameter model of BA and the DRNN neural network power grid parameter model of BA in series.
[0041] The CNN convolutional neural network power grid parameter model usually consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The convolutional layer mainly extracts features of the power grid parameters. The power grid parameter information from the input layer is convolved with the convolution kernel of the convolutional layer. The expression of the CNN convolutional neural network power grid parameter model is:
[0042]
[0043] Where C is the output state of the convolutional layer; is the activation function, and this paper adopts the ReLU function; Wc is the weight matrix of the convolutional layer; is the convolution operation; X is the input multi-feature data matrix; bc is the bias matrix of the convolution layer; P is the output of the pooling layer; H is the output of the fully connected layer; σ(·) is the Sigmoid activation function; Ws is the weight matrix of the fully connected layer; bs is the bias vector of the fully connected layer.
[0044] The Bat Algorithm (BA) is a random search algorithm that simulates the use of sonar by bats in nature to detect prey, avoid obstacles and connect it with the optimization target function. It realizes dynamic control of the conversion between local search and global search, and is far superior to other algorithm optimizations in terms of accuracy and effectiveness. Fitness is the driving force of the Bat Algorithm. The fitness function is also called the evaluation function. By mapping the fitness of an individual to the objective function of the optimization problem, the optimization of the objective function of the optimization problem can be achieved during the evolution of the bat population.
[0045] The CNN neural network power grid parameter model of BA is the principle of BA optimizing the CNN neural network power grid parameter model as follows: replace the connection weights and thresholds between the CNN neural network power grid parameter models with the individual position parameters of the bat population, encode each individual using real number coding, and obtain the initial bat population. The prediction error of the CNN neural network power grid parameter model is used as the fitness function of BA, and the fitness value is calculated and returned to BA. BA updates the position parameters, obtains new bat individuals, calculates new fitness function values, and repeats the update until the fitness function value drops to the minimum. At this time, the bat is the best selected individual, and the bat's position parameter value is the initialization weight and bias of the optimal CNN neural network power grid parameter model to be found.
[0046] The DRNN neural network power grid parameter model is a diagonal regression neural network. Its network structure is divided into input layer, hidden layer and output layer. The input layer is the same as the static network, using a linear function as the action function. The hidden layer contains N neurons, each neuron has a self-feedback regressor, using a Sigmoid function as the action function. The input of each neuron in the input layer is I(k), the regressor of the network regression layer is X(k), the sum of the input of the regression neuron is S(k), and the output layer is O(k). Let I = [I 1 (t),I 2 (t),…,I n (t)] is the input grid parameter vector of the DRNN neural network grid parameter model, where I i (t) is the input of the i-th neuron in the input layer of the DRNN neural network power grid parameter model at time t, and the output of the j-th neuron in the regression layer is X j (t), S j (t) is the sum of the inputs of the jth regression neuron, f(·) is a function of S, and O(t) is the output of the DRNN neural network power grid parameter model. The output layer output of the DRNN neural network power grid parameter model is:
[0047]
[0048] The DRNN neural network power grid parameter model of BA is optimized by BA based on the principle of: replacing the connection weights and thresholds between the DRNN neural network power grid parameter models with the individual position parameters of the bat population, encoding each individual using real number coding, and obtaining the initial bat population. The prediction error of the DRNN neural network power grid parameter model is used as the fitness function of BA, and the fitness value is calculated and returned to BA. BA updates the position parameters, obtains new bat individuals, and calculates new fitness function values. The update is repeated until the fitness function value drops to the minimum. At this time, the bat is the best selected individual, and the bat's position parameter value is the initialization weight and bias of the optimal DRNN neural network power grid parameter model to be found.
[0049] 2. ARIMA power grid parameter model-BA DRNN neural network power grid parameter model design
[0050] ARIMA power grid parameter model - BA's DRNN neural network power grid parameter model is a series connection of the ARIMA power grid parameter model and the BA's DRNN neural network power grid parameter model. The ARIMA (Auto regressive Integrated Moving Average) model organically combines the autoregressive model (Autoregressive, AR) and the moving average model (Moving Average, MA) to make it a comprehensive prediction method for power grid parameters. It is known as the most complex and advanced model in the time series prediction method. Since the original power grid parameter time series often shows certain trends or cyclical characteristics, it does not meet the ARMA model's requirements for the stability of the time series. Taking the difference is a convenient and effective way to eliminate the trend of power grid parameters.
[0051] The model established based on the differential power grid parameter time series is called the ARIMA power grid parameter model, which is recorded as {Xt}-ARIMA (p, d, q), where p and q are called the order of the model, and d represents the number of differences. When d is 0, the ARIMA power grid parameter model is defined as:
[0052] x t =b 1 x t-1 +…+b p x t-p +ε t +a 1 ε t-1 +…+a q ε t-q (3)
[0053] {x t} is the grid parameter data sequence to be predicted, {ε t}WN(0,σ 2 ).
[0054] The establishment of ARIMA power grid parameter model mainly includes model identification, parameter estimation and model diagnosis. Model identification mainly includes the preprocessing of power grid parameter time series and the preliminary order determination of model parameters; after the model order determination is completed, the unknown parameters in the model need to be estimated through the power grid parameter time series observations and combined with p, d, q values.
[0055] The ARIMA (p, d, q) power grid parameter model can be used to fit the power grid parameter time series. The modeling steps of the ARIMA (p, d, q) power grid parameter model are as follows: A. Obtain the smart grid parameter time series value. B. Determine the stationarity of the power grid parameter time series. If the series is non-stationary, it is necessary to preprocess the power grid parameter data and perform differential operations to make the power grid parameter time series stable, and determine the value of the differential order d. C. When the power grid parameter time series is a stationary non-white noise series after differentiation, we can choose an ARMA (p, q) power grid parameter model with an appropriate order to model the power grid parameter series. D. Estimate the unknown parameters in the model based on the identified model and its order. E. Test the residual sequence and use statistical tests to test whether this preliminary model is effective. F. Use the obtained fitting model to predict the future development trend of the stabilized power grid parameter time series.
[0056] The design steps of BA's DRNN neural network power grid parameter model refer to the relevant design methods of this patent.
[0057] 3. BiLSTM neural network power grid parameter model-ARIMA power grid parameter model design
[0058] The BiLSTM neural network power grid parameter model-ARIMA power grid parameter model is a series connection of the BiLSTM neural network power grid parameter model and the ARIMA power grid parameter model. The LSTM neural network model consists of a forget gate k t , input gate i t and output gate o t The grid parameter time series information x(t) at time t and the hidden state h at the previous time are composed t-1 As the input of the forget gate and the input gate; the state unit is connected to the forget gate and the input gate, and the unit state c at the previous moment is t-1 Updated to c t ; Updated cell state c t , x(t) and h t-1 Enter the output gate and update the hidden state to h through the sigmoid and tanh functions tIn order to enhance the learning ability of the LSTM neural network power grid parameter model and improve the extraction effect of the time series characteristics of the input power grid parameter information, the BiLSTM neural network power grid parameter model has a network structure that can bidirectionally extract the time series data characteristics of the input power grid parameter information, and can obtain the past power grid parameter information ignored by the LSTM neural network power grid parameter model. The hidden layer of the BiLSTM neural network power grid parameter model is obtained by weighted splicing of the hidden layer of the forward LSTM neural network power grid parameter model and the hidden layer of the backward LSTM neural network power grid parameter model. The calculation formula of the BiLSTM neural network power grid parameter model is:
[0059] h t '=α t h t +β t h p (4)
[0060] Where: h t ' is the hidden layer of the BiLSTM neural network power grid parameter model; h t is the hidden layer of the backward LSTM neural network power grid parameter model; h p is the hidden layer of the forward LSTM neural network power grid parameter model; α t and β t is the weight coefficient. The design method of the ARIMA power grid parameter model refers to the relevant design steps of this patent.
[0061] 4. BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model design
[0062] The BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model is a BiLSTM neural network power grid parameter model and a BA's DRNN neural network power grid parameter model connected in series. The output of the BiLSTM neural network power grid parameter model is used as the input of the BA's DRNN neural network power grid parameter model. The design method of the BiLSTM neural network power grid parameter model and the BA's DRNN neural network power grid parameter model refers to the relevant design method of this patent.
[0063] 5. BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter model design
[0064] BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter model is a series connection of BA's CNN neural network power grid parameter model and BiLSTM neural network power grid parameter model. The output of BA's CNN neural network power grid parameter model is used as the input of the BiLSTM neural network power grid parameter model. The design method of BA's CNN neural network power grid parameter model and BiLSTM neural network power grid parameter model refers to the relevant design method of this patent.
[0065] 6. Wavelet decomposition power grid parameter model design
[0066] The wavelet decomposition power grid parameter model decomposes the time series data of power supply parameters, and performs autocorrelation and cross-correlation analysis on the decomposed information of each layer; the wavelet decomposition power grid parameter model smoothes the signal, and establishes the corresponding BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model according to the characteristics of each layer of signal analysis to predict power grid parameters. The decomposition process of the wavelet decomposition power grid parameter model generally adopts the Mallat algorithm. The decomposition relationship of the wavelet decomposition power grid parameter model is expressed as follows:
[0067]
[0068] In formula (6), h 0 、h 1 They are low-pass analysis filter and high-pass analysis filter respectively. p 、n p The resolution is 2 -p The algorithm reconstructs the low-frequency coefficients and high-frequency coefficients under the following conditions:
[0069]
[0070] In formula (7), g 0 , g 1 A is the low-pass reconstruction filter and high-pass reconstruction filter of the power supply parameters respectively. p , D p Resolution 2 -p The Mallat algorithm decomposes the low-frequency signal part of the power supply parameter after each layer of decomposition into high frequency and low frequency again, and decomposes it layer by layer. The result obtained after the p-layer decomposition of the time series data X of the power supply parameter is:
[0071] X=D 1 +D 2 +…D p +A p (7)
[0072] In formula (7), A pis the low-frequency signal part of the power supply parameter after decomposition of the pth layer, D p It is the high-frequency part of the power supply parameters after the p-th layer decomposition. The wavelet decomposition power grid parameter model decomposes the power supply parameter time series data into different resolution spaces. The effect after such processing is that the time series data of the power supply parameters decomposed into each resolution space can be predicted more accurately.
[0073] 2. Design of power grid power consumption prediction model
[0074] The power grid power consumption prediction model includes power supply detection module 1-C, FCM power grid parameter classifier, BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2-4, BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 2.
[0075] The FCM grid parameter classifier algorithm is as follows:
[0076] Suppose there is a finite set X = {x 1 ,x 2 ,…x n} is a set of n grid parameter samples, the output of multiple grid parameter sensor groups input by the FCM grid parameter classifier, C is the number of predetermined categories, m i (i=1,2,…c) is the center of each cluster, μ j (x i ) is the membership of the i-th sample to the j-th class, and the clustering criterion function is defined by the membership function:
[0077]
[0078] Where:||x i -m j || is x i to m j b is the fuzzy weighted power index, which is a parameter that can control the fuzziness of the clustering results; M is the fuzzy C partition matrix of X, and V is the cluster center set of X. The result of the FCM power grid parameter classifier algorithm is to obtain M and V that minimize the criterion function. In the FCM power grid parameter classifier algorithm, the sum of the membership of the sample to the output cluster of each power supply parameter sensor group is required to be 1, that is:
[0079]
[0080] The FCM grid parameter classifier can be completed according to the following iterative steps: A. Set the number of grid parameter clusters c and parameter b, the algorithm termination threshold ε, the number of iterations t = 1, and the maximum number of iterations allowed is t max ; B. Initialize the clustering centers m of each power grid parameter i; C. Use the current grid parameter clustering center to calculate the membership function; D. Use the current membership function to update the clustering centers of various grid parameters; E. Select a suitable matrix norm. If ||V(t+1)-V(t)||≤ε or t≥t max , stop the operation; otherwise, t=t+1, return to step C. When the algorithm converges, the cluster centers of various types of power grid parameters and the membership of the output value samples of each type of power grid parameter sensor group to each type are obtained, and the fuzzy clustering division is completed. Finally, the fuzzy clustering results of the power grid parameters are defuzzified, and the fuzzy clustering is transformed into a deterministic power grid parameter classification to achieve the final classification of the output values of multiple power grid parameter sensor groups. The design methods of other models refer to the design steps of the relevant models of the power supply detection module.
[0081] 3. Design of Smart Grid Parameters IoT Management System
[0082] The smart grid parameter IoT management system consists of grid parameter measurement terminals 1 to n, grid control terminals 1 to m, grid information transmission terminals, grid remote monitoring terminals, grid cloud servers and grid monitoring APP. The grid parameter measurement terminals collect smart grid parameters and upload them to the grid cloud servers via the grid information transmission terminals, and store data and publish information on the grid cloud servers. The grid monitoring APP and the grid remote monitoring terminals access the smart grid parameter information of the grid cloud server through the 5G network to monitor the smart grid parameters in real time. The grid parameter measurement terminal and the grid control terminal are responsible for collecting smart grid parameters and controlling smart grid equipment. The grid information transmission terminal realizes two-way communication among the grid parameter measurement terminal, the grid control terminal, the grid remote monitoring terminal, the grid cloud server and the grid monitoring APP, thereby realizing smart grid parameter collection and smart grid equipment control. See the structure and functions of the smart grid parameter IoT management system. Figure 3 shown.
[0083] 1. Design of power grid parameter measurement terminal
[0084] The grid parameter measurement terminal based on the wireless communication module is used as the smart grid parameter perception terminal. The environmental sensor group includes a temperature sensor, a humidity sensor and a wind speed sensor. The grid parameter measurement terminal realizes information interaction between the grid information transmission terminals through the wireless communication module. The grid parameter measurement terminal includes the humidity, temperature, wind speed, current, voltage and corresponding signal conditioning circuits, electric meters, STM32 microprocessors, video image sensors and wireless communication modules for collecting smart grid information. The software of the grid parameter measurement terminal mainly realizes the communication of the wireless communication module, the collection and preprocessing of smart grid parameters. The software is designed with C language programming and has a high degree of compatibility, which greatly improves the work efficiency of software design and development and enhances the reliability, readability and portability of the program code. The structure and function of the grid parameter measurement terminal can be seen in Figure 4 .
[0085] 2. Design of the power grid control terminal
[0086] The grid control end realizes information interaction between the grid information transmission ends through the wireless communication module. The grid control end includes an STM32 microprocessor, a wireless communication module, a load controller, a conversion controller and a switch controller. The load controller, the conversion controller and the switch controller control the load, the power generation device and the power transmission equipment respectively. The structure of the grid control end is shown in Figure 5 .
[0087] 3. Design of power grid information transmission terminal
[0088] The power grid information transmission end includes a wireless communication module, an NB-IoT module, and an STM32 single-chip computer. The power grid information transmission end uses the wireless communication module to realize the communication network with the power grid parameter measurement end and the power grid control end. The NB-IoT module realizes the two-way data interaction between the power grid information transmission end and the power grid cloud server, and realizes the information interaction between the power grid information transmission end and the power grid remote monitoring end and the power grid monitoring APP. The structure of the power grid information transmission end is shown in Figure 6 .
[0089] 4. Power grid remote monitoring software
[0090] The power grid remote monitoring terminal is an industrial control computer. It mainly collects and processes smart grid parameters and interacts with the power grid information transmission terminal through the 5G network and the power grid cloud server. The main functions of the power grid remote monitoring terminal are power grid communication parameter setting, power grid data analysis and data management, and power grid power consumption prediction model. The software functions of the power grid remote monitoring terminal are shown in Figure 7 .
[0091] The present invention also discloses a smart grid parameter Internet of Things monitoring device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the steps of the above-mentioned smart grid parameter Internet of Things monitoring method are implemented.
[0092] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for monitoring smart grid parameters using the Internet of Things, characterized in that: The steps include: Collect output values of multiple current sensors, multiple electric meters, and multiple groups of environmental parameter sensors; Construct a power grid power consumption prediction model, including power supply detection module 1-C, FCM power grid parameter classifier, BiLSTM neural network power grid parameter model - ARIMA power grid parameter model 2-5, BA's CNN convolutional neural network power grid parameter model - BA's DRNN neural network power grid parameter model 2; Multiple current sensor outputs serve as inputs to power supply detection module 1, multiple electricity meter outputs serve as inputs to power supply detection module 2, multiple sets of environmental parameter sensor values serve as inputs to the FCM power grid parameter classifier, and multiple types of environmental parameter sensor values output by the FCM power grid parameter classifier serve as inputs to corresponding power supply detection modules 3-C. Power supply detection modules 1-C predict the output values of the corresponding current sensors, electricity meters, and multiple sets of environmental parameter sensors and output corresponding sensor prediction values. The output of the power supply detection module 1 is used as the corresponding input of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2. The time series difference between the output of the power supply detection module 1 and the output of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2 and the rate of change of the time series difference are used as the corresponding input of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2. The output of the power supply detection module 3-C is used as the corresponding input of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 4. The outputs of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 2, the power supply detection module 2 and the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 4 are used as the corresponding input of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 5. The outputs of the grid parameter model-ARIMA grid parameter model 3 and the BiLSTM neural network grid parameter model-ARIMA grid parameter model 5 are used as the corresponding inputs of the BA's CNN convolutional neural network grid parameter model-BA's DRNN neural network grid parameter model 2. The time series difference and the rate of change of the difference between the output of the power supply detection module 2 and the output of the BiLSTM neural network grid parameter model-ARIMA grid parameter model 5, and the time series difference and the rate of change of the difference between the output of the BA's CNN convolutional neural network grid parameter model-BA's DRNN neural network grid parameter model 2 and the output of the power supply detection module 2 are used as the corresponding inputs of the BiLSTM neural network grid parameter model-ARIMA grid parameter model 3 respectively. The three parameter interval numbers output by the BA's CNN convolutional neural network grid parameter model-BA's DRNN neural network grid parameter model 2 are used as the power prediction values of the detected power grid.
2. A smart grid parameter Internet of Things monitoring method according to claim 1, characterized in that: The power supply detection modules 1 to C have the same structure, and all include BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1, ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model, BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1, BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter models 1-2, wavelet decomposition power grid parameter model, BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter models 1-2; The output values of multiple current sensors or multiple electricity meters or multiple groups of environmental parameter sensor groups are respectively used as the corresponding inputs of BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1, the corresponding inputs of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1 and the corresponding inputs of BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1-2. The outputs of BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1 are respectively used as the inputs of ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model, BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1 and the input of wavelet decomposition power grid parameter model. The outputs of ARIMA power grid parameter model-BA's DRNN neural network power grid parameter model are used as the corresponding inputs of BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1. BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1 and The outputs of the BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1 are respectively used as the corresponding inputs of the BA's CNN convolutional neural network power grid parameter model-BiLSTM neural network power grid parameter model 1, the outputs of the wavelet decomposition power grid parameter model are used as the inputs of the BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 2, the outputs of the BiLSTM neural network power grid parameter model-BA's DRNN neural network power grid parameter model 2 and the BA's CNN convolutional neural network power grid parameter model-BiLSTM neural network power grid parameter model 1 are respectively used as the corresponding inputs of the BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter model 2, the outputs of the BiLSTM neural network power grid parameter model-ARIMA power grid parameter model 1 and the BA's CNN neural network power grid parameter model-BiLSTM neural network power grid parameter model 1-2 are three-parameter intervals consisting of a, b and c, and the number of intervals is [a, b, c] as the output prediction values of multiple current sensors or multiple electricity meters or multiple groups of environmental parameter sensors.
3. A method for monitoring smart grid parameters in the Internet of Things according to claim 2, characterized in that: BA's CNN convolutional neural network power grid parameter model-BA's DRNN neural network power grid parameter model 1-2 is the BA's CNN neural network power grid parameter model and the BA's DRNN neural network power grid parameter model connected in series; The CNN neural network power grid parameter model of the BA replaces the connection weights and thresholds between the CNN neural network power grid parameter models with the individual position parameters of the bat population, and encodes each individual in a real number encoding manner to obtain an initial bat population. The prediction error of the CNN neural network power grid parameter model is used as the fitness function of the BA, and the fitness value is calculated and returned to the BA. The BA updates the position parameters to obtain a new bat individual, calculates a new fitness function value, and repeats the update until the fitness function value drops to the minimum. At this time, the bat is the best selected individual, and the bat's position parameter value is the initialization weight and bias of the optimal CNN neural network power grid parameter model to be found; The BA's DRNN neural network power grid parameter model replaces the connection weights and thresholds between the DRNN neural network power grid parameter models with the individual position parameters of the bat population, and encodes each individual using real number coding to obtain the initial bat population. The prediction error of the DRNN neural network power grid parameter model is used as the fitness function of the BA, and the fitness value is calculated and returned to the BA. The BA updates the position parameters to obtain new bat individuals and calculates the new fitness function value. The update is repeated until the fitness function value drops to the minimum. At this time, the bat is the best selected individual, and the bat's position parameter value is the initialization weight and bias of the optimal DRNN neural network power grid parameter model to be found.
4. The method for monitoring smart grid parameters in the Internet of Things according to claim 1, wherein: The environmental parameter sensor group includes a temperature sensor, a humidity sensor and a wind speed sensor; multiple points are installed in the surrounding environment of the smart grid to be monitored; multiple current sensors and multiple electric meters are installed in the circuit of the smart grid to be detected.
5. A smart grid parameter Internet of Things management system, characterized in that: It includes power grid parameter measurement terminals 1~n, power grid control terminals 1~m, power grid cloud server, and power grid remote monitoring terminal; the power grid cloud server is connected to the power grid parameter measurement terminals 1~n and power grid control terminals 1~m via a power grid information transmission terminal, and the power grid cloud server is connected to the power grid remote monitoring terminal via a 5G network; The grid parameter measurement terminals 1~n collect grid parameters and upload them to the grid cloud server via the grid information transmission terminal. The grid cloud server is used to receive the data collected by the grid parameter measurement terminals 1~n. The grid remote monitoring terminal realizes grid communication parameter setting, grid data analysis and data management, and utilizes the grid power consumption prediction model to realize the smart grid parameter Internet of Things monitoring method as described in any one of claims 1 to 4. The grid cloud server issues control instructions based on the data processing results of the grid remote monitoring terminal; the grid control terminal receives the control instructions from the grid cloud server and controls the grid equipment to execute the instructions.
6. The smart grid parameter Internet of Things management system according to claim 5, characterized in that: It also includes a power grid monitoring APP, and the power grid cloud server is connected to the power grid monitoring APP via a 5G network.
7. The smart grid parameter Internet of Things management system according to claim 5, characterized in that: The power grid parameter measurement terminals 1-n include a temperature sensor, a humidity sensor, a wind speed sensor, a current sensor, a voltage sensor, an electric meter, and corresponding signal conditioning circuits, an STM32 microprocessor, a video image sensor, and a wireless communication module; the temperature sensor, humidity sensor, wind speed sensor, current sensor, voltage sensor, electric meter, and video image sensor are all connected to the STM32 microprocessor through a signal conditioning circuit; an ad hoc network is established between the power grid parameter measurement terminals 1-n through a wireless communication module to achieve information interaction with the power grid cloud server.
8. The smart grid parameter Internet of Things management system according to claim 5, characterized in that: The grid control terminals 1 to m include an STM32 microprocessor, a wireless communication module, a load controller, a conversion controller, and a switch controller. The load controller, the conversion controller, and the switch controller are connected to the STM32 microprocessor to control the load, the power generation device, and the power transmission equipment, respectively. The STM32 microprocessor receives signals from the load controller, the conversion controller, and the switch controller and sends control instructions to the load controller, the conversion controller, and the switch controller. Each grid control terminal communicates through the wireless communication module and realizes information exchange with the grid cloud server.
9. A smart grid parameter Internet of Things monitoring device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, the steps of a smart grid parameter Internet of Things monitoring method according to any one of claims 1 to 4 are implemented.
