Workshop power supply and distribution remote monitoring system based on Internet of Things technology

Abstract

The invention relates to the technical field of intelligent power distribution, and discloses a factory building power supply and distribution remote monitoring system based on the Internet of Things technology, and the system comprises a data collection module which collects the operation parameters of electric equipment; the data processing module generates a feature sequence; the first prediction module is used for predicting state parameters of the first K time periods in the future time period T2 through a first prediction model; the second prediction module is used for predicting state parameters of the last K time periods in the future time period T2 through a second prediction model; the prediction state matrix construction module is used for constructing a prediction state matrix; the power distribution scheme generation module generates a power distribution scheme through a particle swarm algorithm; according to the method, machine learning is combined with the neural network to capture the change characteristics of the operation parameters of the electric equipment in the time dimension, the prediction of the electricity consumption and the energy consumption in multiple time periods in the future is realized, and the unified scheduling of the electric equipment in the plant is realized in combination with the particle swarm optimization, so that the electric power waste is reduced, and the electric power response speed is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent power distribution technology, and more specifically, to a remote monitoring system for power supply and distribution in a plant based on the Internet of Things technology. Background Art

[0002] The power supply and distribution system of a factory building refers to an electrical system that provides power supply to various production equipment, lighting facilities, air-conditioning systems, etc. in industrial factories. The functions of the power supply and distribution system of a factory building mainly include: power transmission, distribution, regulation and protection. It is usually connected to high-voltage power from an external power grid, and converted into low-voltage power through power distribution equipment in the factory building (such as transformers, distribution cabinets, etc.) and distributed to electrical equipment in the factory building.

[0003] Traditional power supply and distribution systems in factories usually collect power consumption data of power-consuming equipment through power monitoring equipment (such as smart meters, sensors, etc.), and realize power distribution through manually preset rules combined with automatic control programs, where the automatic control program can be PLC (programmable logic controller), DCS (distributed control system), etc. However, the traditional power supply and distribution system in factories has the following problems: 1. The power supply and distribution regulation of factories is usually based on a single device or area, and there is a lack of a unified scheduling mechanism, that is, it is impossible to dynamically adjust power distribution according to actual load demand, resulting in unnecessary power waste. 2. There is a lack of the ability to predict the power demand and operating status of power-consuming equipment. Static manually preset rules cannot adapt to sudden load changes in time, resulting in delayed power response. Summary of the invention

[0004] The present invention provides a remote monitoring system for power supply and distribution in a factory building based on the Internet of Things technology, which solves the technical problems in the above-mentioned background technology.

[0005] The present invention provides a remote monitoring system for power supply and distribution in a factory based on the Internet of Things technology, comprising: A data collection module, which is used to collect the operating parameters of M electrical equipment in the factory building at preset time intervals t within a preset time period T1; Operation parameters include: holiday identification, shift identification, ambient temperature, output power, rated power, power consumption, electricity price, energy efficiency ratio, carbon emissions and power supply priority; A data processing module, which is used to process the operating parameters of M electrical devices to generate a feature sequence; The characteristic sequence includes N sequence units, and the nth sequence unit represents the operating parameters at the nth time point after data processing, where 1≤n≤N, N=T1 / t; A first prediction module, which is used to input the feature sequence of each electric device into the first prediction model, and the output value represents the state parameter of the electric device in the first K time periods in the future time period T2; The status parameters include: electricity consumption forecast and energy consumption coefficient; A second prediction module is used to input the feature sequence of each electric device into a second prediction model, and the output value represents the state parameter of the electric device in the next K time periods in the future time period T2; A prediction state matrix construction module is used to concatenate the state parameters of the first K time periods and the state parameters of the last K time periods of the M electrical devices in the future time period T2 to obtain a prediction state matrix; The prediction state matrix includes M rows and H columns, and the element values ​​of the prediction state matrix are represented by state parameters, where H=2×K; The power distribution plan generation module is used to generate a power distribution plan according to the predicted state matrix through a particle swarm algorithm.

[0006] Furthermore, the preset time period T1, the preset time interval t, the number of electrical devices M, the future time period T2 and the number of time periods K are all custom parameters.

[0007] Furthermore, data processing is performed on the operating parameters to generate a characteristic sequence, including the following steps: Step S201, converting the holiday identifier, shift identifier and power supply priority in the operating parameters into numerical representation; The holiday flag is represented by 0 or 1, where 0 represents a non-holiday and 1 represents a holiday; The shift identifier is represented by a positive integer ranging from 1 to 3, where 1 represents the morning shift, 2 represents the midday shift, and 3 represents the evening shift; The power supply priority is manually set and represented by a positive integer ranging from 1 to 5. The larger the value, the higher the priority. Step S202, performing interpolation and filling processing on the missing values ​​in the operating parameters; For the missing values ​​of ambient temperature, output power, power consumption, electricity price and carbon emissions in the operating parameters, interpolation filling is performed by taking the average value of the values ​​of two adjacent time points corresponding to the missing values; For the missing values ​​of holiday identification, shift identification, rated power, energy efficiency ratio and power supply priority in the operating parameters, interpolation filling is performed by randomly taking the value corresponding to an adjacent time point of the missing value; Step S203: normalize the operating parameters at each time point using the Min-Max method to generate a feature sequence.

[0008] Further, the first prediction model is composed of two first units, two second units, two third units, two fourth units, K first integrated units and K second integrated units; The inputs of the first, second, third, and fourth units are all feature sequences; The outputs of one of the first unit, the second unit, the third unit and the fourth unit are all the predicted power consumption of the power equipment in the first K time periods in the future time period T2, and the outputs of the other first unit, the second unit, the third unit and the fourth unit are all the energy consumption coefficients of the power equipment in the first K time periods in the future time period T2; The K first integrated units are respectively used to perform weighted summation on the predicted power consumption of the power-consuming equipment in the first K time periods in the future time period T2 output by one of the first units, the second unit, the third unit and the fourth unit, and the sum of the weight coefficients of the first unit, the second unit, the third unit and the fourth unit in each first integrated unit is 1; The K second integrated units are respectively used to perform weighted summation on the energy consumption coefficients of the electrical equipment output by another first unit, second unit, third unit and fourth unit in the first K time periods in the future time period T2, and the sum of the weight coefficients of the first unit, the second unit, the third unit and the fourth unit in each second integrated unit is 1; The first unit is built on random forest, the second unit is built on lightweight gradient boosting machine, the third unit is built on classification gradient boosting machine, and the fourth unit is built on extreme gradient boosting machine.

[0009] Furthermore, after the preset time period T1, the power consumption and output power of the electrical equipment in the first K time periods in the future time period T2 are collected, and the energy consumption coefficient is calculated in combination with the energy efficiency ratio of the electrical equipment to obtain a sample label for training samples for training the first prediction model, where the energy consumption coefficient is equal to power consumption / (output power × energy efficiency ratio).

[0010] Further, the second prediction model is composed of the first prediction model excluding the K first integrated units and the K second integrated units, a first time series analysis layer, a first hidden layer, a first splicing layer, a second time series analysis layer, a second hidden layer, a second splicing layer, K first classifiers and K second classifiers; The first time series analysis layer inputs a feature sequence and outputs a first hidden vector; The input of the first hidden layer is a first vector, the dimension of the first vector is 4×K, which respectively corresponds to the predicted power consumption of the power-consuming equipment in the first K time periods in the future time period T2 output by one of the first unit, the second unit, the third unit and the fourth unit of the first prediction model, and the output of the first hidden layer is a first update vector; The first concatenation layer is used to concatenate the first hidden vector and the first update vector to obtain a first combined vector; The second time series analysis layer inputs the feature sequence and outputs the second hidden vector; The input of the second hidden layer is a second vector, the dimension number of the second vector is 4×K, which respectively corresponds to the energy consumption coefficient of the electrical equipment in the first K time periods in the future time period T2 output by another first unit, second unit, third unit and fourth unit of the first prediction model, and the output of the second hidden layer is a second update vector; The second concatenation layer is used for concatenating the second hidden vector and the second update vector to obtain a second combined vector; The first combination vector is input into K first classifiers, and the classification space of the K first classifiers represents the predicted power consumption of the electrical equipment in the next K time periods in the future time period T2; The second combination vector is input into K second classifiers, and the classification space of the K second classifiers represents the energy consumption coefficient of the electrical equipment in the next K time periods in the future time period T2; The first time series analysis layer and the second time series analysis layer are both built based on the GRU model.

[0011] Furthermore, the calculation formula of the first hidden layer is the same as the calculation formula of the second hidden layer, and the calculation formula of the first hidden layer is as follows: ; in represents the first update vector of the first hidden layer output, represents the first vector of the first hidden layer input, and denote the first weight parameter and the second weight parameter respectively, and They represent the first bias parameter and the second bias parameter respectively, Swish represents the Swish activation function, and GELU represents the GELU activation function.

[0012] Furthermore, after the preset time period T1, the power consumption and output power of the electrical equipment in the next K time periods in the future time period T2 are collected, and the energy consumption coefficient is calculated in combination with the energy efficiency ratio of the electrical equipment to serve as the sample label of the training sample for training the second prediction model, where the energy consumption coefficient is equal to power consumption / (output power × energy efficiency ratio).

[0013] Furthermore, all individuals in the initialization population of the particle swarm algorithm are represented by matrix coding, and the matrix coding includes M rows and H columns. The element value of the mth row and the hth column represents the output power adjustment value of the mth electrical equipment in the hth time period in the future time period T2, wherein 1≤m≤M, 1≤h≤H, and the element values ​​of the matrix coding all meet the constraint conditions. In each iteration of the particle swarm algorithm, the fitness values ​​of all individuals in the initialization population are obtained by calculating the objective function until the iteration termination condition is met, and the matrix code output of the individual with the largest fitness value in the initialization population is used as the power distribution plan; The constraints include: the output power adjustment value of each power-consuming device in each time period cannot exceed the upper and lower limits of its output power, where the upper and lower limits of the output power of each power-consuming device are custom parameters; the output power adjustment value of each power-consuming device in each time period multiplied by the length of a time period must be greater than or equal to the corresponding power consumption forecast in the prediction state matrix; the output power adjustment value of all power-consuming devices in each time period cannot exceed the total distribution power, where the total distribution power is a custom parameter; The iteration termination conditions include: the number of iterations reaches the maximum number of iterations, where the maximum number of iterations is a custom parameter; within three consecutive iterations, the change rate of the maximum value of the fitness value does not exceed 5%.

[0014] Furthermore, the objective function is calculated as follows: ; Where Fit represents the fitness value, represents the output power adjustment value corresponding to the mth row and hth column of the matrix code, represents the energy consumption coefficient corresponding to the mth row and hth column of the predicted state matrix, and They represent the electricity price and carbon emissions of the mth electrical equipment at the Nth time point respectively.

[0015] The beneficial effects of the present invention are as follows: the present invention uses machine learning combined with neural networks to capture the changing characteristics of the operating parameters of electrical equipment in the time dimension, realizes the prediction of power consumption and energy consumption in multiple time periods in the future, and combines the particle swarm algorithm to realize the unified scheduling of electrical equipment in the factory, thereby reducing power waste and improving the speed of power response. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of a remote monitoring system for power supply and distribution in a factory based on the Internet of Things technology of the present invention; Figure 2 is a flow chart of data processing to generate a feature sequence according to the present invention; Figure 3 is a schematic diagram of a first prediction model of the present invention; Figure 4 is a schematic diagram of a second prediction model of the present invention.

[0017] In the figure: a data acquisition module 101, a data processing module 102, a first prediction module 103, a second prediction module 104, a prediction state matrix construction module 105, and a power distribution plan generation module 106. DETAILED DESCRIPTION

[0018] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0019] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] like Figure 1 to Figure 4 As shown, a remote monitoring system for power supply and distribution in a factory based on the Internet of Things technology includes: The data collection module 101 is used to collect the operating parameters of M electrical equipment in the factory building at preset time intervals t within a preset time period T1; Operation parameters include: holiday identification, shift identification, ambient temperature, output power, rated power, power consumption, electricity price, energy efficiency ratio, carbon emissions and power supply priority; A data processing module 102, which is used to perform data processing on the operating parameters of M electrical devices to generate a feature sequence; The characteristic sequence includes N sequence units, and the nth sequence unit represents the operating parameters at the nth time point after data processing, where 1≤n≤N, N=T1 / t; The first prediction module 103 is used to input the feature sequence of each electric device into the first prediction model, and the output value represents the state parameter of the electric device in the first K time periods in the future time period T2; The status parameters include: electricity consumption forecast and energy consumption coefficient; The second prediction module 104 is used to input the feature sequence of each electric device into the second prediction model, and the output value represents the state parameter of the electric device in the next K time periods in the future time period T2; A prediction state matrix building module 105 is used to concatenate the state parameters of the first K time periods and the state parameters of the last K time periods of the M electrical devices in the future time period T2 to obtain a prediction state matrix; The prediction state matrix includes M rows and H columns, and the element values ​​of the prediction state matrix are represented by state parameters, where H=2×K; The power distribution plan generation module 106 is used to generate a power distribution plan according to the predicted state matrix by using a particle swarm algorithm.

[0021] In one embodiment of the present invention, the preset time period T1, the preset time interval t, the number of electrical equipment M, the future time period T2 and the number of time periods K are all custom parameters. Preferably, the preset time period T1 is set to 24 hours, the preset time interval t is set to 15 minutes, then the number of sequence units of the feature sequence N=96, the number of electrical equipment M is set according to the actual equipment running in the factory, the future time period T2 is set to 8 hours, K is set to 4, then the number of columns of the prediction state matrix H=8.

[0022] In one embodiment of the present invention, the ambient temperature is acquired through a temperature sensor; the output power is acquired through an electric power monitoring device, which may be a smart meter or a power meter; the rated power is provided by the equipment manufacturer; the power consumption is also acquired through the electric power monitoring device; the electricity price is acquired through an interface query provided by the power company; the energy efficiency ratio is provided by the equipment manufacturer; the carbon emissions are equal to the power consumption multiplied by the carbon emission factor, which indicates the carbon dioxide emissions corresponding to each kilowatt-hour of electricity generated and is provided by the power company. For example, the carbon emission factor of thermal power generation is 0.9kgCO2 / kWh.

[0023] It should be noted that the carbon emission factor of natural gas power generation is 0.4kgCO2 / kWh, and the carbon emission factor of new energy power generation such as wind power generation, photovoltaic power generation and hydropower generation is 0kgCO2 / kWh. The carbon emissions can be calculated based on the proportion of the above-mentioned power generation types in actual power generation, which will not be elaborated here.

[0024] In one embodiment of the present invention, Figure 2 As shown, data processing of the operating parameters is performed to generate a characteristic sequence, including the following steps: Step S201, converting the holiday identifier, shift identifier and power supply priority in the operating parameters into numerical representation; The holiday flag is represented by 0 or 1, where 0 represents a non-holiday and 1 represents a holiday; The shift identifier is represented by a positive integer ranging from 1 to 3, where 1 represents the morning shift, 2 represents the midday shift, and 3 represents the evening shift; For example, the morning shift starts at 6 a.m. and ends at 2 p.m.; the mid-day shift starts at 2 p.m. and ends at 10 p.m.; the evening shift starts at 10 p.m. and ends at 6 a.m. the next day. The power supply priority is manually set and represented by a positive integer ranging from 1 to 5. The larger the value, the higher the priority. Step S202, performing interpolation and filling processing on the missing values ​​in the operating parameters; For the missing values ​​of ambient temperature, output power, power consumption, electricity price and carbon emissions in the operating parameters, interpolation filling is performed by taking the average value of the values ​​of two adjacent time points corresponding to the missing values; For the missing values ​​of holiday identification, shift identification, rated power, energy efficiency ratio and power supply priority in the operating parameters, interpolation filling is performed by randomly taking the value corresponding to an adjacent time point of the missing value; Step S203: normalize the operating parameters at each time point using the Min-Max method to generate a feature sequence.

[0025] It should be noted that, taking the electrical equipment in industrial plants as an example, the power supply priority of production line equipment is set to 5 (core production equipment must be powered first to ensure the normal operation of the production line), the power supply priority of key mechanical equipment is set to 4 (high-precision processing equipment requires a higher priority power supply to avoid processing interruptions), the power supply priority of storage and transportation equipment is set to 3 (as auxiliary equipment for production services, it has a lower priority than core production equipment), the power supply priority of lighting equipment is set to 2 (interruptions in the short term will not directly affect production), and the power supply priority of air-conditioning equipment is set to 1 (as comfort auxiliary equipment, it has a lower priority).

[0026] In one embodiment of the present invention, the first prediction model is composed of two first units, two second units, two third units, two fourth units, K first integrated units and K second integrated units; The inputs of the first, second, third, and fourth units are all feature sequences; The outputs of one of the first unit, the second unit, the third unit and the fourth unit are all the predicted power consumption of the power equipment in the first K time periods in the future time period T2, and the outputs of the other first unit, the second unit, the third unit and the fourth unit are all the energy consumption coefficients of the power equipment in the first K time periods in the future time period T2; The K first integrated units are respectively used to perform weighted summation on the predicted power consumption of the power-consuming equipment in the first K time periods in the future time period T2 output by one of the first units, the second unit, the third unit and the fourth unit, and the sum of the weight coefficients of the first unit, the second unit, the third unit and the fourth unit in each first integrated unit is 1; The K second integrated units are respectively used to perform weighted summation on the energy consumption coefficients of the electrical equipment output by another first unit, second unit, third unit and fourth unit in the first K time periods in the future time period T2, and the sum of the weight coefficients of the first unit, second unit, third unit and fourth unit in each second integrated unit is 1.

[0027] In one embodiment of the present invention, the first unit is constructed based on Random Forest, the second unit is constructed based on LightGBM (Lightweight Gradient Boosting Machine), the third unit is constructed based on CatBoost (Classification Gradient Boosting Machine), and the fourth unit is constructed based on XGBoost (Extreme Gradient Boosting Machine).

[0028] It should be noted that the weight coefficients of the first unit, the second unit, the third unit and the fourth unit in the first integrated unit and the second integrated unit can be custom settings or learnable hyperparameters. The weight coefficients are automatically updated by a gradient descent algorithm (such as AdaGrad, RMSProp, etc.) to minimize the prediction error. In addition, the first prediction model integrates multiple sub-units to make up for the shortcomings of a single model, improve the robustness of the first prediction model, and improve the accuracy of the prediction.

[0029] For example, the calculation formula for customizing the weight coefficient of the first integrated unit is as follows: ; Where 1≤k≤K, represents the weight coefficient of the i-th unit (the first unit, the second unit, the third unit and the fourth unit) in the k-th first integrated unit, It represents the mean square error of the i-th unit in the k-th first integrated unit. The smaller the error, the larger the corresponding weight coefficient.

[0030] In one embodiment of the present invention, after a preset time period T1, the power consumption and output power of the electrical equipment in the first K time periods in the future time period T2 are collected, and the energy consumption coefficient is calculated in combination with the energy efficiency ratio of the electrical equipment as a sample label of the training sample for training the first prediction model, where the energy consumption coefficient is equal to power consumption / (output power × energy efficiency ratio).

[0031] In one embodiment of the present invention, the second prediction model is composed of the first prediction model excluding K first integrated units and K second integrated units, a first time series analysis layer, a first hidden layer, a first splicing layer, a second time series analysis layer, a second hidden layer, a second splicing layer, K first classifiers and K second classifiers; The first time series analysis layer inputs a feature sequence and outputs a first hidden vector; The input of the first hidden layer is a first vector, the dimension of the first vector is 4×K, which respectively corresponds to the predicted power consumption of the power-consuming equipment in the first K time periods in the future time period T2 output by one of the first unit, the second unit, the third unit and the fourth unit of the first prediction model, and the output of the first hidden layer is a first update vector; The first concatenation layer is used to concatenate the first hidden vector and the first update vector to obtain a first combined vector; The second time series analysis layer inputs the feature sequence and outputs the second hidden vector; The input of the second hidden layer is a second vector, the dimension number of the second vector is 4×K, which respectively corresponds to the energy consumption coefficient of the electrical equipment in the first K time periods in the future time period T2 output by another first unit, second unit, third unit and fourth unit of the first prediction model, and the output of the second hidden layer is a second update vector; The second concatenation layer is used for concatenating the second hidden vector and the second update vector to obtain a second combined vector; The first combination vector is input into K first classifiers, and the classification space of the K first classifiers represents the predicted power consumption of the electrical equipment in the next K time periods in the future time period T2; The second combination vector is input into K second classifiers, and the classification space of the K second classifiers represents the energy consumption coefficient of the electrical equipment in the next K time periods in the future time period T2.

[0032] In one embodiment of the present invention, the first time series analysis layer and the second time series analysis layer are both constructed based on the GRU model, and may also be constructed based on the LSTM model, which will not be described in detail here.

[0033] It should be noted that the K first classifiers do not share weight parameters and bias parameters, and the K second classifiers do not share weight parameters and bias parameters. The first prediction model focuses on short-term prediction, and the second prediction model focuses on long-term prediction. The second prediction model adds a neural network on the basis of the first prediction model, which can increase the nonlinear capture ability of the feature sequence, thereby improving the accuracy of the long-term prediction.

[0034] In one embodiment of the present invention, the calculation formula of the first hidden layer is the same as the calculation formula of the second hidden layer. The calculation formula of the first hidden layer is as follows: ; in represents the first update vector of the first hidden layer output, represents the first vector of the first hidden layer input, and denote the first weight parameter and the second weight parameter respectively, and They represent the first bias parameter and the second bias parameter respectively, Swish represents the Swish activation function, and GELU represents the GELU activation function.

[0035] It should be noted that the weight parameters and bias parameters of the first hidden layer and the second hidden layer are all learnable hyperparameters, for example It can be designed as a 4K×32 matrix. Instead of multiplying to get a 1×32 vector, It can be designed as a matrix of size 32×16, and the size of the first update vector finally outputted is 1×16.

[0036] In one embodiment of the present invention, after a preset time period T1, the power consumption and output power of the electrical equipment in the next K time periods in the future time period T2 are collected, and the energy consumption coefficient is calculated in combination with the energy efficiency ratio of the electrical equipment to serve as a sample label for training samples for training a second prediction model, where the energy consumption coefficient is equal to power consumption / (output power × energy efficiency ratio).

[0037] In one embodiment of the present invention, all individuals of the initialization population of the particle swarm algorithm are represented by matrix coding, the matrix coding includes M rows and H columns, the element value of the mth row and the hth column represents the output power adjustment value of the mth electrical equipment in the hth time period in the future time period T2, wherein 1≤m≤M, 1≤h≤H, the element values ​​of the matrix coding all meet the constraint conditions, in each iteration of the particle swarm algorithm, the fitness values ​​of all individuals of the initialization population are obtained by calculating the objective function until the iteration termination condition is met, and the matrix code of the individual with the largest fitness value in the initialization population is output as the power distribution plan; The constraints include: the output power adjustment value of each power-consuming device in each time period cannot exceed the upper and lower limits of its output power, where the upper and lower limits of the output power of each power-consuming device are custom parameters; the output power adjustment value of each power-consuming device in each time period multiplied by the length of a time period must be greater than or equal to the corresponding power consumption forecast in the prediction state matrix; the output power adjustment value of all power-consuming devices in each time period cannot exceed the total distribution power, where the total distribution power is a custom parameter; The iteration termination conditions include: the number of iterations reaches the maximum number of iterations, where the maximum number of iterations is a custom parameter, preferably, the maximum number of iterations is set to 20; within 3 consecutive iterations, the change rate of the maximum value of the fitness value does not exceed 5%.

[0038] In one embodiment of the present invention, the calculation formula of the objective function is as follows: ; Where Fit represents the fitness value, represents the output power adjustment value corresponding to the mth row and hth column of the matrix code, represents the energy consumption coefficient corresponding to the mth row and hth column of the predicted state matrix, and They represent the electricity price and carbon emissions of the mth electrical equipment at the Nth time point respectively.

[0039] In one embodiment of the present invention, the output power of M electrical equipment in H time periods in the future time period T2 is adjusted according to the power distribution plan. When the total distribution power does not meet the actual operating requirements, the electrical equipment is distributed in order of power supply priority from large to small.

[0040] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are protected by the present embodiment.

Claims

1. A remote monitoring system for power supply and distribution in a factory based on Internet of Things technology, characterized in that: include: A data collection module, which is used to collect the operating parameters of M electrical equipment in the factory building at preset time intervals t within a preset time period T1; Operation parameters include: holiday identification, shift identification, ambient temperature, output power, rated power, power consumption, electricity price, energy efficiency ratio, carbon emissions and power supply priority; A data processing module, which is used to process the operating parameters of M electrical devices to generate a feature sequence; The characteristic sequence includes N sequence units, and the nth sequence unit represents the operating parameters at the nth time point after data processing, where 1≤n≤N, N=T1 / t; A first prediction module, which is used to input the feature sequence of each electric device into the first prediction model, and the output value represents the state parameter of the electric device in the first K time periods in the future time period T2; The status parameters include: electricity consumption forecast and energy consumption coefficient; A second prediction module is used to input the feature sequence of each electric device into a second prediction model, and the output value represents the state parameter of the electric device in the next K time periods in the future time period T2; A prediction state matrix construction module is used to concatenate the state parameters of the first K time periods and the state parameters of the last K time periods of the M electrical devices in the future time period T2 to obtain a prediction state matrix; The prediction state matrix includes M rows and H columns, and the element values ​​of the prediction state matrix are represented by state parameters, where H=2×K; The power distribution plan generation module is used to generate a power distribution plan according to the predicted state matrix through a particle swarm algorithm.

2. According to the remote monitoring system for power supply and distribution in a factory building based on Internet of Things technology in claim 1, it is characterized in that: The preset time period T1, the preset time interval t, the number of electrical devices M, the future time period T2 and the number of time periods K are all custom parameters.

3. According to the remote monitoring system for power supply and distribution in a factory building based on Internet of Things technology in claim 1, it is characterized in that: The operation parameters are processed to generate a characteristic sequence, including the following steps: Step S201, converting the holiday identifier, shift identifier and power supply priority in the operating parameters into numerical representation; The holiday flag is represented by 0 or 1, where 0 represents a non-holiday and 1 represents a holiday; The shift identifier is represented by a positive integer ranging from 1 to 3, where 1 represents the morning shift, 2 represents the midday shift, and 3 represents the evening shift; The power supply priority is manually set and represented by a positive integer ranging from 1 to 5. The larger the value, the higher the priority. Step S202, performing interpolation and filling processing on the missing values ​​in the operating parameters; For the missing values ​​of ambient temperature, output power, power consumption, electricity price and carbon emissions in the operating parameters, interpolation filling is performed by taking the average value of the values ​​of two adjacent time points corresponding to the missing values; For the missing values ​​of holiday identification, shift identification, rated power, energy efficiency ratio and power supply priority in the operating parameters, interpolation filling is performed by randomly taking the value corresponding to an adjacent time point of the missing value; Step S203: normalize the operating parameters at each time point using the Min-Max method to generate a feature sequence.

4. According to the remote monitoring system for power supply and distribution in a factory based on Internet of Things technology as described in claim 1, it is characterized in that: The first prediction model consists of two first units, two second units, two third units, two fourth units, K first integrated units and K second integrated units; The inputs of the first, second, third, and fourth units are all feature sequences; The outputs of one of the first unit, the second unit, the third unit and the fourth unit are all the predicted power consumption of the power equipment in the first K time periods in the future time period T2, and the outputs of the other first unit, the second unit, the third unit and the fourth unit are all the energy consumption coefficients of the power equipment in the first K time periods in the future time period T2; The K first integrated units are respectively used to perform weighted summation on the predicted power consumption of the power-consuming equipment in the first K time periods in the future time period T2 output by one of the first units, the second unit, the third unit and the fourth unit, and the sum of the weight coefficients of the first unit, the second unit, the third unit and the fourth unit in each first integrated unit is 1; The K second integrated units are respectively used to perform weighted summation on the energy consumption coefficients of the electrical equipment output by another first unit, second unit, third unit and fourth unit in the first K time periods in the future time period T2, and the sum of the weight coefficients of the first unit, the second unit, the third unit and the fourth unit in each second integrated unit is 1; The first unit is built on random forest, the second unit is built on lightweight gradient boosting machine, the third unit is built on classification gradient boosting machine, and the fourth unit is built on extreme gradient boosting machine.

5. According to the remote monitoring system for power supply and distribution in a factory building based on Internet of Things technology in claim 1, it is characterized in that: After the preset time period T1, the power consumption and output power of the electrical equipment in the first K time periods in the future time period T2 are collected, and the energy consumption coefficient is obtained in combination with the energy efficiency ratio of the electrical equipment as the sample label of the training sample for training the first prediction model, where the energy consumption coefficient is equal to power consumption / (output power × energy efficiency ratio).

6. According to the remote monitoring system for power supply and distribution of a factory building based on Internet of Things technology in claim 1, it is characterized in that: The second prediction model is composed of the first prediction model excluding the K first integrated units and the K second integrated units, a first time series analysis layer, a first hidden layer, a first splicing layer, a second time series analysis layer, a second hidden layer, a second splicing layer, K first classifiers and K second classifiers; The first time series analysis layer inputs a feature sequence and outputs a first hidden vector; The input of the first hidden layer is a first vector, the dimension of the first vector is 4×K, which respectively corresponds to the predicted power consumption of the power-consuming equipment in the first K time periods in the future time period T2 output by one of the first unit, the second unit, the third unit and the fourth unit of the first prediction model, and the output of the first hidden layer is a first update vector; The first concatenation layer is used to concatenate the first hidden vector and the first update vector to obtain a first combined vector; The second time series analysis layer inputs the feature sequence and outputs the second hidden vector; The input of the second hidden layer is a second vector, the dimension number of the second vector is 4×K, which respectively corresponds to the energy consumption coefficient of the electrical equipment in the first K time periods in the future time period T2 output by another first unit, second unit, third unit and fourth unit of the first prediction model, and the output of the second hidden layer is a second update vector; The second concatenation layer is used for concatenating the second hidden vector and the second update vector to obtain a second combined vector; The first combination vector is input into K first classifiers, and the classification space of the K first classifiers represents the predicted power consumption of the electrical equipment in the next K time periods in the future time period T2; The second combination vector is input into K second classifiers, and the classification space of the K second classifiers represents the energy consumption coefficient of the electrical equipment in the next K time periods in the future time period T2; The first time series analysis layer and the second time series analysis layer are both built based on the GRU model.

7. The remote monitoring system for power supply and distribution in a factory based on Internet of Things technology according to claim 6 is characterized in that: The calculation formula of the first hidden layer is the same as that of the second hidden layer. The calculation formula of the first hidden layer is as follows: ； in represents the first update vector of the first hidden layer output, represents the first vector of the first hidden layer input, and denote the first weight parameter and the second weight parameter respectively, and They represent the first bias parameter and the second bias parameter respectively, Swish represents the Swish activation function, and GELU represents the GELU activation function.

8. The remote monitoring system for power supply and distribution in a factory based on Internet of Things technology according to claim 1 is characterized in that: After the preset time period T1, the power consumption and output power of the electrical equipment in the next K time periods in the future time period T2 are collected, and the energy consumption coefficient is obtained in combination with the energy efficiency ratio of the electrical equipment as the sample label of the training sample for training the second prediction model, where the energy consumption coefficient is equal to power consumption / (output power × energy efficiency ratio).

9. The remote monitoring system for power supply and distribution in a factory based on Internet of Things technology according to claim 1 is characterized in that: All individuals in the initialization population of the particle swarm algorithm are represented by matrix coding, and the matrix coding includes M rows and H columns. The element value of the mth row and the hth column represents the output power adjustment value of the mth electrical equipment in the hth time period in the future time period T2, where 1≤m≤M, 1≤h≤H, and the element values ​​of the matrix coding all meet the constraint conditions. In each iteration of the particle swarm algorithm, the fitness values ​​of all individuals in the initialization population are obtained by calculating the objective function until the iteration termination condition is met, and the matrix code output of the individual with the largest fitness value in the initialization population is used as the power distribution plan; The constraints include: the output power adjustment value of each power-consuming device in each time period cannot exceed the upper and lower limits of its output power, where the upper and lower limits of the output power of each power-consuming device are custom parameters; the output power adjustment value of each power-consuming device in each time period multiplied by the length of a time period must be greater than or equal to the corresponding power consumption forecast in the prediction state matrix; the output power adjustment value of all power-consuming devices in each time period cannot exceed the total distribution power, where the total distribution power is a custom parameter; The iteration termination conditions include: the number of iterations reaches the maximum number of iterations, where the maximum number of iterations is a custom parameter; within three consecutive iterations, the change rate of the maximum value of the fitness value does not exceed 5%.

10. A remote monitoring system for power supply and distribution in a factory based on Internet of Things technology according to claim 9, characterized in that: The objective function is calculated as follows: ； Where Fit represents the fitness value, represents the output power adjustment value corresponding to the mth row and hth column of the matrix code, represents the energy consumption coefficient corresponding to the mth row and hth column of the predicted state matrix, and They represent the electricity price and carbon emissions of the mth electrical equipment at the Nth time point respectively.

Images