A cloud computing-based electric energy demand management method

Abstract

The application relates to a cloud-computing-based electric energy demand management method, which is suitable for general industrial and commercial enterprises, large-scale industrial enterprises and the like. A database is constructed, the database stores instantaneous active power of each municipal incoming line and corresponding time; for i municipal incoming lines, if the number of data with instantaneous active power greater than 0 exceeds 80% of the total number of data, the municipal incoming line power supply mark is marked as 1; after performing AND operation on the power supply marks of all municipal incoming lines, a working mode mark bit is obtained; if the operation result is 1, the working mode is simultaneous power supply; if the operation result is 0, the working mode is one standby; the application aims to combine cloud computing technology, big data technology and industrial field monitoring technology, automatically predict the demand change of electric energy, achieve timely early warning, intelligent analysis, reasonable suggestion, finally realize enterprise cost reduction and reduction of electric load fluctuation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial internet of things, and is an electricity demand management method based on cloud computing. BACKGROUND

[0002] As the most important part of various energy consumption in large enterprises, electricity cost is directly related to the energy consumption and cost of the entire enterprise. The traditional electricity management takes saving electricity as the core, emphasizes reducing unnecessary electricity consumption and peak load shifting in the electricity process, and the basic electricity billing mode has become a new breakthrough in the environment of "capacity change demand". Electricity demand management not only has important effects of saving and reducing costs on enterprise operation and development, but also has important social benefits of energy saving and emission reduction and reducing electricity costs. Through scientific electricity demand management, the rational allocation of electricity resources can be realized, the influence of electricity side load fluctuation on power transmission side and power grid can be reduced, the electricity supply-demand gap can be greatly reduced, and the overall electricity cost of the society can be reduced, and the supply quality and efficiency can be improved.

[0003] For various large industrial and commercial enterprises, it is still complex to realize electricity demand management, and lacks operability. It is almost impossible to realize high-frequency monitoring and analysis of a large amount of electricity data through manual operation. If only the instantaneous power of the enterprise gateway meter is monitored, the alarm form of exceeding the early warning value has great hysteresis, and it is almost useless to avoid the intervention of the management to avoid the demand value. At the same time, the demand value alarm form also limits the attention of the electricity data to the data of the gateway meter, and lacks the information support for the manager to find the information support for the intervention management and the peak of the energy use. Therefore, only by introducing an intelligent big data platform and intelligent algorithm for management and control, the current bottleneck of electricity management can be broken through. SUMMARY

[0004] The present application relates to the technical field of industrial internet of things, and is an electricity demand management method based on cloud computing.

[0005] The present application relates to the technical field of industrial internet of things, and is an electricity demand management method based on cloud computing. The present application relates to the technical field of industrial internet of things, and is an electricity demand management method based on cloud computing.

[0006] The division step of the electric energy demand management:

[0007] Simplify the physical model of the enterprise power distribution architecture to form a total and branch radiation architecture with only municipal power supply lines and multiple important power consumption points. For the total input part of the simplified physical model, usually the municipal double power supply lines of the enterprise, intelligent electric energy meters are installed on both lines for the collection of power supply side electric energy data. For the branch output part of the simplified physical model, the selection of end power consumption points is selected, focusing on electric power lines and other power supply lines that have large changes in power consumption load and play a decisive role in total load changes.

[0008] All metering points are equipped with intelligent electric meters as metering equipment, and real-time electric energy collection data are sent to NB-IoT DTU wireless terminals through RS485 serial ports. Narrowband Internet of Things technology is used to realize communication with the cloud server. The DTU wireless terminal sends serial data to the server of the system set up in the cloud, and the server completes real-time data storage. When storing, the sequence number Id field and the parent sequence number PId field are used to effectively distinguish the subordinate relationship between the municipal total line meter and the end metering point meter.

[0009] According to the current time t, the time length T judge , the data in the database that meet the conditions (t≥(now-T judge ) and (t<now) are filtered, and the instantaneous active power of each municipal total line meter is traversed in turn. For i municipal lines, the number of data with instantaneous active power P i (n) greater than 0 exceeds 80% of the total number of data, and the municipal line power supply flag InLineOn i is marked as 1. According to the power supply flags of all municipal lines, the working mode flag Flag mode is obtained by performing an AND operation. If the operation result is 1, the working mode is simultaneous power supply; if the operation result is 0, the working mode is one standby.

[0010] For the municipal lines in the working state, cloud computing analysis is performed, and according to the historical data of the instantaneous active power P(n) stored in the database, an algorithm model is established, which includes a linear autoregressive algorithm and a white noise influence prediction base, to realize the prediction of the future total power load change of the enterprise.

[0011] The future load trend change of the total line is predicted and compared with the set demand warning value to realize the demand warning function. Each time the prediction algorithm is executed, the instantaneous active power P t in the database for the current time is taken as the prediction base.

[0012] According to the algorithm of the prediction model, the instantaneous active power P at time {t+1, t+2, …, t+n} is sequentially predicted t+n , realizing continuous prediction of the instantaneous power in the future period. According to the demand early warning value set by the enterprise, when there are a large number of predicted values higher than the demand early warning value, a demand exceeding alarm is given.

[0013] When the demand exceeding alarm is triggered, the end trend analysis algorithm is enabled to judge the main inducing power points of the demand early warning, assisting the management personnel in optimizing power consumption decision-making, etc.

[0014] The setting of a large number of parameters in the algorithm formula of the prediction model requires that after the installation of all meters at the selected monitoring points and the collection of a certain period of data, the model is trained according to the historical data of the measured instantaneous power.

[0015] In the training process, the noise part of the algorithm is first ignored, and by accessing the continuous m historical data in any period of time in the database, the autoregressive coefficient of the linear autoregressive part is solved.

[0016] By accessing multiple continuous m historical data in different time periods in the database, the autoregressive coefficient is repeatedly solved multiple times, and finally the median value of the parameter value of each is taken to filter out the influence of part of the distorted data, realizing the setting of the autoregressive coefficient .

[0017] After determining the autoregressive coefficient, by accessing the continuous 2*m historical data in any period of time in the database, the estimation parameters θ m-1 and white noise sequence ε t of the noise part of the algorithm model are solved. In the model training process, the {1, 2, …, m} historical data are taken as historical values, and the {m+1, m+2, …, 2m} historical data are taken as known predicted values to solve the parameters. The solving process is given by the following formula, and the estimation parameters and white noise sequence are obtained by the maximum likelihood estimation method.

[0018] When the future trend of the total power consumption load of the entire enterprise exceeds the demand warning line, trend analysis needs to be performed on the end important metering points corresponding to the power supply line that generates the warning to obtain the main reasons affecting the upward trend of the total power consumption load.

[0019] According to the working mode flag Flag mode , the end objects that need to be analyzed are screened. If the working mode flag Flag mode= 0, the working mode of the current feeder is one active and one standby mode, all the terminals of the whole plant need to be analyzed, and then the power load trend of all the meters with non-empty parent sequence number PId field in the data table is analyzed in sequence.If the working mode flag bit Flag mode = 1, the working mode of the current feeder is two-way simultaneous power supply, the terminal corresponding to the feeder with a warning signal needs to be analyzed, and then the power load trend of all the meters with the sequence number PId field being the sequence number Id of the current warning line in the data table is analyzed in sequence. The analysis process adopts a sign test algorithm of interval difference, and the small sample terminal power consumption data is analyzed.

[0020] According to the collection frequency f collect and the analysis time length T analysis , a data set P 末 composed of instantaneous active power is obtained, and the sample number n.

[0021] According to the data set P composed of instantaneous active power, a trend data set Dis is generated, and the sample number N.

[0022] In the trend data set Dis, the number of data Dis i > 0 is Dis + . When the power of this metering point has no obvious upward or downward trend in the analysis time length T analysis , the number of positive data and the number of negative data in the trend data set Dis should be binomial distribution. If the result value is greater than a certain amplitude, it is considered that the metering point is in a severe upward trend, which is one of the reasons for the change of the main load, and the returned sequence number Id value is used to identify the corresponding metering point.

[0023] The present application has the following beneficial effects:

[0024] 1. For large industrial and commercial enterprises, the construction time is earlier, the power distribution equipment is old, and it is not possible to stop production and work for large-scale modification during the long-term operation process. The present application simplifies the range of technical modification required on site. Through the installation of Internet of Things collection equipment and smart meters, the basic data collection work of key power points is realized, which greatly reduces the hardware investment and construction period in the technical modification process, optimizes the economy in the implementation process, and improves the implementability of the demand management technology application.

[0025] 2, the actual power distribution architecture of the comprehensive enterprise is comprehensively considered, monitoring points of main power consumption points are selected, through monitoring and joint analysis of high-voltage total incoming line of the enterprise and low-voltage total incoming line of important power consumption terminal, the maximization of demand management effect is realized under the limited transformation cost. The joint prediction of total and distribution also becomes a big lever for management to intervene in the power consumption points with the trend of exceeding the demand under the condition of demand management process, and completely avoids the problem that management has no direction. In view of the power trend change of the terminal metering point, the sign test algorithm of interval difference is selected, which can effectively identify the power consumption points with real rising trend in the data sample with small data volume and non-linear data change.

[0026] 3, for the instantaneous power of the power supply total incoming line collected in real time on the spot, the energy management comprehensive platform side carries out timing data storage, and the future value of instantaneous load is predicted by using the prediction model. The prediction model algorithm designed by the method is used to analyze the historical data, and the effective model training strategy is used to obtain the actual algorithm parameter value. In the actual use process, the load change in the future time period can be accurately predicted. And in the prediction process, this method can also process the influence of temporary, sudden change or large noise historical data on the result, and has higher guarantee for the accuracy of demand prediction and early warning, and can really realize the effective early warning of demand. BRIEF DESCRIPTION OF DRAWINGS

[0027] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments, made with reference to the accompanying drawings:

[0028] Figure 1 The system diagram of the power meter collection point arrangement in the present application is shown in the figure;

[0029] Figure 2 The trend chart of predicting future load based on historical load value in the present application is shown in the figure;

[0030] Figure 3 The comparison trend chart of predicting future load based on historical load value and actual load change in the present application is shown in the figure;

[0031] Figure 4 The power sampling chart for identifying no obvious trend change of terminal power consumption in the present application is shown in the figure;

[0032] Figure 5 The power sampling chart for identifying rising trend of terminal power consumption in the present application is shown in the figure; DETAILED DESCRIPTION

[0033] The present application will be described below in combination with the drawings and formula calculation.

[0034] As Figures 1-4The application is a power demand management method based on cloud computing, comprising the following steps:

[0035] 1. The internal power distribution architecture of large-scale industrial and commercial enterprises is usually large, which is a radial structure with step-by-step voltage reduction from the total voltage reduction station to the center station and finally to the distribution room. In the process of power transformation and distribution at each level, there are usually thousands of power consumption points that need to be monitored in the power line and feeder loop. If all of them are monitored, the cost is high and the analysis is difficult. In the actual analysis process of the method, only the metering points in the power distribution architecture that can represent the total power load of the enterprise and the metering points with high proportion and high change rate of power load in the power consumption end are monitored. The addition of the former metering point is mainly used for the statistics of the instantaneous load of the whole enterprise, so as to realize the algorithm prediction and intelligent early warning of demand trend. The addition of the latter metering point is mainly used for the main cause analysis of demand exceeding under demand early warning, so as to find the power consumption end that can be intervened and managed. Therefore, the first step of the implementation of the application is to simplify the physical model of the enterprise power distribution architecture to form a total and divided radial architecture with only municipal power supply lines and multiple important power consumption metering points, as shown in Figure 1 .

[0036] For the total input part of the simplified physical model, the two power supply lines of the enterprise are usually installed with intelligent electric energy meters for the collection of power data on the power supply side. In the process of use, the two power supply lines exist in two working conditions of one use and one standby and double supply. When the power supply line is in one use and one standby working condition for power supply, one of the lines will bear the entire power load of the enterprise, and the other line will be idle and not consume power. At this time, the load change of all end metering in the whole factory only affects the load of the line in use. When the two power supply lines supply power at the same time, the contact cabinet at the total voltage reduction station side is usually disconnected, and the two lines are responsible for part of the load of the whole factory, so when analyzing the power demand management, the two lines need to be analyzed respectively, and when the demand of one line is warned, only the end metering point is searched for the reason of load change. Therefore, in the process of power demand management calculation of the method, first, the working condition mode corresponding to the collected and stored line power data needs to be automatically identified and compatible.

[0037] For the selection of the end power consumption metering point of the simplified physical model, the method does not need to consider the power load stable or the power load too small line, such as office building, commercial lighting, etc. The selection of the end power consumption metering point focuses on the power line and other power feeder lines with large power load change and decisive factors for total load change, such as air compression station, commercial water chiller room, water pump room, etc.

[0038] All metering points are equipped with smart meters as metering devices, and real-time power collection data is sent to NB-IoT DTU wireless terminal through RS485 serial port. Narrowband Internet of Things technology is used to realize communication with cloud server. The collection frequency f collect of DTU wireless terminal is set according to the accuracy of the algorithm. The DTU wireless terminal sends the serial data to the server in the cloud built by the system, and the server stores the data in real time. The data table structure for analyzing the power demand management is as follows.

[0039] Table 1. Power data storage table

[0040] Serial number Parent serial number Instantaneous active power (kW) Data update time Id1 PId1 P1 Timestamp1 Id2 PId2 P2 Timestamp2 … … … … Id(n) PId(n) P(n) Timestamp(n)

[0041] The sequence number Id field in Table 1 is used to distinguish the Id of the meter itself of each metering point. The parent sequence number PId field is used to attribute the Id of the meter corresponding to the previous power distribution line of each metering point. (The parent sequence number PId field is stored as null, which is the municipal total incoming line meter; the parent sequence number PId field is non-empty data, which distinguishes the subordinate relationship between the end metering point and the municipal incoming line)

[0042] According to the current time t, the time length T judge , the data is screened, and the time length T judge is usually 10 minutes. In the data that meet the conditions (t≥(now-T judge ) and (t<now) in the database, the instantaneous active power of each municipal total incoming line meter is traversed in turn. For i municipal incoming lines, the instantaneous active power P i (n) is greater than 0, and the number of data is more than 80% of the total number of data, the municipal incoming line power supply flag InLineOn i is marked as 1.

[0043] Flag mode = And[InLineOn i ]

[0044] According to the logical AND operation of all municipal incoming line power supply flags, the working mode flag Flag mode is obtained. If the result of the working mode flag is 1, the working mode is simultaneous power supply; if the result of the working mode flag is 0, the working mode is one active and one standby.

[0045] For the meter with the parent sequence number PId field stored as null, i.e. the municipal total incoming line meter, if the corresponding municipal incoming line power supply flag InLineOn iIf 1, it represents that the incoming line is in working state. For all municipal incoming lines in working state, cloud computing analysis is carried out, a prediction model is established and real-time trend prediction is carried out according to the historical data of instantaneous active power P(n) stored in the database. The future load trend change of the total incoming line is predicted and compared with the set demand alarm value, so as to realize the demand early warning function.

[0046] Since the data change curve of electric energy load usually shows macro-continuity and micro-fluctuation in time sequence, such characteristic data shows that the change implies law is not completely disordered, and has the characteristic of gradually changing with the passage of time. Through the big data platform, the historical instantaneous power data of enterprise municipal incoming line is stored for a long time, so that the electric load change becomes a data group with a long historical trend. The basic idea of the prediction stage of the method is that the instantaneous power at a time point is composed of the influence of its historical data, and is also affected by the fluctuation deviation caused by accidental events in the past period. Therefore, when predicting the instantaneous power, two parts are processed, one is to consider the influence of the past historical value on the current existence by using linear autoregressive algorithm, and the other is to assume that the influence of each past time on the present exists noise relationship, each time point white noise is independent and subject to the same distribution, and has constant mean and variance.

[0047] Each time the prediction algorithm is executed, the instantaneous active power P t is taken as the prediction base. The number num is determined according to the accuracy requirement of prediction and the hardware computing capacity of the server, and the size of the set number num usually needs to reach all the data of several days ago.

[0048] Prediction algorithm expression:

[0049]

[0050] t: the time point when the prediction is performed;

[0051] num: the set prediction base number;

[0052] c: constant, linear deviation value of autoregressive algorithm part;

[0053] for num historical instantaneous power data, corresponding autoregressive coefficient;

[0054] ε t-num : for num historical instantaneous power data, corresponding white noise sequence;

[0055] θ num-1 : for num white noise, corresponding estimation parameter;

[0056] According to the algorithm of the prediction model, the instantaneous active power P at time {t+1, t+2, …, t+n} is sequentially predicted t+n , realizing continuous prediction of the instantaneous power in the future period. According to the demand early warning value set by the enterprise, when there are a large number of predicted values higher than the demand early warning value, a demand exceeding standard alarm is given, such as Figure 2 .

[0057] When the demand exceeding standard alarm is triggered, step 4 is entered, the end trend analysis algorithm is enabled, the main inducing power points of the demand early warning are judged, and the management personnel are assisted to optimize the power consumption decision, etc.

[0058] The setting of a large number of parameters in the algorithm formula of the prediction model requires that after the installation of all the meters of the selected monitoring points and the collection of a certain period of data, the model is trained according to the historical data of the measured instantaneous power. The model training time is usually 20 to 30 days.

[0059] In the training process, the noise part of the algorithm is first ignored, and by accessing the continuous m historical data in any period of time in the database, the autoregressive coefficient of the linear autoregressive part is solved. The solving process is as follows:

[0060]

[0061] By accessing the continuous m historical data in multiple different time periods in the database, the autoregressive coefficient is repeatedly solved multiple times, and finally the median value of the parameter value of each is taken, the influence of part of the distorted data is filtered out, and the autoregressive coefficient is set.

[0062] After the autoregressive coefficient is determined, by accessing the continuous 2*m historical data in any period of time in the database, the estimation parameters θ m-1 and white noise sequence ε t of the noise part of the algorithm model are continued to be solved. In the model training process, the {1, 2, …, m} historical data are taken as the historical values, and the {m+1, m+2, …, 2m} historical data are taken as the known predicted values to solve the parameters. The solving process is to solve the following equations, and the estimation parameters and white noise sequence are obtained by maximum likelihood estimation method.

[0063]

[0064] After the model training is completed, the algorithm is more accurate for the trend prediction of the instantaneous active power of the total incoming line of the enterprise, and can effectively identify the upcoming power peak, such as Figure 3 .

[0065] According to the calculation of step 2, 3, the total power load of the whole enterprise future trend has more than the demand warning line, the trend analysis needs to be carried out for the corresponding end important metering point of the power supply line, and the main reason for the total power load rising situation is obtained.

[0066] First, according to the working mode flag bit Flag mode , the end object that needs to be analyzed is screened. If the working mode flag bit Flag mode = 0, the working mode of the current incoming line is one use one standby mode, and all the ends of the factory need to be analyzed, and the power load trend of the parent sequence number PId field in the data table is analyzed in turn. If the working mode flag bit Flag mode = 1, the working mode of the current incoming line is two-way power supply, and the end corresponding to the incoming line with warning signal needs to be analyzed, and the power load trend of the sequence number PId field in the data table is analyzed in turn. The sequence number Id of the current warning line.

[0067] Because when analyzing the trend change of a terminal power metering point, the measured data often does not have an intuitive rising characteristic, that is, it is often nonlinear change and not a straight line rising trend, and only according to the slope change between different time data to obtain whether there is an upward trend is too one-sided, so it is necessary to use cloud computing algorithm form to analyze the trend. Different from the analysis in the total incoming line prediction stage, the power consumption of the branch line is more inclined to micro analysis, and the incremental of the terminal power often does not have time characteristics, and tracing the data in the early stage does not have practical significance, and is not suitable for using linear regression calculation method. In view of the analysis of the terminal power that causes the total load to rise, the data quantity used is often very small, only in a period before the warning occurs, so this method uses interval difference sign test algorithm to analyze the trend of small sample terminal power data.

[0068] According to the collection frequency f collect and the analysis time length T analysis , a data set P 末 composed of instantaneous active power is obtained, and the sample number n.

[0069] P 末 = {P1, P2, …, P n}

[0070]

[0071] According to the data set P composed of instantaneous active power, a trend data set Dis is generated, and the sample number N.

[0072] Dis = {Dis1, Dis2, …, Dis N}

[0073] Dis i = (P i+N - P i )

[0074]

[0075] The statistical trend data set Dis, Dis i > 0, is the number of data in Dis + When the electrical energy of this metering point does not have a significant upward or downward trend within the analysis time length T analysis , the number of positive data and the number of negative data in the trend data set Dis should be binomially distributed, such as Figure 4 If the result value of is greater than a certain amplitude, it is considered that this end metering point is in a sharp upward trend, such as Figure 5 is one of the reasons for the main load change, and the returned serial number Id value is used to identify the corresponding metering point.

[0076] The above shows and describes the basic principles and main features of the present application and the advantages of the present application, and it is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be considered as limiting the claims involved.

[0077] In addition, it should be understood that although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the description manner of the specification is only for clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be properly combined to form other embodiments that those skilled in the art can understand.

Claims

1. A cloud computing-based method for electricity demand management, characterized in that... Includes the following steps: Construct a database that stores the instantaneous active power and corresponding time for each municipal power line: The physical model of the enterprise's power distribution architecture is simplified to form a total-branch radiation architecture with only municipal power input and multiple important electricity metering points. For the total input part of the simplified physical model, there are two municipal power inputs for the enterprise, and smart energy meters are installed on both inputs for the collection of power supply side power data. For the output part of the simplified physical model, power feeders are selected for power monitoring. All metering points are equipped with smart meters as metering devices, and the collected energy data is sent to the NB-IoT DTU wireless terminal in real time via RS485 serial port. Narrowband IoT technology is used to achieve communication with the cloud server. The DTU wireless terminal packages the serial port data and sends it to the server set up in the cloud for this system. The server stores the data in real time. During storage, the data is stored using serial numbers. Field, Parent Sequence Number The field effectively distinguishes the hierarchical relationship between the municipal main incoming line meter and the terminal metering point meter; Based on the current time "now", the corresponding data update time "timestamp" and time length for instantaneous active power. Filter and select those that meet the requirements in the database. and In the data provided, the instantaneous active power of each municipal incoming line is sequentially traversed; for Municipal incoming line, instantaneous active power If the number of data points greater than 0 exceeds 80% of the total number of data points, then the municipal incoming power supply flag will be displayed. =1; After performing a bitwise AND operation on all municipal incoming power supply signs, the operating mode flag is obtained. : If the calculation result is 1, the working mode is simultaneous power supply; if the calculation result is 0, the working mode is one in use and one in standby. If the working mode flag is If the current operating mode of the incoming line is two lines supplying power simultaneously, and we need to analyze the terminal of the incoming line with the warning signal, then we need to analyze the parent serial number in all meters in the data table in turn. The field is the sequence number of the current warning line. The trend of electrical load; The sign test algorithm based on the interval difference is used to perform situational analysis on a small sample of end-point electricity consumption data: Based on the sampling frequency and analysis time length This yields a dataset consisting of instantaneous active power. Number of samples ; Data set based on instantaneous active power Generate trend data sets Number of samples ; Statistical trend data set middle, Data volume When the electrical energy at this metering point is analyzed over a period of time... When there is no obvious upward or downward trend within the data set, the trend data set The number of positive and negative data points should follow a binomial distribution. like If the result value exceeds a certain range, it is considered that this terminal metering point is experiencing a sharp upward trend, which is one of the causes of major load changes. The returned sequence number will then be used. The value indicates the corresponding measurement point.

2. The cloud computing-based power demand management method according to claim 1, characterized in that, The analysis of electricity consumption patterns at end-metering points during total load demand early warning also includes the following steps: According to the working mode flag Filter the end objects that need to be analyzed; if the working mode flag is set... If the current operating mode of the incoming line is one in use and one on standby, then it is necessary to analyze all terminals in the entire plant. Therefore, in the data table, analyze the parent serial number of each meter sequentially. The field represents the electricity load trend when it is not empty.

Images