A power consumption monitoring system and method based on power big data

Abstract

The present application discloses a power consumption monitoring system and method based on power big data to solve the technical problem that existing solutions require a large amount of manpower and material resources. It includes: an acquisition module for acquiring the original data set uploaded by a preset monitored enterprise; a cleaning module for removing the values of the data in the original data set that do not conform to the preset specifications and removing the outliers beyond the reasonable range to obtain a second data set; a clustering module for clustering the power consumption status corresponding to the preset monitored enterprise and determining the status threshold corresponding to the preset monitored enterprise; a training module for using the second data set and the marked violation information as training data to train a bidirectional long short-term memory neural network model to alarm the violation information of the power data uploaded by the preset monitored enterprise through the trained bidirectional long short-term memory neural network model. The present application realizes the real-time monitoring of the illegal production behavior of the preset monitored enterprise through the above method.

Description

Technical Field

[0001] The present application relates to the technical field of electricity consumption monitoring, and in particular to an electricity consumption monitoring system and method based on power big data. Background Art

[0002] As an important indicator of whether an enterprise has started production, electricity consumption information plays a pivotal role in the management of illegal production in enterprises. In recent years, illegal and irregular production, open shutdown and dark operation, daytime shutdown and night operation, and other illegal and irregular production behaviors have become important factors in causing production safety accidents. The existing method of supervising production safety accidents is mainly for relevant staff to discover illegal and irregular situations through inspections.

[0003] However, the above methods require a lot of manpower and material resources, and because they are manual prevention, there is a certain degree of error. In order to prevent related production safety accidents, there is an urgent need for an electricity consumption monitoring system and method based on power big data. By analyzing the abnormal electricity consumption of enterprises, exploring the use of power big data, cloud computing, artificial intelligence algorithms and other technologies, conducting enterprise electricity consumption law analysis, building a monitoring and alarm model, and developing and launching a "power-assisted emergency management" monitoring and analysis system, timely discover illegal production, illegal construction, overcapacity production and other illegal behaviors of closed enterprises, give full play to the supporting role of power big data in the analysis and judgment of illegal production of closed enterprises, improve the ability of enterprises to produce safely, and promote the modernization of safety governance capabilities. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides an electricity consumption monitoring system and method based on power big data to solve the above-mentioned technical problems.

[0005] In a first aspect, an electricity consumption monitoring system based on power big data provided by an embodiment of the present application includes: an acquisition module for acquiring an original data set uploaded by a preset monitored enterprise; wherein the original data set includes enterprise basic data and power data; a cleaning module for removing values of data in the original data set that do not conform to preset specifications to obtain a first data set; by determining a reasonable range of each data in the first data set, removing outliers in the first data set that exceed the reasonable range; and further obtaining a second data set; a clustering module for performing clustering processing on the power data in the second data set to cluster the electricity consumption status corresponding to the preset monitored enterprise; and determining a status threshold corresponding to the preset monitored enterprise according to the electricity consumption status and the power data; wherein the status threshold includes a production stop threshold, a normal production threshold, and an over-limit production threshold; a training module for determining marked violation information corresponding to the second data set according to violation information calculated based on the status threshold and actually collected violation information; using the second data set and the marked violation information as training data to train a bidirectional long short-term memory neural network model; and performing violation information warning on the power data uploaded by the preset monitored enterprise through the trained bidirectional long short-term memory neural network model.

[0006] Further, the cleaning module includes a box plot unit; the box plot unit is used to calculate a reasonable range through a preset box plot formula: LowerLimit = Max{Q 1 - 1.5IQR, Mininum};

[0007] UpperLimit = Min{Q 3 + 1.5IQR, Maxinum}, where any set of data is sorted from smallest to largest and divided into 4 equal parts, Q1 represents the number at the first equal part point among the three equal part points; Q3 represents the number at the third equal part point among the three equal part points; IQR represents the value of Q3 - Q1; Minimum represents the smallest number in any set of data, and Maximum represents the largest number in any set of data; determining the range from LowerLimit to UpperLimit as the reasonable range, and further removing outliers in the first data set that exceed the reasonable range.

[0008] Further, the clustering module includes a status threshold unit; the status threshold unit is used to determine the specific quantity of the second data set corresponding to the preset monitored enterprise; when the specific data is less than a preset quantity threshold, determining the status threshold corresponding to the preset monitored enterprise through a clustering algorithm for calculating a data center and a class interval; when the specific data is equal to or greater than the preset quantity threshold, determining the boundary point of the power data in the second data set as the status threshold corresponding to the preset monitored enterprise through a preset boundary point clustering algorithm.

[0009] Second aspect, an embodiment of the present application provides a power consumption monitoring method based on power big data, and the method includes: obtaining an original data set uploaded by a preset monitored enterprise; wherein, the original data set includes enterprise basic data and power data; removing values of data in the original data set that do not conform to preset specifications to obtain a first data set; determining a reasonable range of each data in the first data set, and removing outliers in the first data set that exceed the reasonable range; thereby obtaining a second data set; performing clustering processing on the power data in the second data set to cluster the power consumption status corresponding to the preset monitored enterprise; determining a status threshold corresponding to the preset monitored enterprise according to the power consumption status and the power data; wherein, the status threshold includes a production suspension threshold, a normal production threshold, and an over-limit production threshold; determining the marked violation information corresponding to the second data set according to the violation information calculated based on the status threshold and the actually collected violation information; using the second data set and the marked violation information as training data to train a bidirectional long short-term memory neural network model; and performing violation information warning on the power data uploaded by the preset monitored enterprise through the trained bidirectional long short-term memory neural network model.

[0010] Further, determining a reasonable range of each data in the first data set and removing outliers in the first data set that exceed the reasonable range specifically includes: through a preset box plot formula:

[0011] LowerLimit = Max{Q 1 - 1.5IQR, Mininum}; UpperLimit = Min{Q 3 + 1.5IQR, Maxinum}, calculating the reasonable range; wherein, sorting any group of data from smallest to largest and dividing it into 4 equal parts, Q1 represents the number at the first equal part point among the three equal part points; Q3 represents the number at the third equal part point among the three equal part points; IQR represents the value of Q3 - Q1; Minimum represents the smallest number in any group of data, and Maximum represents the largest number in any group of data; determining the range between LowerLimit and UpperLimit as the reasonable range, and thereby removing outliers in the first data set that exceed the reasonable range.

[0012] Further, performing clustering processing on the power data in the second data set to cluster the power consumption status corresponding to the preset monitored enterprise specifically includes: randomly selecting several power data in the second data set as sample data through the k-means algorithm, determining that each sample data initially represents the average value or center μ j of a power consumption status, and marking the power consumption status closest to the sample data as the affiliated power consumption status; that is: wherein, X represents the sample data, i represents the number of sample data; C represents the power consumption status, and j represents the number of centers or average values of the power consumption status.

[0013] Furthermore, determine the power consumption status corresponding to the power data in each second dataset.

[0014] Further, according to the power consumption status and the power data, determine the status threshold corresponding to the preset monitoring enterprise, specifically including: determining the specific quantity of the power data in the second dataset corresponding to the preset monitoring enterprise; when the specific data is less than the preset quantity threshold, determine the status threshold corresponding to the preset monitoring enterprise through the clustering algorithm of the preset calculation data center and class interval; when the specific data is equal to or greater than the preset quantity threshold, determine the boundary point corresponding to the power data in the second dataset as the status threshold corresponding to the preset monitoring enterprise through the preset calculation boundary point clustering algorithm.

[0015] Further, the basic enterprise data at least includes enterprise status data, enterprise location data, and enterprise legal person data; the power data at least includes enterprise power customer file data, daily power consumption data, current data, power data, load data, business expansion and installation data, and derating data.

[0016] Those skilled in the art can understand that the present invention has at least the following beneficial effects: Through the acquisition module, the acquisition of the original dataset of the preset monitoring enterprise (key supervision enterprise) is realized, and then through the real-time acquisition and monitoring of the power consumption of the key supervision enterprise, the work efficiency of strengthening safety production supervision by using power big data is improved. Through the cleaning module, the cleaning of invalid data and abnormal data is realized. Through the clustering module, the power consumption status corresponding to the preset monitoring enterprise is obtained; furthermore, the status threshold corresponding to the preset monitoring enterprise is determined. Through the training module, the training of the bidirectional long short-term memory neural network model is completed, and a bidirectional long short-term memory neural network model capable of giving warnings about violation information is obtained. By continuously obtaining the actually collected violation information, the bidirectional long short-term memory neural network model can obtain the law of illegal production of enterprises changing in fitting, and further achieve the technical effect of "one enterprise, one model". BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The following describes some embodiments of the present disclosure with reference to the drawings, in which:

[0018] Figure 1 is a schematic internal structure diagram of a power consumption monitoring system based on power big data provided by an embodiment of the present application.

[0019] Figure 2 is a flowchart of a power consumption monitoring method based on power big data provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Those skilled in the art should understand that the embodiments described below are only the preferred embodiments of the present disclosure, and do not mean that the present disclosure can only be implemented through these preferred embodiments. These preferred embodiments are only used to explain the technical principles of the present disclosure, rather than to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts should still fall within the protection scope of the present disclosure.

[0021] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent in such process, method, commodity or device. Without further limitation, the element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, commodity or device including the said element.

[0022] The technical solutions proposed in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] Figure 1 A power consumption monitoring system based on power big data provided for the embodiments of the present application. As Figure 1 shown, the power consumption monitoring system provided in the embodiments of the present application mainly includes: an acquisition module 110, a cleaning module 120, a clustering module 130, and a training module 140.

[0024] Among them, the acquisition module 110 is any feasible device or apparatus capable of acquiring data, etc., and is mainly used to acquire the original data set uploaded by the preset monitored enterprise.

[0025] It should be noted that the original data set includes enterprise basic data and power data; by way of example, the enterprise basic data at least includes enterprise status data, enterprise location data, and enterprise legal person data; the power data at least includes enterprise power customer file data, daily power consumption data, current data, power data, load data, business expansion application data, and capacity reduction data. The preset monitored enterprises are key supervised enterprises, and the specific enterprises corresponding to the key supervised enterprises can be determined by those skilled in the art according to the actual situation.

[0026] Among them, the cleaning module 120 is any feasible device or apparatus capable of cleaning data, etc., and is mainly used to remove the values of the data in the original data set that do not conform to the preset specifications to obtain a first data set; determine the reasonable range of each data in the first data set, and remove the outliers in the first data set that exceed the reasonable range; and then obtain a second data set. By way of example, the preset specification data can be non-zero values, non-negative values, and non-infinite values.

[0027] As an example, the cleaning module 120 includes a box-line unit 121. The box-line unit 121 is used to calculate the reasonable range through the preset box-line formula: LowerLimit = Max{Q 1 - 1.5IQR, Mininum};

[0028] UpperLimit = Min{Q 3 + 1.5IQR, Maxinum}, where any set of data is sorted from smallest to largest and divided into four equal parts. Q1 represents the number at the first equal-part point among the three equal-part points; Q3 represents the number at the third equal-part point among the three equal-part points; IQR represents the value of Q3 - Q1; Minimum represents the smallest number in any set of data, and Maximum represents the largest number in any set of data. It is determined that the range from LowerLimit to UpperLimit is the reasonable range, and then the outliers in the first dataset that exceed the reasonable range are removed.

[0029] Among them, the clustering module 130 is any feasible device or apparatus that can perform data distance, etc. It is mainly used to cluster the power data in the second dataset to cluster the power consumption status corresponding to the preset monitored enterprises; and to determine the status threshold corresponding to the preset monitored enterprises according to the power consumption status and the power data. Among them, the status threshold includes a production shutdown threshold, a normal production threshold, and an over-limit production threshold. As an example, the power consumption status includes that the shutdown enterprise has hidden production while officially shutting down, the shutdown enterprise has hidden production during the day while shutting down at night, the shutdown production line has hidden production while officially shutting down, the enterprise resumes production after shutdown, the enterprise automatically (for a long time) shuts down production, the enterprise has an emergency production shutdown, privately adds production lines / increases power consumption loads (devices), hot work operations, and hot work operations, etc.

[0030] As an example, the clustering module 130 includes a status threshold unit 131; the status threshold unit 131 is used to determine the specific quantity of the second data set corresponding to the preset monitored enterprise; when the specific data is less than the preset quantity threshold, the status threshold corresponding to the preset monitored enterprise is determined through a clustering algorithm for pre-calculating the data center and the class interval; it should be noted that since the amount of power data of some newly opened enterprises is small (the specific data is less than the preset quantity), in the case of insufficient data, to improve the credibility of the threshold result and ensure the generalization ability of the model, the method of using the boundary points of different power consumption states as the threshold is cancelled, and instead, the class center and the class interval are calculated respectively. Through the inter-class relationship of different power consumption states combined with the preset on-site verification experience value, any neural learning network is imported to obtain the rule logic of power data and the determined output threshold. It should be noted that the preset on-site verification experience value can be obtained by those skilled in the art through multiple experiments. According to the neural learning network and the rule logic, the production suspension threshold, the normal production threshold, and the over-limit production threshold are obtained. When the specific data is equal to or greater than the preset quantity threshold, the boundary point corresponding to the power data in the second data set is determined as the status threshold corresponding to the preset monitored enterprise through a pre-calculated boundary point clustering algorithm.

[0031] It should be noted that the process of calculating the threshold by the clustering algorithm for pre-calculating the data center and the class interval and the pre-calculated boundary point clustering algorithm can be implemented by existing technologies, and the present invention does not make excessive limitations on this.

[0032] Among them, the training module 140 is any feasible device or apparatus capable of performing model training, etc., and is mainly used to mark the violation information corresponding to the second data set according to the status threshold and the actually collected violation information; the second data set with the marked violation information is used as the training data to train the bidirectional long short-term memory neural network model; and the power data uploaded by the preset monitored enterprise is alarmed for violation information through the trained bidirectional long short-term memory neural network model. As an example, the violation information includes violation and non-violation; among them, if an enterprise meets any one of the following, the violation information corresponding to the enterprise is a violation: the daily power consumption of a production suspension enterprise is significantly higher than the daily life power consumption, the night power of a production suspension enterprise is significantly higher than the day power consumption, the power consumption suddenly rises, does not exceed the normal peak power consumption, and maintains a high load, the daily power consumption of a production suspension enterprise is continuously higher than the daily life power consumption for multiple days, the daily power of a normal operating enterprise is continuously lower than the daily production power consumption for multiple days, the daily power consumption of a hazardous chemical enterprise drops significantly compared with the previous day, the daily power consumption of an enterprise is significantly higher than the daily production power consumption, the branch line current shows a similar sawtooth-like fluctuation, and the current increase conforms to the welding power of the enterprise, the current of the fire protection special branch line instantaneously rises and maintains a high load, etc. The high load can be determined by those skilled in the art according to the actual situation.

[0033] It should be noted that after the violation information warning (abnormal warning) is issued, this embodiment can also perform warning push, on-site verification, video verification, law enforcement verification feedback, alarm cancellation, and rule optimization on the abnormal warning. Furthermore, a closed-loop management process is realized to ensure the realization of the real-time monitoring function of the electricity-using enterprise.

[0034] In addition, the embodiment of the present application also provides a power consumption monitoring method based on power big data, and its execution entity is a server, such as Figure 2 As shown, the power consumption monitoring method provided by the embodiment of the present application mainly includes the following steps:

[0035] Step 210: Obtain the original data set uploaded by the preset monitored enterprise.

[0036] It should be noted that the original data set includes enterprise basic data and power data; by way of example, the enterprise basic data at least includes enterprise status data, enterprise location data, and enterprise legal person data; the power data at least includes enterprise power customer file data, daily power consumption data, current data, power data, load data, business expansion application data, and capacity reduction data. The preset monitored enterprise is an enterprise under key supervision, and the specific enterprises corresponding to the enterprises under key supervision can be determined by those skilled in the art according to the actual situation.

[0037] Step 220: Remove the values of the data in the original data set that do not conform to the preset specifications to obtain a first data set; determine the reasonable range of each data in the first data set, and remove the abnormal values in the first data set that exceed the reasonable range; thereby obtaining a second data set.

[0038] It should be noted that the box plot rule can be any feasible algorithm, formula, rule, etc. that can calculate the reasonable range of the data set.

[0039] Among them, determining the reasonable range of each data in the first data set and removing the abnormal values in the first data set that exceed the reasonable range can be specifically:

[0040] Through the preset box plot formula: LowerLimit = Max{Q 1 - 1.5IQR, Mininum};

[0041] UpperLimit = Min{Q 3{+1.5IQR, Maximum}, calculate the reasonable range; among them, sort any group of data from small to large and divide it into four equal parts. Q1 represents the number at the first equal division point among the three equal division points; Q3 represents the number at the third equal division point among the three equal division points; IQR represents the value of Q3 - Q1; Minimum represents the smallest number in any group of data, and Maximum represents the largest number in any group of data; determine the range between LowerLimit and UpperLimit as the reasonable range, and then remove the outliers in the first dataset that exceed the reasonable range.

[0042] Step 230: Perform clustering processing on the power data in the second dataset to cluster the power consumption status corresponding to the preset monitored enterprises; and determine the status threshold corresponding to the preset monitored enterprises based on the power consumption status and the power data.

[0043] It should be noted that the status threshold includes a production suspension threshold, a normal production threshold, and an over-limit production threshold.

[0044] Among them, performing clustering processing on the power data in the second dataset to cluster the power consumption status corresponding to the preset monitored enterprises can be specifically:

[0045] Randomly select several power data in the second dataset as sample data through the k-means algorithm, and initially determine that each sample data represents the average value or center μ of a power consumption status j , and mark the power consumption status closest to the sample data as the affiliated power consumption status; that is: Among them, X represents the sample data, i represents the number of sample data; C represents the power consumption status, and j represents the number of centers or average values of the power consumption status; and then determine the power consumption status corresponding to the power data in each second dataset.

[0046] As an example, (1) Randomly select several power data (such as power consumption data) as sample data, and each sample data initially represents the center or average value of a cluster (power consumption status); (2) Mark according to the distance of each power data from the cluster center; (3) Update each cluster center to the mean value of all samples belonging to this cluster; repeat and iterate steps (1), (2), and (3) until the iteration upper limit function converges.

[0047] In addition, determining the status threshold corresponding to the preset monitored enterprises based on the power consumption status and the power data can be specifically:

[0048] Determine the specific quantity of power data in the second dataset corresponding to the preset monitored enterprises; it should be noted that since the power data volume of some newly opened enterprises is small (the specific data is less than the preset quantity), in the case of insufficient data, to improve the credibility of the threshold result and ensure the generalization ability of the model, the method of using the boundary points of different power consumption states as thresholds is cancelled, and instead, the class center and class interval are calculated respectively. By combining the inter-class relationship of different power consumption states with the preset on-site verification empirical value and importing it into any neural learning network, the rule logic of power data and the determined output threshold is obtained. According to this neural learning network and rule logic, the production suspension threshold, normal production threshold, and over-limit production threshold are obtained. When the specific data is less than the preset quantity threshold, the state threshold corresponding to the preset monitored enterprises is determined through the clustering algorithm for preset calculation of the data center and class interval; when the specific data is equal to or greater than the preset quantity threshold, the boundary point corresponding to the power data in the second dataset is determined as the state threshold corresponding to the preset monitored enterprises through the boundary point clustering algorithm for preset calculation.

[0049] It should be noted that the specific process of calculating the threshold by the clustering algorithm for preset calculation of the data center and class interval and the boundary point clustering algorithm for preset calculation can be realized by the existing technology, and the present invention does not limit it too much.

[0050] Step 240: Determine the marked violation information corresponding to the second dataset according to the violation information calculated based on the state threshold and the actually collected violation information; use the second dataset and the marked violation information as training data to train the bidirectional long short-term memory neural network model; use the trained bidirectional long short-term memory neural network model to alarm the violation information for the power data uploaded by the preset monitored enterprises.

[0051] It should be noted that the violation information includes violation and non-violation; among them, the violation information calculated based on the state threshold is mainly whether the power data in the second dataset exceeds the threshold, and when it exceeds the threshold, this second dataset is the violation information.

[0052] It should be noted that the method of importing input data to train the bidirectional long short-term memory neural network model can be realized by the existing method, and this application does not limit it.

[0053] After the violation information alarm (abnormal alarm) is carried out, this embodiment can also perform alarm push, on-site verification, video verification, law enforcement verification feedback, alarm cancellation and rule optimization on this abnormal alarm. Furthermore, a closed-loop management process is realized to ensure the realization of the real-time monitoring function of power consumption enterprises.

[0054] So far, the technical solutions of the present disclosure have been described in combination with multiple embodiments in the foregoing text. However, those skilled in the art can easily understand that the protection scope of the present disclosure is not limited to these specific embodiments. Without departing from the technical principles of the present disclosure, those skilled in the art can split and combine the technical solutions in the above-mentioned various embodiments, and can also make equivalent changes or substitutions to the relevant technical features. Any changes, equivalent substitutions, improvements, etc. made within the technical concept and / or technical principle of the present disclosure will fall within the protection scope of the present disclosure.

Claims

1. An electricity consumption monitoring system based on power big data, characterized in that, the system includes: an acquisition module, configured to acquire an original data set uploaded by a preset monitored enterprise; wherein, the original data set includes enterprise basic data and power data; a cleaning module, configured to remove the values of the data in the original data set that do not conform to the preset specifications to obtain a first data set; determine the reasonable range of each data in the first data set, and remove the outliers in the first data set that exceed the reasonable range; and then obtain a second data set; a clustering module, configured to perform clustering processing on the power data in the second data set to cluster the electricity consumption status corresponding to the preset monitored enterprise; and determine the status threshold corresponding to the preset monitored enterprise according to the electricity consumption status and the power data; wherein, the status threshold includes a production stop threshold, a normal production threshold, and an over-limit production threshold; a training module, configured to determine the marked violation information corresponding to the second data set according to the violation information calculated based on the status threshold and the actually collected violation information; use the second data set and the marked violation information as training data to train a bidirectional long short-term memory neural network model; and perform violation information warning on the power data uploaded by the preset monitored enterprise through the trained bidirectional long short-term memory neural network model; the clustering module includes a status threshold unit; the status threshold unit is configured to determine the specific quantity of the second data set corresponding to the preset monitored enterprise; when the specific quantity is less than the preset quantity threshold, determine the status threshold corresponding to the preset monitored enterprise through a clustering algorithm for calculating the data center and class interval; when the specific quantity is equal to or greater than the preset quantity threshold, determine the boundary point of the power data in the second data set as the status threshold corresponding to the preset monitored enterprise.

2. The electricity consumption monitoring system based on power big data according to claim 1, characterized in that, the cleaning module includes a box plot unit; the box plot unit is configured to use a preset box plot formula: LowerLimit = Max{Q 1 - 1.5IQR, Minimum}; UpperLimit = Min{Q 3 + 1.5IQR, Maximum}, calculate the reasonable range; among them, sort any group of data from small to large and divide it into 4 equal parts, Q1 represents the number at the first equal part point of the three equal part points; Q3 represents the number at the third equal part point of the three equal part points; IQR represents the value of Q3 - Q1; Minimum represents the smallest number in any group of data, and Maximum represents the largest number in any group of data; determine the range from LowerLimit to UpperLimit as the reasonable range, and then remove the outliers in the first dataset that exceed the reasonable range.

3. An electricity consumption monitoring method based on power big data, characterized in that, the method includes: acquiring an original data set uploaded by a preset monitored enterprise; wherein, the original data set includes enterprise basic data and power data; removing the values of the data in the original data set that do not conform to the preset specifications to obtain a first data set; determining the reasonable range of each data in the first data set, and removing the outliers in the first data set that exceed the reasonable range; and then obtaining a second data set; performing clustering processing on the power data in the second data set to cluster the electricity consumption status corresponding to the preset monitored enterprise; and determining the status threshold corresponding to the preset monitored enterprise according to the electricity consumption status and the power data; wherein, the status threshold includes a production stop threshold, a normal production threshold, and an over-limit production threshold; Determine the marked violation information corresponding to the second data set based on the violation information calculated according to the status threshold and the actually collected violation information; use the second data set and the marked violation information as training data to train a bidirectional long short-term memory neural network model; and use the trained bidirectional long short-term memory neural network model to alarm violation information for the power data uploaded by a preset monitored enterprise. Determine the status threshold corresponding to a preset monitored enterprise according to the electricity consumption status and power data, specifically including: Determine the specific quantity of the power data in the second data set corresponding to the preset monitored enterprise. When the specific quantity is less than the preset quantity threshold, determine the status threshold corresponding to the preset monitored enterprise through a clustering algorithm for pre-calculating data centers and class intervals. When the specific quantity is equal to or greater than the preset quantity threshold, determine the boundary point corresponding to the power data in the second data set as the status threshold corresponding to the preset monitored enterprise through a pre-calculated boundary point clustering algorithm.

4. The electricity consumption monitoring method based on power big data according to claim 3, characterized in that Determine the reasonable range of each data in the first data set and remove the outliers in the first data set that exceed the reasonable range, specifically including: By presetting the box line formula: LowerLimit = Max{Q 1 - 1.5IQR, Mininum}; UpperLimit = Min{Q 3 + 1.5IQR, Maximum}, calculate the reasonable range; Among them, sort any group of data from small to large and divide it into 4 equal parts. Q1 represents the number at the first equal part point among the three equal part points; Q3 represents the number at the third equal part point among the three equal part points; IQR represents the value of Q3 - Q1; Minimum represents the smallest number in any group of data, and Maximum represents the largest number in any group of data. Determine the range from LowerLimit to UpperLimit as the reasonable range, and then remove the outliers in the first data set that exceed the reasonable range.

5. The electricity consumption monitoring method based on power big data according to claim 3, characterized in that Perform clustering processing on the power data in the second data set to cluster the electricity consumption status corresponding to the preset monitored enterprise, specifically including: Randomly select a number of power data in the second dataset as sample data through the k-means algorithm, and determine that each sample data initially represents the average or center μ of a power consumption state j , and mark the power consumption state closest to the sample data as the membership power consumption state; that is: Among them, X represents the sample data, and i represents the quantity of the sample data; C represents the electricity consumption status, and j represents the number of electricity consumption status centers or averages. Then determine the electricity consumption status corresponding to the power data in each second data set.

6. The electricity consumption monitoring method based on power big data according to claim 3, characterized in that The basic enterprise data at least includes enterprise status data, enterprise location data, and enterprise legal person data; The power data at least includes enterprise power customer file data, daily electricity consumption data, current data, power data, load data, business expansion application data, and capacity reduction data.

Images