Power distribution dispatching method and system based on artificial intelligence

Abstract

The present invention belongs to the technical field of power distribution dispatching, and specifically relates to a power distribution dispatching method and system based on artificial intelligence, the method comprising: a data preparation module collects power consumption data and condition data, and generates different first training data, and sends the different first training data to a model generation module; the model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, and uses different second training data to repeat the training for estimation model B at different time points after a preset number of dates; the model generation module selects one of the estimation model A at the estimated time point and the estimation model B at the estimated time point as the final estimation model; the scheduling plan generation module formulates a scheduling plan for power generation according to the estimated power consumption data. Through the present invention, even if the power consumption mode changes, the accuracy of the estimated power consumption data can be ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution dispatching, and in particular relates to a power distribution dispatching method and system based on artificial intelligence. Background Art

[0002] By predicting electricity consumption data, power resources can be dispatched more effectively, ensuring that electricity demand is met while avoiding energy waste and improving the efficiency of power grid operation.

[0003] A Chinese invention application with publication number CN107248740A discloses a method for predicting the load of household electrical equipment, the method comprising: collecting environmental data characterizing various environmental states of household electrical equipment; collecting power load data of household electrical equipment under various environmental states; performing data cleaning on the collected power load data; performing attribute value encoding processing on the power load data after data cleaning; constructing more than one autoencoder model, and initializing the autoencoder model parameters in turn; training each autoencoder model in turn; constructing a deep stacked autoencoder model from all the trained autoencoder models, and setting initialization parameters for them; performing deep training on the deep stacked autoencoder model; and using the trained deep stacked autoencoder model to realize power load prediction of household electrical equipment. In addition, a Chinese invention application with publication number CN117895477A proposes a short-term electricity load forecasting method, device and load forecasting system, the method comprising: obtaining first predicted meteorological data corresponding to the target area at the time to be measured and first historical load data in the historical time period before the time to be measured; processing the first historical load data based on at least two time scales to obtain at least two types of second historical load data; each type of second historical load data includes load data corresponding to multiple unit times under the time scale corresponding to the second historical load data; based on at least two types of second historical load data and the first predicted meteorological data, predicting the target load data corresponding to the target area at the time to be measured.

[0004] However, when the electricity consumption pattern changes, the prediction accuracy of the electricity consumption data of the above two invention applications will deteriorate. Therefore, the present invention provides a power distribution scheduling method and system based on artificial intelligence. Summary of the invention

[0005] In the present invention, estimation model A and estimation model B are trained and generated for each time point, and estimation model B is repeatedly trained for different time points after a certain number of days, and one of the estimation model A at the estimated time point and the estimation model B at the estimated time point is selected as the final estimation model, and a scheduling plan for power generation is formulated based on the estimated power consumption data output by the final estimation model. The purpose of the present invention is to ensure the accuracy of the estimated power consumption data when the power consumption mode changes.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides the following artificial intelligence-based power distribution dispatching method, which mainly includes the following steps:

[0007] The data preparation module generates different first training data using the power consumption data and condition data at different time points of each day within a date interval before a preset date, and sends the different first training data to the model generation module;

[0008] The model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, and after a preset number of days, the model generation module uses different second training data to repeatedly train estimation model B for different time points;

[0009] Based on the estimated time point of the estimated date, the model generation module calculates the estimation accuracy of the estimation model A at the estimated time point and the estimation model B at the estimated time point, determines the final estimation model, and sends the final estimation model to the estimation processing module;

[0010] The model generation module calculates the estimation accuracy, including the following steps:

[0011] The model generation module determines a test date interval before the estimated date, and acquires power consumption data and condition data at the estimated time point of each day within the test date interval to generate different test data;

[0012] The model generation module sequentially inputs the conditional data in different test data into the estimation model A at the estimation time point and the estimation model B at the estimation time point, and calculates the estimation accuracy of the estimation model A at the estimation time point and the estimation model B at the estimation time point by the following formula:

[0013] ,

[0014] in, To estimate the accuracy, is the total number of test data, For the The electricity consumption data in the test data, For the The output data of the estimated model corresponding to the conditional data in the test data, for The maximum value among the power consumption data in the test data.

[0015] As a preferred technical solution of the present invention, the estimation processing module uses the final estimation model to estimate the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date to obtain estimated electricity consumption data, and the scheduling plan generation module formulates a scheduling plan for power generation based on the estimated electricity consumption data.

[0016] As a preferred technical solution of the present invention, the model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, including: first classifying different first training data according to different time points to obtain first training data at different time points, and then using the first training data at different time points to train and generate estimation model A and estimation model B at different time points.

[0017] As a preferred technical solution of the present invention, the model generation module uses different second training data to repeatedly train the estimation model B at different time points, including: first classifying the different second training data according to different time points to obtain the second training data at different time points, and then using the second training data at different time points to repeatedly train the estimation model B at different time points.

[0018] As a preferred technical solution of the present invention, the model generation module selects one of the estimation model A at the estimation time point and the estimation model B at the estimation time point as the final estimation model, comprising the following steps:

[0019] The model generation module regards the estimation accuracy of the estimation model A at the estimation time point as the first estimation accuracy, and regards the estimation accuracy of the estimation model B at the estimation time point as the second estimation accuracy;

[0020] The model generation module selects any one of the estimation model A at the estimation time point and the estimation model B at the estimation time point as the final estimation model when both the first estimation accuracy and the second estimation accuracy are less than a preset estimation accuracy threshold; the model generation module regards the estimation model A at the estimation time point as the final estimation model when the first estimation accuracy is less than the estimation accuracy threshold and the second estimation accuracy is greater than or equal to the estimation accuracy threshold; the model generation module regards the estimation model B at the estimation time point as the final estimation model when the first estimation accuracy is greater than or equal to the estimation accuracy threshold and the second estimation accuracy is less than the estimation accuracy threshold; the model generation module obtains different third training data from the data preparation module when both the first estimation accuracy and the second estimation accuracy are greater than or equal to the estimation accuracy threshold, the different third training data are generated by the electricity consumption data and condition data at the estimation time point of each day within a specific date interval before the estimation date, and retrains and generates the estimation model A at the estimation time point and the estimation model B at the estimation time point according to the different third training data.

[0021] As a preferred technical solution of the present invention, after the model generation module retrains and generates the estimation model A at the estimated time point and the estimation model B at the estimated time point, it also includes: every time the number of dates passes, the model generation module obtains different fourth training data from the data preparation module, the different fourth training data are generated by the electricity consumption data and condition data at the estimated time point of each day within the date range corresponding to the passed dates, and the estimation model B at the estimated time point is repeatedly trained using different fourth training data.

[0022] As a preferred technical solution of the present invention, after the model generation module regards the estimation model A at the estimation time point as the final estimation model, it also includes: the model generation module rolls back the model state of the estimation model B at the estimation time point, so that the model state of the estimation model B at the estimation time point is the same as the model state of the estimation model A at the estimation time point.

[0023] As a preferred technical solution of the present invention, after the model generation module regards the estimation model B at the estimation time point as the final estimation model, it also includes: the model generation module repeatedly trains the estimation model A at the estimation time point so that the model state of the estimation model A at the estimation time point is the same as the model state of the estimation model B at the estimation time point.

[0024] As a preferred technical solution of the present invention, the estimation processing module uses the final estimation model to estimate the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date, including the following steps:

[0025] The estimation processing module obtains the condition data of the estimated time point of each day within the date interval before the estimated date from the data preparation module, and determines the first condition data with the smallest value and the second condition data with the largest value among all the condition data, and also calculates the average value of all the condition data;

[0026] The estimation processing module arbitrarily generates an estimation condition data whose value is greater than the value of the first condition data and less than the value of the second condition data, and determines whether the difference between the value of the estimation condition data and the average value is less than a preset difference threshold. If so, the estimation condition data is input into the final estimation model; if not, this step is repeated.

[0027] The present invention also provides a power distribution dispatching system based on artificial intelligence, which mainly includes the following modules:

[0028] The power consumption module includes power consumption equipment, which is used to consume power and generate power consumption data;

[0029] Condition module, used to store condition data;

[0030] A data preparation module, used to collect and store the power consumption data of the power consumption module at different time points every day in the past, and to collect and store the condition data at different time points every day in the past, and to generate and send different first training data and different second training data;

[0031] A model generation module, used to use different first training data to train, generate and store estimation model A and estimation model B for each time point, and used to obtain different second training data from the data preparation module every time a preset number of dates pass, and use different second training data to repeatedly train estimation model B for different time points, and further used to calculate the estimation accuracy of estimation model A at the estimated time point and estimation model B at the estimated time point based on the estimated date, and select one of the estimation model A at the estimated time point and the estimation model B at the estimated time point as the final estimation model;

[0032] An estimation processing module, used to estimate and process the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date using the final estimation model to obtain estimated electricity consumption data;

[0033] The scheduling plan generation module is used to formulate a scheduling plan for power generation based on the estimated power consumption data.

[0034] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0035] In the present invention, first, the data preparation module collects the power consumption data and condition data at different time points of each day in the past, and uses the power consumption data and condition data at different time points of each day within the date interval before the preset date to generate different first training data; secondly, the model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, and after each preset number of dates, the model generation module uses different second training data to repeat the training for estimation model B at different time points; thirdly, the model generation module calculates the estimation accuracy of estimation model A at the estimation time point and estimation model B at the estimation time point, and selects one of the estimation model A at the estimation time point and estimation model B at the estimation time point as the final estimation model; finally, the estimation processing module obtains the estimated power consumption data through the final estimation model, and the scheduling plan generation module formulates the scheduling plan for power generation according to the estimated power consumption data. Through the present invention, even if the power consumption mode of the power module suddenly changes for a short time, the most suitable estimation model for estimation processing can be selected, so as to ensure the accuracy of the estimated power consumption data, and further ensure the smooth formulation of the scheduling plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A flowchart of the steps of the power distribution dispatching method based on artificial intelligence of the present invention;

[0037] Figure 2 This is a structural diagram of the power distribution dispatching system based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0039] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0040] The present invention provides Figure 1 The power distribution dispatching method based on artificial intelligence is mainly implemented by executing the following steps:

[0041] Step 1, the data preparation module collects and stores the power consumption data of the power consumption module at different time points in the past every day, and at the same time, the data preparation module collects and stores the condition data at different time points in the past every day from the condition module, and the data preparation module uses the power consumption data and condition data at different time points every day within a date range before a preset date to generate different first training data, and sends the generated different first training data to the model generation module;

[0042] Step 2, the model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, and stores estimation model A at different time points and estimation model B at different time points respectively, and after each preset number of dates, the model generation module obtains different second training data from the data preparation module, and the different second training data are generated by the power consumption data and condition data at different time points of each day within the date interval corresponding to the passed date, and at the same time, the model generation module uses different second training data to repeatedly train estimation model B at different time points;

[0043] Step 3: Based on the estimated time point of the estimated date, the model generation module calculates the estimation accuracy of the estimation model A at the estimated time point and the estimation model B at the estimated time point, selects one of the estimation model A at the estimated time point and the estimation model B at the estimated time point as the final estimation model according to the different estimation accuracies calculated, and sends the final estimation model to the estimation processing module;

[0044] Furthermore, the estimation processing module uses the final estimation model to estimate the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date to obtain estimated electricity consumption data, and the scheduling plan generation module formulates a scheduling plan for power generation based on the estimated electricity consumption data.

[0045] Specifically, in step 1, the data preparation module collects and stores the electricity consumption data of the electricity consumption module at different time points in the past every day. The different time points can be from 0 o'clock to 24 o'clock. For example, the electricity consumption data of the electricity consumption module at 4 o'clock refers to the electricity consumption data generated by the electricity consumption module as of 4 o'clock in the day. At the same time, the data preparation module collects and stores conditional data at different time points in the past every day from the conditional module. For example, the conditional data can be the temperature. The data preparation module also uses the electricity consumption data and conditional data at different time points of each day within the date range before the preset date to generate different first training data, and the generated different The same first training data is sent to the model generation module, and the preset date is specifically set in the actual application scenario. For example, if the preset date is 2024.06.30, then every day in the date range before the preset date refers to every day before 2024.06.30. The content of each first training data includes power consumption data and condition data. The power consumption data is the power consumption data at a certain time point on a certain date. The date and time point corresponding to the condition data are respectively the same as the date and time point corresponding to the power consumption data. In step 2, the model generation module uses different first training data to train each time point separately. Generate and store estimation model A and estimation model B. Estimation model A and estimation model B can use machine learning models in the prior art, which will not be described in detail. Every time a preset number of dates passes, which refers to the number of dates from a preset date, the preset number of dates is specifically set in an actual application scenario, for example, 5 days. The model generation module obtains different second training data from the data preparation module. The content of the second training data also includes electricity consumption data and corresponding condition data. Different second training data are generated by electricity consumption data and condition data at different time points of each day within the date interval corresponding to the passed date. For example, the preset number of dates is 5 days, and the date interval corresponding to the passed date is the date interval corresponding to the 5 days that have passed. The model generation module uses different second training data to repeatedly train estimation model B at different time points. In step 3, according to the estimated time point of the estimated date, the model generation module calculates the estimation accuracy of estimation model A and estimation model B at the estimated time point, and selects one of them from estimation model A at the estimated time point and estimation model B at the estimated time point according to the different estimation accuracies calculated, as the final estimation model, and sends the final estimation model to the estimation processing module. Afterwards, the estimation processing module uses the final estimation model to estimate the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date to obtain the estimated electricity consumption data, so that the scheduling plan generation module can formulate a scheduling plan according to the estimated electricity consumption data. The content of the scheduling plan includes the combined working methods of different generating sets, the purpose of which is to produce electricity that meets the estimated electricity consumption data without causing waste of resources.

[0046] Furthermore, the model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, including: first classifying different first training data according to different time points to obtain first training data at different time points, and then using the first training data at different time points to train and generate estimation model A and estimation model B at different time points.

[0047] Specifically, this introduces how to use different first training data to train and generate estimation model A and estimation model B for each time point. First, different first training data are classified and processed according to different time points to obtain first training data for different time points. For example, the first training data at 6 o'clock includes the first training data corresponding to 6 o'clock on different dates. Secondly, the first training data at different time points are used to train and generate estimation models A and estimation models B for different time points. For example, the first training data at 6 o'clock is used to train and generate estimation model A at 6 o'clock and estimation model B at 6 o'clock. At this time, the model states of the two are the same, which can be understood as the various parameters in the model are the same.

[0048] Furthermore, the model generation module uses different second training data to repeatedly train the estimation model B at different time points, including: first classifying the different second training data according to different time points to obtain the second training data at different time points, and then using the second training data at different time points to repeatedly train the estimation model B at different time points.

[0049] Specifically, here we introduce how to use different second training data to repeatedly train the estimation model B at different time points. First, the different second training data are classified and processed according to different time points to obtain the second training data at different time points. This process is the same as the process of obtaining the first training data at different time points. Secondly, the second training data at different time points are used to re-train the estimation model B at different time points. For example, the second training data at 6 points is used to re-train the estimation model B at 6 points. At this time, due to the re-training, the model state of the estimation model B at 6 points and the model state of the estimation model A at 6 points are different.

[0050] Further, based on the estimated time point of the estimated date, the model generation module calculates the estimation accuracy of the estimation model A at the estimated time point and the estimation model B at the estimated time point, including the following steps:

[0051] Step 1: The model generation module determines a test date interval before the estimated date, and the model generation module obtains power consumption data and condition data at the estimated time point of each day within the test date interval to generate different test data;

[0052] Step 2: The model generation module sequentially inputs the conditional data in different test data into the estimation model A at the estimation time point and the estimation model B at the estimation time point, and calculates the estimation accuracy of the estimation model A at the estimation time point and the estimation model B at the estimation time point by the following formula:

[0053] ,

[0054] in, To estimate the accuracy, is the total number of test data, For the The electricity consumption data in the test data, For the The output data of the estimated model corresponding to the conditional data in the test data, for The maximum value among the power consumption data in the test data.

[0055] Specifically, in step 1, the model generation module determines the inspection date interval before the estimated date. The inspection date interval can be set in the actual application scenario. For example, the inspection date interval is two weeks before the estimated date. The model generation module obtains the electricity consumption data and condition data at the estimated time point of each day within the inspection date interval, and generates different inspection data. The content of the inspection data also includes electricity consumption data and corresponding condition data. In step 2, the model generation module takes the condition data in all the inspection data as input data in turn, so that the estimation model A at the estimated time point and the estimation model B at the estimated time point generate output data respectively, and then uses the above formula to calculate the estimation accuracy of the estimation model A at the estimated time point and the estimation model B at the estimated time point. It should be noted that The smaller the value of , the better the performance of the estimation model in the estimation process. The larger the value of , the worse the performance of the estimation model in the estimation process.

[0056] Furthermore, the model generation module selects one of the estimation model A at the estimation time point and the estimation model B at the estimation time point as the final estimation model according to the different estimation accuracies calculated, including the following steps:

[0057] Step 1: The model generation module regards the estimation accuracy of the estimation model A at the estimation time point as the first estimation accuracy, and regards the estimation accuracy of the estimation model B at the estimation time point as the second estimation accuracy;

[0058] Step 2, when the first estimation accuracy and the second estimation accuracy are both less than the preset estimation accuracy threshold, the model generation module selects any one of the estimation model A at the estimation time point and the estimation model B at the estimation time point as the final estimation model; when the first estimation accuracy is less than the estimation accuracy threshold and the second estimation accuracy is greater than or equal to the estimation accuracy threshold, the model generation module regards the estimation model A at the estimation time point as the final estimation model; when the first estimation accuracy is greater than or equal to the estimation accuracy threshold and the second estimation accuracy is less than the estimation accuracy threshold, the model generation module regards the estimation model B at the estimation time point as the final estimation model; when the first estimation accuracy and the second estimation accuracy are both greater than or equal to the estimation accuracy threshold, the model generation module obtains different third training data from the data preparation module, the different third training data are generated by the electricity consumption data and condition data at the estimation time point of each day within a specific date interval before the estimation date, and the estimation model A at the estimation time point and the estimation model B at the estimation time point are retrained and generated according to the different third training data.

[0059] Specifically, in step 1, the model generation module regards the estimation accuracy of the estimation model A at the estimation time point as the first estimation accuracy, and regards the estimation accuracy of the estimation model B at the estimation time point as the second estimation accuracy. In step 2, the model generation module processes four situations. The first situation is that both the first estimation accuracy and the second estimation accuracy are less than the estimation accuracy threshold. The estimation accuracy threshold can be set in the actual application scenario. Any one of the estimation model A at the estimation time point and the estimation model B at the estimation time point is selected as the final estimation model, because both have good performance in estimation processing. The estimation model corresponding to the smaller estimation accuracy can also be selected. The second situation is that the first estimation accuracy is less than the estimation accuracy threshold, and the second estimation accuracy is greater than or equal to the estimation accuracy threshold. The estimation model A at the estimation time point is regarded as the final estimation model because it performs better in estimation processing. The third situation is that the first estimation accuracy The first estimation accuracy is greater than or equal to the estimation accuracy threshold, the second estimation accuracy is less than the estimation accuracy threshold, the estimation model B at the estimation time point that performs better during the estimation processing is used as the final estimation model. The fourth case is that the first estimation accuracy and the second estimation accuracy are both greater than or equal to the estimation accuracy threshold, different third training data are obtained from the data preparation module, and the estimation model A at the estimation time point and the estimation model B at the estimation time point are retrained and generated according to the different third training data. This is done because the performance of both in the estimation processing is poor and should not be used for estimation processing, that is, the estimation processing cannot be performed this time. Among them, different third training data are generated by the electricity consumption data and conditional data at the estimation time point of each day within a specific date interval before the estimation date. The specific date interval can be set in the actual application scenario. For example, the specific date interval is two months before the estimation date, and the content of the third training data still includes electricity consumption data and corresponding conditional data.

[0060] Furthermore, after the model generation module retrains to generate the estimation model A for the estimated time point and the estimation model B for the estimated time point, it also includes: every time a certain number of dates pass, the model generation module obtains different fourth training data from the data preparation module, and the different fourth training data are generated by the electricity consumption data and condition data at the estimated time point of each day within the date range corresponding to the passed dates, and the estimation model B for the estimated time point is repeatedly trained using the different fourth training data.

[0061] Specifically, following the fourth situation mentioned above, after retraining and generating the estimation model A for the estimated time point and the estimation model B for the estimated time point, every time the above number of dates passes, which refers to the above number of dates from the estimated date, the model generation module obtains different fourth training data from the data preparation module. The content of the fourth training data still includes electricity consumption data and condition data. The different fourth training data are generated by the electricity consumption data and condition data at the estimated time point of each day within the date range corresponding to the passed dates. For example, if the number of dates is 7 days, the date range corresponding to the passed dates is the date range corresponding to the 7 days that have passed. The model generation module uses different fourth training data to repeat the training for the estimation model B for the estimated time point.

[0062] Furthermore, after the model generation module regards the estimation model A at the estimation time point as the final estimation model, it also includes: the model generation module rolls back the model state of the estimation model B at the estimation time point, so that the model state of the estimation model B at the estimation time point is the same as the model state of the estimation model A at the estimation time point.

[0063] Specifically, the reason for the occurrence of the second situation mentioned above is believed to be that during the repeated training of the estimation model B at the estimation time point, the power usage mode of the power usage module accidentally changed. For example, the power usage equipment included in the power usage module failed. During this period, the estimation model B at the estimation time point learned the intrinsic connection between the conditional data and the power usage data when the power usage mode changed, which is different from the previously learned intrinsic connection between the conditional data and the power usage data when the power usage mode did not change, resulting in the estimation accuracy of the calculated estimation model B at the estimation time point deteriorating. Therefore, the model generation module can roll back the model state of the estimation model B at the estimation time point, so that the model state of the estimation model B at the estimation time point is the same as the model state of the estimation model A at the estimation time point.

[0064] Furthermore, after the model generation module regards the estimation model B at the estimation time point as the final estimation model, it also includes: the model generation module repeatedly trains the estimation model A at the estimation time point so that the model state of the estimation model A at the estimation time point is the same as the model state of the estimation model B at the estimation time point.

[0065] Specifically, the reason for the occurrence of the third situation mentioned above is believed to be that during the repeated training of the estimation model B at the estimation time point, the electricity consumption mode of the electricity consumption module has not changed, and the repeated training of the estimation model B at the estimation time point has improved the calculated estimation accuracy. Therefore, the model generation module performs the same repeated training on the estimation model A at the estimation time point as the repeated training on the estimation model B at the estimation time point, so that the model state of the estimation model A at the estimation time point is the same as the model state of the estimation model B at the estimation time point.

[0066] Furthermore, the estimation processing module uses the final estimation model to perform estimation processing on the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date, including the following steps:

[0067] Step 1, the estimation processing module obtains the condition data of the estimation time point of each day within the date interval before the estimation date from the data preparation module, and determines the first condition data with the smallest value and the second condition data with the largest value among all the condition data, and also calculates the average value of all the condition data;

[0068] Step 2: The estimation processing module arbitrarily generates an estimated condition data whose value is greater than the value of the first condition data and less than the value of the second condition data, and determines whether the difference between the value of the estimated condition data and the average value is less than a preset difference threshold. If so, the estimated condition data is input into the final estimation model; if not, repeat this step.

[0069] Specifically, according to the above content, actual conditional data are used when training the estimation model. However, if the actual conditional data is input into the final estimation model to obtain the estimated power consumption data at the estimated time point of the estimated date, the estimated power consumption data will be obtained too late, and there will not be sufficient time to formulate a scheduling plan. Therefore, by estimating the conditional data at the estimated time point of the estimated date through the above method, the estimated power consumption data can be obtained in advance, thereby leaving sufficient time to formulate a scheduling plan. In step 1, the estimation processing module obtains the conditional data at the estimated time point of each day within the date range before the estimated date from the data preparation module. Here, the conditional data at the estimated time point before the estimated date is obtained. The date range can be determined according to the actual application scenario. For example, half a month before the estimated date, the first condition data with the smallest value and the second condition data with the largest value are determined among all the condition data, and the average value of all the condition data is calculated. In step 2, the estimation processing module arbitrarily generates an estimated condition data with a value greater than the value of the first condition data and less than the value of the second condition data, and determines whether the difference between the value of the estimated condition data and the average value is less than the difference threshold. The difference threshold can be set in the actual application scenario. If so, the estimated condition data is input into the final estimation model. Otherwise, this step is repeated until the estimated condition data that meets the requirements is generated.

[0070] References Figure 2 As shown, the present invention also provides an artificial intelligence-based distribution scheduling system, including a power consumption module, a condition module, a data preparation module, a model generation module, an estimation processing module, and a scheduling plan generation module, which is used to implement the artificial intelligence-based distribution scheduling method described above.

[0071] The functions of each module are as follows:

[0072] The power consumption module includes power consumption equipment, which is used to consume power and generate power consumption data;

[0073] Condition module, used to store condition data;

[0074] A data preparation module, used to collect and store the power consumption data of the power consumption module at different time points every day in the past, and to collect and store the condition data at different time points every day in the past, and to generate and send different first training data and different second training data;

[0075] A model generation module, used to use different first training data to train, generate and store estimation model A and estimation model B for each time point, and used to obtain different second training data from the data preparation module every time a preset number of dates pass, and use different second training data to repeatedly train estimation model B for different time points, and further used to calculate the estimation accuracy of estimation model A at the estimated time point and estimation model B at the estimated time point based on the estimated date, and select one of the estimation model A at the estimated time point and the estimation model B at the estimated time point as the final estimation model;

[0076] An estimation processing module, used to estimate and process the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date using the final estimation model to obtain estimated electricity consumption data;

[0077] The scheduling plan generation module is used to formulate a scheduling plan for power generation based on the estimated power consumption data.

[0078] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0079] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0080] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0081] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0082] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power distribution dispatching method based on artificial intelligence, characterized in that: The steps include: The data preparation module generates different first training data using the power consumption data and condition data at different time points of each day within a date interval before a preset date, and sends the different first training data to the model generation module; The model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, and after a preset number of days, the model generation module uses different second training data to repeatedly train estimation model B for different time points; Based on the estimated time point of the estimated date, the model generation module calculates the estimation accuracy of the estimation model A at the estimated time point and the estimation model B at the estimated time point, determines the final estimation model, and sends the final estimation model to the estimation processing module; The model generation module calculates the estimation accuracy, including the following steps: The model generation module determines a test date interval before the estimated date, and acquires power consumption data and condition data at the estimated time point of each day within the test date interval to generate different test data; The model generation module sequentially inputs the conditional data in different test data into the estimation model A at the estimation time point and the estimation model B at the estimation time point, and calculates the estimation accuracy of the estimation model A at the estimation time point and the estimation model B at the estimation time point by the following formula: ， in, To estimate the accuracy, is the total number of test data, For the The electricity consumption data in the test data, For the The output data of the estimated model corresponding to the conditional data in the test data, for The maximum value among the power consumption data in the test data; The model generation module selects one of the estimation model A at the estimation time point and the estimation model B at the estimation time point as the final estimation model, comprising the following steps: The model generation module regards the estimation accuracy of the estimation model A at the estimation time point as the first estimation accuracy, and regards the estimation accuracy of the estimation model B at the estimation time point as the second estimation accuracy; The model generation module selects any one of the estimation model A at the estimation time point and the estimation model B at the estimation time point as the final estimation model when both the first estimation accuracy and the second estimation accuracy are less than a preset estimation accuracy threshold; the model generation module regards the estimation model A at the estimation time point as the final estimation model when the first estimation accuracy is less than the estimation accuracy threshold and the second estimation accuracy is greater than or equal to the estimation accuracy threshold; the model generation module regards the estimation model B at the estimation time point as the final estimation model when the first estimation accuracy is greater than or equal to the estimation accuracy threshold and the second estimation accuracy is less than the estimation accuracy threshold; the model generation module obtains different third training data from the data preparation module when both the first estimation accuracy and the second estimation accuracy are greater than or equal to the estimation accuracy threshold, the different third training data are generated by the electricity consumption data and condition data at the estimation time point of each day within a specific date interval before the estimation date, and retrains and generates the estimation model A at the estimation time point and the estimation model B at the estimation time point according to the different third training data.

2. The method according to claim 1, characterized in that The estimation processing module uses the final estimation model to estimate the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date to obtain estimated electricity consumption data, and the scheduling plan generation module formulates a scheduling plan for power generation based on the estimated electricity consumption data.

3. The method according to claim 1, characterized in that: The model generation module uses different first training data to train and generate estimation model A and estimation model B for each time point, including: first classifying different first training data according to different time points to obtain first training data at different time points, and then using the first training data at different time points to train and generate estimation model A and estimation model B at different time points.

4. The method according to claim 1, characterized in that: The model generation module uses different second training data to repeatedly train the estimation model B at different time points, including: first classifying the different second training data according to different time points to obtain the second training data at different time points, and then using the second training data at different time points to repeatedly train the estimation model B at different time points.

5. The method according to claim 1, characterized in that After the model generation module retrains to generate the estimation model A for the estimated time point and the estimation model B for the estimated time point, it also includes: every time the number of dates passes, the model generation module obtains different fourth training data from the data preparation module, the different fourth training data are generated by the electricity consumption data and condition data at the estimated time point of each day within the date range corresponding to the passed dates, and the estimation model B for the estimated time point is repeatedly trained using the different fourth training data.

6. The method according to claim 1, characterized in that After the model generation module regards the estimation model A at the estimation time point as the final estimation model, it also includes: the model generation module rolls back the model state of the estimation model B at the estimation time point, so that the model state of the estimation model B at the estimation time point is the same as the model state of the estimation model A at the estimation time point.

7. The method according to claim 1, characterized in that After the model generation module regards the estimation model B at the estimation time point as the final estimation model, it also includes: the model generation module repeatedly trains the estimation model A at the estimation time point so that the model state of the estimation model A at the estimation time point is the same as the model state of the estimation model B at the estimation time point.

8. The method according to claim 2, characterized in that: The estimation processing module uses the final estimation model to perform estimation processing on the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date, including the following steps: The estimation processing module obtains the condition data of the estimated time point of each day within the date interval before the estimated date from the data preparation module, and determines the first condition data with the smallest value and the second condition data with the largest value among all the condition data, and also calculates the average value of all the condition data; The estimation processing module arbitrarily generates an estimation condition data whose value is greater than the value of the first condition data and less than the value of the second condition data, and determines whether the difference between the value of the estimation condition data and the average value is less than a preset difference threshold. If so, the estimation condition data is input into the final estimation model; if not, this step is repeated.

9. A power distribution dispatching system based on artificial intelligence, used to implement the method according to any one of claims 1 to 8, characterized in that: Includes the following modules: The power consumption module includes power consumption equipment, which is used to consume power and generate power consumption data; Condition module, used to store condition data; A data preparation module, used to collect and store the power consumption data of the power consumption module at different time points every day in the past, and to collect and store the condition data at different time points every day in the past, and to generate and send different first training data and different second training data; A model generation module, used to use different first training data to train, generate and store estimation model A and estimation model B for each time point, and used to obtain different second training data from the data preparation module every time a preset number of dates pass, and use different second training data to repeatedly train estimation model B for different time points, and further used to calculate the estimation accuracy of estimation model A at the estimated time point and estimation model B at the estimated time point based on the estimated date, and select one of the estimation model A at the estimated time point and the estimation model B at the estimated time point as the final estimation model; An estimation processing module, used to estimate and process the electricity consumption data of the electricity consumption module at the estimated time point of the estimated date using the final estimation model to obtain estimated electricity consumption data; The scheduling plan generation module is used to formulate a scheduling plan for power generation based on the estimated power consumption data.

Images