Intelligent energy-saving power distribution method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of intelligent power distribution, and discloses an intelligent energy-saving power distribution method and system based on artificial intelligence. Comprising the steps that the total electricity utilization power of a user and the electricity utilization power of all sub-devices are collected in real time through an intelligent electric meter, environment data (temperature, humidity and illumination intensity) and historical meteorological data (temperature, humidity and rainfall probability) of the user are obtained at the same time, and a data set marked by timestamps is constructed in combination with electricity utilization preference of the user and the maximum power supply capacity of a region; and training the first machine learning model based on a sliding window method to realize short-term load prediction, and training the second machine learning model to realize medium and long-term load prediction. And combining output of the two models, calculating a load evaluation value, evaluating the load condition in real time through the evaluation value, and further making a standby power supply starting and load regulation and control strategy.

Description

Technical Field

[0001] The present invention relates to the field of intelligent power distribution technology, and more specifically, to an intelligent energy-saving power distribution method and system based on artificial intelligence. Background Art

[0002] Power distribution refers to the process of transmitting high-voltage electrical energy from the transmission system to end users, including the voltage reduction, distribution and control of electrical energy.

[0003] The Chinese patent with the authorization announcement number CN117010573B discloses an intelligent power distribution energy-saving optimization method and system, which obtains basic information of equipment by consulting power consumption files, analyzes them to draw power consumption curves, determines key energy-saving points based on the curves, and extracts the usage and power data of related equipment. These data are used to optimize the power distribution coefficient, obtain the optimal power distribution plan, and improve the efficiency of the power distribution system.

[0004] The Chinese patent with the authorization announcement number CN113162238B discloses an incoming line cabinet for an energy-saving power distribution device. The device realizes intelligent control and management of the incoming line cabinet by integrating multiple functional modules such as a control module, a conversion module, and a contact module. The control module is responsible for directing the operation of each component in the incoming line cabinet, and the contact module is responsible for controlling the connectivity between the incoming line cabinet and the external power grid. It can remotely operate the on and off of the output circuit of the incoming line cabinet, and can automatically identify and cut off the circuit of non-continuous power supply equipment, thereby reducing standby energy consumption.

[0005] However, the above invention fails to more flexibly carry out macro-control of long-term and short-term electricity consumption, fails to reduce the investment in backup power supply, and fails to reduce the waste of resources caused by excessive power generation.

[0006] In view of this, the present invention proposes an intelligent energy-saving power distribution method and system based on artificial intelligence to solve the above problems. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an intelligent energy-saving power distribution method and system based on artificial intelligence.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] Intelligent energy-saving power distribution method based on artificial intelligence;

[0010] Collect real-time power consumption data, user environment data, historical meteorological data, power preference data and regional maximum power supply capacity. Real-time power consumption data includes total power consumption and sub-device power consumption per minute; user environment data includes temperature, humidity and light intensity;

[0011] Real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data are timestamped and used as data sets; the timestamp includes the characteristics of time, quarter, and whether it is a weekday;

[0012] The first machine learning model trained based on the dataset to predict short-term total user power consumption in real time;

[0013] A second machine learning model is trained based on the data set to predict the total power consumption of users in the medium and long term in real time;

[0014] Calculate a load evaluation value based on the first machine learning model and the second machine learning model;

[0015] The load evaluation value is compared with the preset load evaluation threshold. When the load evaluation value is greater than or equal to the load evaluation threshold, the backup power supply is enabled; when the load evaluation value is less than 90% of the load evaluation threshold but greater than or equal to 80% of the load evaluation threshold, a load regulation strategy is adopted.

[0016] Preferably, the sub-devices are different types of electrical equipment of the user, which are divided into: automobile electricity, air conditioner and refrigerator electricity, heating electricity and small household appliances electricity;

[0017] Historical meteorological data includes the temperature, humidity and rainfall probability for the next 24 hours. The temperature and humidity in the historical meteorological data represent the temperature and humidity of the user's geographical environment.

[0018] The power usage preference data is data reflecting the user's habitual power usage time. The power usage preference data is a set of vector data, which maps different types of sub-devices to numbers as the x component of the vector; the y component and the z component are the preferred start and end time of the sub-device;

[0019] The maximum power supply capacity of a region is the maximum amount of power that the power system can provide in a unit time period during the power distribution process;

[0020] The timestamp marking process includes: formatting the data into a unified structure; appending unified time features; and adding features of quarter and whether it is a working day.

[0021] Preferably, the training process of the first machine learning model includes:

[0022] Use sliding window method to convert the data set into training set;

[0023] The data set is divided into n time windows of equal size. Multiple sets of real-time power consumption data, user environment data, historical meteorological data and power consumption preference data in each window are taken as a sample. The total power consumption of users at the future k moments in the window is taken as a label. One sample corresponds to one label. One sample and the corresponding label constitute a set of training data. Multiple sets of training data constitute a training set.

[0024] A data set contains 12 hours of real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data. Each hour is regarded as a time step. The 12 hours of real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data are marked as a feature data set A, where A = {A1, A2, A3, A4, A5..., A 12}, A i It includes the real-time power consumption data, user environment data, historical meteorological data and power consumption preference data of the ith hour. The length of the sliding window is defined as 6, so 6 hours of data is a sample. The total power consumption of users in the next hour is predicted, that is, k = 1, then A1, A2, A3, A4, A5 and A6 are a group of samples as the input of the first machine learning model, and the total power consumption of users in the 7th hour is used as the sample label.

[0025] Using the training set generated by the sliding window method as the input of the first machine learning model, the first machine learning model takes the future total power consumption prediction value as the output, takes the real-time sample as the prediction target, and takes minimizing the loss function value of the first machine learning model as the training target; when the loss function value of the first machine learning model is less than or equal to the preset first target loss value, the training is stopped;

[0026] The first machine learning model loss function is a mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the sample group number; u is the number of sample groups; y i is the label corresponding to the i-th group of samples, The total power consumption predicted for the i-th group of samples; the first machine learning model is a recurrent neural network model;

[0027] Preferably, the training process of the second machine learning model includes:

[0028] The training set is converted into a feature matrix and projected into a high-dimensional embedding space with a dimension of d;

[0029] Add position encoding to each time step; use sine and cosine functions to encode position information;

[0030] Compute query Q, key K and value V based on time step;

[0031] Calculate the attention score: d k is the dimension of K;

[0032] The attention scores are concatenated and mapped to the original dimension through linear transformation to output the total power consumption of users in the medium and long term. The step size of the second machine learning model is 12 hours and the time window is 7. The second machine learning model is a transformer model.

[0033] Preferably, the load rating value Where D1(t) represents the short-term total power consumption of users predicted by the first machine learning model at time t, P total (t) is the maximum power supply capacity of the region at time t; D2(t) is the total power consumption of users in the medium and long term predicted by the second machine learning model at time t; is the average maximum power supply capacity in the future time window T; w1 is 0.8; w2 is 0.2.

[0034] Preferably, the load regulation strategy includes: recommending users to delay power operation to a period of low power consumption through the APP, and providing a low-price electricity rate strategy.

[0035] An intelligent energy-saving power distribution system based on artificial intelligence, the system comprising:

[0036] Data collection module, which collects real-time power consumption data, user environment data, historical meteorological data, power preference data and regional maximum power supply capacity. Real-time power consumption data includes total power consumption and sub-device power consumption per minute; user environment data includes temperature, humidity and light intensity;

[0037] The data preprocessing module timestamps the real-time power consumption data, user environment data, historical meteorological data, and power consumption preference data as a data set; the timestamp includes the characteristics of time, quarter, and whether it is a weekday;

[0038] A short-term model training module, which trains the first machine learning model for real-time prediction of the total power consumption of short-term users based on the data set;

[0039] The medium- and long-term model training module trains a second machine learning model based on the data set to predict the total power consumption of users in the medium and long term in real time;

[0040] A post-processing module, calculating a load evaluation value based on the first machine learning model and the second machine learning model;

[0041] The decision module compares the load evaluation value with the preset load evaluation threshold. When the load evaluation value is greater than or equal to the load evaluation threshold, the backup power supply is enabled; when the load evaluation value is less than 90% of the load evaluation threshold but greater than or equal to 80% of the load evaluation threshold, a load regulation strategy is adopted.

[0042] An electronic device comprises: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-mentioned intelligent energy-saving power distribution method based on artificial intelligence by calling the computer program stored in the memory.

[0043] A computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned intelligent energy-saving power distribution method based on artificial intelligence.

[0044] The technical effects and advantages of the intelligent energy-saving power distribution method and system based on artificial intelligence of the present invention are as follows: the total power consumption of users and the power consumption of each sub-device are collected in real time through smart meters, and the user's environmental data (temperature, humidity and light intensity) and historical meteorological data (temperature, humidity and rainfall probability) are obtained at the same time, and a timestamp-marked data set is constructed in combination with the user's power preferences and the maximum power supply capacity of the region. The first machine learning model is trained based on the sliding window method to achieve short-term load forecasting, and the second machine learning model is trained to achieve medium- and long-term load forecasting. Combining the outputs of the two models, the load evaluation value is calculated, and the load status is evaluated in real time through the evaluation value, so as to formulate the backup power supply activation and load regulation strategy.

[0045] This application realizes the short-term dynamic response and medium- and long-term trend analysis of power load, and provides intelligent support for user power management and power grid load regulation. The model's ability to learn short-term dynamics and long-term time series rules has been significantly improved, and the short-term and medium- and long-term load forecasting models have been combined to achieve accurate and comprehensive load forecasting, adapting to complex and changing load scenarios. The power load status is evaluated in real time through the load evaluation value, and the activation of backup power sources is accurately controlled to improve the grid dispatch efficiency and user experience. The introduction of power preference data makes the prediction model more personalized and refined, and the load regulation strategy avoids overload risks, improves the reliability of power grid operation, and reduces the operating cost of the power supply system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a smart energy-saving power distribution method based on artificial intelligence and a schematic diagram of the method of the present invention;

[0047] Figure 2 A schematic diagram of an intelligent energy-saving power distribution method and system based on artificial intelligence of the present invention;

[0048] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application;

[0049] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present application. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Example 1

[0052] See also Figure 1 As shown, the intelligent energy-saving power distribution method based on artificial intelligence described in this embodiment includes:

[0053] Collect real-time power consumption data, including total power consumption and sub-device power consumption per minute, which is obtained through smart meters installed on the user side;

[0054] Sub-devices are different types of electrical equipment used by users; specifically, they are divided into: automobile electricity, air conditioner and refrigerator electricity, heating electricity (such as ovens and electric hair dryers, etc.) and small household appliances electricity;

[0055] Collect user environment data; user environment data includes temperature, humidity and light intensity;

[0056] The temperature is obtained by a temperature sensor installed at the user end; the humidity is obtained by a humidity sensor installed at the user end; and the light intensity is obtained by a photoresistor sensor installed at the user end.

[0057] The total power consumption data reflects the dynamic fluctuation of the user's overall power load and provides time series characteristics for short-term load forecasting.

[0058] The power consumption of sub-devices can identify high-energy-consuming devices and provide clear targets for control priorities. The power consumption of sub-devices can also refine the input features of the load forecasting model and help more accurately learn the contribution of each type of equipment to the total power.

[0059] Temperature changes are an important driving factor for the use of equipment such as air conditioning and heating. Collecting temperature data can significantly improve the accuracy of load forecasting.

[0060] Humidity affects the operating efficiency of refrigeration equipment. Higher humidity will increase the dehumidification burden of the air conditioner, thereby increasing power consumption. When humidity decreases, the energy consumption of the air conditioner in cooling mode decreases, which has a reducing effect on the overall load. Combined with humidity information, the overall load capacity can be evaluated with higher accuracy.

[0061] The light intensity takes special circumstances into consideration, such as when the user is not at home at night. When the light intensity is reduced and the power of small household appliances is not in use, it may reflect that the user is not at home. In this case, the judgment of the power load is more accurate for the special circumstances of whether the user is at home or not.

[0062] Real-time power consumption data and user environment data are uploaded to the cloud database through the wireless communication module;

[0063] Collect historical meteorological data, including temperature, humidity and rainfall probability in the next 24 hours; the temperature and humidity in the historical meteorological data represent the temperature and humidity of the user's geographical environment.

[0064] Historical meteorological data is automatically pulled through the weather platform API interface and updated every hour; obtaining historical meteorological data can effectively capture the impact of the environment on electricity demand and improve the accuracy of model predictions; rainfall will affect people’s travel patterns and may lead to increased household electricity consumption (for example, the time spent using electrical appliances at home on rainy days increases).

[0065] Collect electricity preference data. Electricity preference data is data reflecting the user's habitual electricity usage time. It is obtained by users filling out electronic questionnaires. The electricity preference data is uploaded through the mobile app and stored as time series features to enhance the model's ability to learn personalized behaviors.

[0066] The power usage preference data is a set of vector data, which maps different types of sub-devices to numbers as the x component of the vector; the y component and the z component are the preferred start and end time of the sub-device;

[0067] For example, for a sub-device of type automobile power, the mapping number is 1, and the user prefers to start using automobile power at 20:30 and end automobile power at 07:00, then the power preference data is (1, 2030, 0700);

[0068] It should be noted that each user has multiple sets of electricity usage preference data, and the electricity usage preference data of multiple users are collected.

[0069] Collect the maximum power supply capacity of the region; obtain the specific power supply capacity in real time from the power grid operation system;

[0070] The maximum power supply capacity of a region is the maximum amount of power that the power system can provide in a unit time period during the power distribution process;

[0071] The real-time power consumption data, user environment data, historical meteorological data and power consumption preference data are timestamped, including the characteristics of time, season and whether it is a weekday; the addition of time characteristics can enhance the model's learning ability for load periodicity and seasonal changes and improve prediction accuracy; the real-time power consumption data, user environment data, historical meteorological data and power consumption preference data with timestamps are used as data sets;

[0072] The timestamp marking process includes: formatting the data into a unified structure; appending unified time features; adding features of quarter and whether it is a working day;

[0073] Exemplary codes for the working day / holiday information include:

[0074] from datetime import datetime

[0075] now = datetime.now()

[0076] quarter=(now.month-1) / / 3+1

[0077] weekday=now.strftime("%A")

[0078] is_holiday=check_if_holiday(now)

[0079] timestamp=f″(now.year}-Q{quarter}-{now.day}({weekday if not is_holiday else′holiday′}){now.hour}:{now.minute}:{now.second}″

[0080] The first machine learning model for real-time prediction of short-term user total power consumption (short-term user total load) is trained based on the data set;

[0081] The training process of the first machine learning model includes:

[0082] Use sliding window method to convert the data set into training set;

[0083] The present invention provides the following example of a sliding window method: a data set is divided into n time windows of equal size, and multiple groups of real-time power consumption data, user environment data, historical meteorological data, and power consumption preference data in each window are taken as a sample, and the total power consumption of users at k moments in the future of the window is taken as a label. It should be noted that one sample corresponds to one label, one sample and the corresponding label constitute a set of training data, and multiple groups of training data constitute a training set.

[0084] Assume that a data set contains 12 hours of real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data. Each hour is regarded as a time step. The 12 hours of real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data are marked as a feature data set A, where A = {A1, A2, A3, A4, A5..., A 12}, A iIt includes the real-time power consumption data, user environment data, historical meteorological data and power consumption preference data of the ith hour. The length of the sliding window is defined as 6, so 6 hours of data is a sample. The total power consumption of users in the next hour is predicted, that is, k = 1, then A1, A2, A3, A4, A5 and A6 are a group of samples as the input of the first machine learning model, and the total power consumption of users in the 7th hour is used as the sample label.

[0085] Using the training set generated by the sliding window method as the input of the first machine learning model, the first machine learning model takes the future total power consumption prediction value as the output, takes the real-time sample as the prediction target, and takes minimizing the loss function value of the first machine learning model as the training target; when the loss function value of the first machine learning model is less than or equal to the preset first target loss value, the training is stopped;

[0086] The first machine learning model loss function is a mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the sample group number; u is the number of sample groups; y i is the label corresponding to the i-th group of samples, The total electric power predicted for the i-th group of samples; preferably, the first machine learning model is a recurrent neural network model, and the recurrent neural network model can be a long short-term memory network (LSTM).

[0087] Other model parameters of the first machine learning model, such as the depth of the network model, the number of neurons in each layer, the activation function used by the network model, the convergence condition, the ratio of the training set, the test set, and the validation set, and the loss function, are all obtained through actual engineering implementation and continuous experimental tuning.

[0088] A second machine learning model is trained based on the training set to predict the total power consumption of users in the medium and long term in real time;

[0089] The training process of the second machine learning model includes:

[0090] The training set is converted into a feature matrix and projected into a high-dimensional embedding space with a dimension of d; the high-dimensional embedding space refers to the process of mapping the original data points into a high-dimensional space. In this process, each data point is represented as a vector whose dimension is much larger than the dimension of the original data point. This space is used for data analysis and machine learning tasks such as dimensionality reduction, clustering, classification, and visualization.

[0091] Add position encoding to each time step; position encoding provides the model with temporal information in the time series. Specifically, sine and cosine functions are used to encode position information.

[0092] Compute query Q, key K and value V;

[0093] Calculate the attention score: d k is the dimension of K;

[0094] After concatenating the attention scores, they are mapped to the original dimension through linear transformation, and the total power consumption of users in the medium and long term is output.

[0095] The second machine learning model has a step size of 12 hours and a time window of 7. The second machine learning model provides a global perspective, focusing on trends and periodic changes in long time steps. The first machine learning model gradually memorizes and processes the sequential information in the time series, revealing the gradual change process of short-term user load power over time. The combination of the two can fully handle complex scenarios in power load forecasting tasks that require both long-term regularity analysis and short-term dynamic response. The second machine learning model is a transform model.

[0096] Calculate a load evaluation value based on the first machine learning model and the second machine learning model;

[0097] Load rating Where D1(t) represents the short-term total power consumption of users predicted by the first machine learning model at time t, P total (t) is the maximum power supply capacity of the region at time t; D2(t) is the total power consumption of users in the medium and long term predicted by the second machine learning model at time t; is the average maximum power supply capacity in the future time window T; w1 is 0.8; w2 is 0.2;

[0098] The load evaluation value combines short-term and medium-term load assessments, while taking into account the impact of dynamic load changes, and is suitable for load evaluation analysis in multiple scenarios. The higher the load evaluation value, the closer the power load is to the total power supply capacity.

[0099] Compare the load evaluation value with the preset load evaluation threshold, and when the load evaluation value is greater than or equal to the load evaluation threshold, enable the backup power supply;

[0100] When the load evaluation value is less than 90% of the load evaluation threshold but greater than or equal to 80% of the load evaluation threshold, a load regulation strategy is adopted to recommend users through the APP to delay operating electricity to the low-consumption period and provide a low-price electricity rate strategy.

[0101] This application uses smart meters to collect the total power consumption of users and the power consumption of each sub-device in real time, and simultaneously obtains user environmental data (temperature, humidity, and light intensity) and historical meteorological data (temperature, humidity, and rainfall probability), and combines user power preferences and regional maximum power supply capabilities to construct a timestamp-tagged data set. The first machine learning model (LSTM) is trained based on the sliding window method to achieve short-term load forecasting, and the second machine learning model (based on the attention mechanism) is trained to achieve medium- and long-term load forecasting. Combining the outputs of the two models, the load evaluation value is calculated, and the load status is evaluated in real time through the evaluation value, and then the backup power supply activation and load control strategy are formulated. This application realizes the short-term dynamic response and medium- and long-term trend analysis of power loads, providing intelligent support for user power management and power grid load control.

[0102] This application comprehensively considers real-time electricity consumption, environmental data, meteorological data and user electricity preferences, significantly improving the model's ability to learn short-term dynamics and long-term time series rules, combining short-term with medium- and long-term load forecasting models to achieve accurate and comprehensive load forecasting, and adapt to complex and changing load scenarios. The load evaluation value is used to evaluate the power load status in real time, accurately control the activation of backup power supplies, and improve the grid dispatch efficiency and user experience based on the peak and valley optimization strategy. The introduction of electricity preference data makes the prediction model more personalized and refined, meets the needs of different users, and optimizes the electricity consumption structure. The load control strategy is used to avoid overload risks, improve the reliability of grid operation, and reduce the operating costs of the power supply system.

[0103] Example 2

[0104] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a smart energy-saving power distribution method and system based on artificial intelligence are provided, and the system includes:

[0105] Data collection module, which collects real-time power consumption data, user environment data, historical meteorological data, power preference data and regional maximum power supply capacity. Real-time power consumption data includes total power consumption and sub-device power consumption per minute; user environment data includes temperature, humidity and light intensity;

[0106] The data preprocessing module timestamps the real-time power consumption data, user environment data, historical meteorological data, and power consumption preference data as a data set; the timestamp includes the characteristics of time, quarter, and whether it is a weekday;

[0107] A short-term model training module, which trains the first machine learning model for real-time prediction of the total power consumption of short-term users based on the data set;

[0108] The medium- and long-term model training module trains a second machine learning model based on the data set to predict the total power consumption of users in the medium and long term in real time;

[0109] A post-processing module, calculating a load evaluation value based on the first machine learning model and the second machine learning model;

[0110] The decision module compares the load evaluation value with the preset load evaluation threshold. When the load evaluation value is greater than or equal to the load evaluation threshold, the backup power supply is enabled; when the load evaluation value is less than 90% of the load evaluation threshold but greater than or equal to 80% of the load evaluation threshold, a load regulation strategy is adopted.

[0111] The modules are connected via wired and / or wireless networks.

[0112] Example 3

[0113] See also Figure 3 As shown, according to another aspect of the present application, an intelligent energy-saving power distribution method and system electronic device 500 based on artificial intelligence is also provided. The electronic device 500 may include one or more processors and one or more memories. Among them, the memory stores a computer-readable code, and when the computer-readable code is executed by one or more processors, the intelligent energy-saving power distribution method and system based on artificial intelligence as described above can be executed.

[0114] The method or system according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as ROM 503 or hard disk 507, may store the intelligent energy-saving power distribution method and system based on artificial intelligence provided by the present application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0115] Example 4

[0116] See also Figure 4As shown, a computer-readable storage medium 600 according to an embodiment of the present application is shown. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are executed by the processor, the intelligent energy-saving power distribution method and system based on artificial intelligence according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0117] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0118] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0119] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0121] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only for the intelligent energy-saving power distribution method and system based on artificial intelligence. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0122] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0123] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0124] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0125] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent energy-saving power distribution method based on artificial intelligence, characterized in that: include: Collect real-time power consumption data, user environment data, historical meteorological data, power preference data and regional maximum power supply capacity. Real-time power consumption data includes total power consumption and sub-device power consumption per minute; user environment data includes temperature, humidity and light intensity; Real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data are timestamped and used as data sets; the timestamp includes the characteristics of time, quarter, and whether it is a weekday; The first machine learning model trained based on the dataset to predict short-term total user power consumption in real time; A second machine learning model is trained based on the data set to predict the total power consumption of users in the medium and long term in real time; Calculate a load evaluation value based on the first machine learning model and the second machine learning model; The load evaluation value is compared with the preset load evaluation threshold. When the load evaluation value is greater than or equal to the load evaluation threshold, the backup power supply is enabled; when the load evaluation value is less than 90% of the load evaluation threshold but greater than or equal to 80% of the load evaluation threshold, a load regulation strategy is adopted.

2. The intelligent energy-saving power distribution method based on artificial intelligence according to claim 1 is characterized in that: Sub-devices are different types of electrical equipment used by users, including automobile electricity, air conditioner and refrigerator electricity, heating electricity and small household appliances electricity; Historical meteorological data includes the temperature, humidity and rainfall probability for the next 24 hours. The temperature and humidity in the historical meteorological data represent the temperature and humidity of the user's geographical environment. The power usage preference data is data reflecting the user's habitual power usage time. The power usage preference data is a set of vector data, which maps different types of sub-devices to numbers as the x component of the vector; the y component and the z component are the preferred start and end time of the sub-device; The maximum power supply capacity of a region is the maximum amount of power that the power system can provide in a unit time period during the power distribution process; The timestamp marking process includes: formatting the data into a unified structure; appending unified time features; and adding features of quarter and whether it is a working day.

3. The intelligent energy-saving power distribution method based on artificial intelligence according to claim 2 is characterized in that: The training process of the first machine learning model includes: Use sliding window method to convert the data set into training set; The data set is divided into n time windows of equal size. Multiple sets of real-time power consumption data, user environment data, historical meteorological data and power consumption preference data in each window are taken as a sample. The total power consumption of users at the future k moments in the window is taken as a label. One sample corresponds to one label. One sample and the corresponding label constitute a set of training data. Multiple sets of training data constitute a training set. A data set contains 12 hours of real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data. Each hour is regarded as a time step. The 12 hours of real-time electricity consumption data, user environment data, historical meteorological data, and electricity preference data are marked as a feature data set A, where A = {A1, A2, A3, A4, A5..., A 12 }, A i It includes the real-time power consumption data, user environment data, historical meteorological data and power consumption preference data of the ith hour. The length of the sliding window is defined as 6, so 6 hours of data is a sample. The total power consumption of users in the next hour is predicted, that is, k = 1, then A1, A2, A3, A4, A5 and A6 are a group of samples as the input of the first machine learning model, and the total power consumption of users in the 7th hour is used as the sample label. Using the training set generated by the sliding window method as the input of the first machine learning model, the first machine learning model takes the future total power consumption prediction value as the output, takes the real-time sample as the prediction target, and takes minimizing the loss function value of the first machine learning model as the training target; when the loss function value of the first machine learning model is less than or equal to the preset first target loss value, the training is stopped; The first machine learning model loss function is a mean square error; the mean square error is calculated by Minimize to train the model; in the loss function, MSE is the loss function value, i is the sample group number; u is the number of sample groups; y i is the label corresponding to the i-th group of samples, The total power consumption predicted for the i-th group of samples; the first machine learning model is a recurrent neural network model.

4. The intelligent energy-saving power distribution method based on artificial intelligence according to claim 3 is characterized in that: The training process of the second machine learning model includes: The training set is converted into a feature matrix and projected into a high-dimensional embedding space with a dimension of d; Add position encoding to each time step; use sine and cosine functions to encode position information; Compute query Q, key K and value V based on time step; Calculate the attention score: d k is the dimension of K; The attention scores are concatenated and mapped to the original dimension through linear transformation to output the total power consumption of users in the medium and long term. The step size of the second machine learning model is 12 hours and the time window is 7. The second machine learning model is a transformer model.

5. The intelligent energy-saving power distribution method based on artificial intelligence according to claim 4 is characterized in that: Load rating Where D1(t) represents the short-term total power consumption of users predicted by the first machine learning model at time t, P total (t) is the maximum power supply capacity of the region at time t; D2(t) is the total power consumption of users in the medium and long term predicted by the second machine learning model at time t; is the average maximum power supply capacity in the future time window T; w1 is 0.8; w2 is 0.

2.

6. The intelligent energy-saving power distribution method based on artificial intelligence according to claim 5 is characterized in that: The load regulation strategy includes: recommending users to delay power operation to the low-consumption period through the APP, and providing low-price electricity tariff strategies.

7. The intelligent energy-saving power distribution system based on artificial intelligence according to claim 1 is characterized in that: Data collection module, which collects real-time power consumption data, user environment data, historical meteorological data, power preference data and regional maximum power supply capacity. Real-time power consumption data includes total power consumption and sub-device power consumption per minute; user environment data includes temperature, humidity and light intensity; The data preprocessing module timestamps the real-time power consumption data, user environment data, historical meteorological data, and power consumption preference data as a data set; the timestamp includes the characteristics of time, quarter, and whether it is a weekday; A short-term model training module, which trains the first machine learning model for real-time prediction of the total power consumption of short-term users based on the data set; The medium- and long-term model training module trains a second machine learning model based on the data set to predict the total power consumption of users in the medium and long term in real time; A post-processing module, calculating a load evaluation value based on the first machine learning model and the second machine learning model; The decision module compares the load evaluation value with the preset load evaluation threshold. When the load evaluation value is greater than or equal to the load evaluation threshold, the backup power supply is enabled; when the load evaluation value is less than 90% of the load evaluation threshold but greater than or equal to 80% of the load evaluation threshold, a load regulation strategy is adopted.

8. An electronic device, characterized in that include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the intelligent energy-saving power distribution method based on artificial intelligence as described in any one of claims 1 to 6 by calling the computer program stored in the memory.

9. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are executed on a computer, the computer executes the intelligent energy-saving power distribution method based on artificial intelligence as described in any one of claims 1 to 6.

Images