Training method and device of electricity consumption information prediction model and electricity consumption information prediction method

Abstract

The invention provides a training method and device of an electricity consumption information prediction model and an electricity consumption information prediction method, and relates to the technical field of Internet such as artificial intelligence and deep learning. Comprising the following steps: inputting a sample tetrad consisting of a head entity object representing a power utilization area, a relation object representing a power utilization unit value attribute, a tail entity object representing a power utilization total value attribute and a time object representing timestamp information into a first neural network to obtain a sample embedding matrix representation of the sample tetrad; inputting the sample embedded matrix representation into a second neural network to obtain respective sensitive characteristic change rates of the power utilization area, the power utilization unit value attribute and the power utilization total value attribute; inputting the sensitivity characteristic change rate into a third neural network, and calculating a matching degree evaluation value prediction result of the prediction information relative to other real information; and according to the sample embedding matrix representation, the matching degree evaluation value real label and the matching degree evaluation value prediction result, calculating the training loss, and adjusting the model parameters.

Description

Technical Field

[0001] The present invention relates to the field of Internet technologies, and in particular, to fields such as artificial intelligence and deep learning. More specifically, the present invention relates to a training method and device for an electricity consumption information prediction model and an electricity consumption information prediction method. Background Art

[0002] With the opening of the electricity market and the increasing proportion of renewable energy, the electricity price fluctuation has intensified. For example, the "hybrid prediction model" that combines the decomposition algorithm and deep learning and the method of "considering external variables such as weather and policies" have become research hotspots. In addition, "probability prediction" and "uncertainty quantification" technologies have also gradually received attention to cope with the risks in the electricity market.

[0003] However, how to balance prediction accuracy and computational efficiency remains a key challenge in the field of electricity consumption information prediction. Summary of the Invention

[0004] In view of this, the present invention provides a training method and device for an electricity consumption information prediction model and an electricity consumption information prediction method.

[0005] One aspect of the present invention provides a training method for an electricity consumption information prediction model, including: inputting a sample quadruple composed of a head entity object representing an electricity consumption area, a relationship object representing the value attribute of an electricity consumption unit, a tail entity object representing the total value attribute of electricity consumption, and a time object representing timestamp information into a first neural network of the electricity consumption information prediction model to obtain a sample embedding matrix representation of the sample quadruple; inputting the sample embedding matrix representation into a second neural network of the electricity consumption information prediction model to obtain the sensitivity feature change rates of the electricity consumption area, the value attribute of the electricity consumption unit, and the total value attribute of electricity consumption represented by the sample quadruple with respect to the target moment represented by the time object; inputting the sensitivity feature change rates into a third neural network of the electricity consumption information prediction model to calculate a matching degree evaluation value prediction result of the prediction information determined for the target entity object in the sample quadruple with respect to the real information represented by other objects in the sample quadruple, where the target entity object represents the head entity object or the tail entity object, and other objects represent the objects in the sample quadruple except the target entity object, and the sample quadruple has a real label of the matching degree evaluation value; calculating a training loss according to the sample embedding matrix representation, the real label of the matching degree evaluation value, and the matching degree evaluation value prediction result; and adjusting the model parameters of the first neural network, the second neural network, and the third neural network respectively by using the training loss to obtain a trained electricity consumption information prediction model.

[0006] Another aspect of the present invention provides a method for predicting electricity consumption information, including: obtaining an incomplete quadruple composed of a head entity object representing an electricity consumption area, a relationship object representing the value attribute of an electricity consumption unit, a tail entity object representing the total value attribute of electricity consumption, and a time object representing timestamp information, wherein the information represented by the incomplete object in the incomplete quadruple is empty, and the incomplete object represents the head entity object or the tail entity object; inputting the incomplete quadruple into an electricity consumption information prediction model to obtain a target completed result determined for the incomplete object, and outputting a completed quadruple including the target completed result, wherein the electricity consumption information prediction model is trained by using the training method of the electricity consumption information prediction model of the present disclosure.

[0007] Another aspect of the present invention provides a training device for an electricity consumption information prediction model, including: a first neural network module, configured to input a sample quadruple composed of a head entity object representing an electricity consumption area, a relationship object representing the value attribute of an electricity consumption unit, a tail entity object representing the total value attribute of electricity consumption, and a time object representing timestamp information into the first neural network of the electricity consumption information prediction model to obtain a sample embedding matrix representation of the sample quadruple; a second neural network module, configured to input the sample embedding matrix representation into the second neural network of the electricity consumption information prediction model to obtain the sensitivity feature change rates of the electricity consumption area, the value attribute of the electricity consumption unit, and the total value attribute of electricity consumption represented by the sample quadruple with respect to the target time represented by the time object; a third neural network module, configured to input the sensitivity feature change rates into the third neural network of the electricity consumption information prediction model to calculate a matching degree evaluation value prediction result of the predicted information determined for the target entity object in the sample quadruple with respect to the true information represented by other objects in the sample quadruple, the target entity object represents the head entity object or the tail entity object, and other objects represent the objects in the sample quadruple except the target entity object, and the sample quadruple has a true label of the matching degree evaluation value; a loss module, configured to calculate a training loss according to the sample embedding matrix representation, the true label of the matching degree evaluation value, and the matching degree evaluation value prediction result; a training module, configured to adjust the model parameters of the first neural network, the second neural network, and the third neural network respectively by using the training loss to obtain a trained electricity consumption information prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Through the following description of the embodiments of the present invention with reference to the drawings, the above and other objects, features, and advantages of the present invention will become clearer. In the drawings:

[0009] Figure 1 An exemplary system architecture in which at least one of the training method and the prediction method of the electricity consumption information prediction model according to the embodiments of the present invention can be applied is shown;

[0010] Figure 2Shows a flowchart of a method for training an electricity consumption information prediction model according to an embodiment of the present invention;

[0011] Figure 3 Shows a flowchart of a method for predicting electricity consumption information according to an embodiment of the present invention;

[0012] Figure 4 Shows an architecture diagram of a model for predicting electricity consumption information based on a sensitivity-driven method using differential equations according to an embodiment of the present invention;

[0013] Figure 5A Shows a functional module diagram of an electricity consumption information prediction system constructed based on a sensitivity-driven method using differential equations according to an embodiment of the present invention;

[0014] Figure 5B Shows an overall flowchart of predicting electricity consumption information based on a sensitivity-driven method using differential equations according to an embodiment of the present invention;

[0015] Figure 6 Shows a block diagram of a training device for an electricity consumption information prediction model according to an embodiment of the present invention;

[0016] Figure 7 Shows a block diagram of an electricity consumption information prediction device according to an embodiment of the present invention;

[0017] Figure 8 Shows a block diagram of an electronic device suitable for implementing at least one of the method for training an electricity consumption information prediction model and the method for predicting electricity consumption information according to an embodiment of the present invention. Detailed implementation manners

[0018] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present invention. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0019] For an electricity consumption information prediction system, it can be understood as a task of predicting unknown facts based on existing facts to assist cost calculation. Therefore, as a structured representation of real-world quadruples, the electricity consumption information prediction system can provide assistance in many fields such as search, recommendation systems, natural language processing, and question answering.

[0020] Generally, a quadruple is in the form of (head entity, relation, tail entity, timestamp). Currently, existing quadruple databases already contain very complex information. However, there is still a serious lack of information in most databases, and these missing information severely restricts data accuracy, and it is urgent to improve the information. In related technologies, most rely on manual search and completion methods. However, severely restricted by cost and accuracy, it is very necessary to apply algorithms to intelligent search and completion.

[0021] Figure 1 FIG. 100 shows an exemplary system architecture to which at least one of a training method and a prediction method of an electricity consumption information prediction model according to an embodiment of the present invention can be applied. It should be noted that, Figure 1 The figure shown is only an example of a system architecture to which the embodiments of the present invention can be applied, to help those skilled in the art understand the technical content of the present invention, but it does not mean that the embodiments of the present invention cannot be used in other devices, systems, environments or scenarios.

[0022] As Figure 1 shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0023] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only examples).

[0024] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, and desktop computers, etc.

[0025] The server 105 may be a server that provides various services. For example, it may be a background management server (only for example) that supports the websites browsed by the user using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process data such as user requests received, and feedback the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0026] It should be noted that at least one of the training method and the electricity consumption information prediction method of the application electricity consumption information prediction model provided by the embodiments of the present invention can generally be executed by the server 105. Correspondingly, at least one of the training device and the electricity consumption information prediction device of the application electricity consumption information prediction model provided by the embodiments of the present invention can generally be set in the server 105. At least one of the training method and the electricity consumption information prediction method of the application electricity consumption information prediction model provided by the embodiments of the present invention can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Correspondingly, at least one of the training device and the electricity consumption information prediction device of the application electricity consumption information prediction model provided by the embodiments of the present invention can also be set in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Or, at least one of the training method and the electricity consumption information prediction method of the application electricity consumption information prediction model provided by the embodiments of the present invention can also be executed by the first terminal device 101, the second terminal device 102, or the third terminal device 103, or can also be executed by other terminal devices different from the first terminal device 101, the second terminal device 102, or the third terminal device 103. Correspondingly, at least one of the training device and the electricity consumption information prediction device of the application electricity consumption information prediction model provided by the embodiments of the present invention can also be set in the first terminal device 101, the second terminal device 102, or the third terminal device 103, or set in other terminal devices different from the first terminal device 101, the second terminal device 102, or the third terminal device 103.

[0027] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the servers in

[0028] Figure 2 shows a flowchart of a training method for an electricity consumption information prediction model according to an embodiment of the present invention.

[0029] As shown in Figure 2As shown, the method includes operations S201 to S205.

[0030] In operation S201, a sample quadruple composed of a head entity object representing the electricity-consuming area, a relationship object representing the value attribute of the electricity-consuming unit, a tail entity object representing the total electricity value attribute, and a time object representing the timestamp information is input into the first neural network of the electricity consumption information prediction model to obtain a sample embedding matrix representation of the sample quadruple.

[0031] According to an embodiment of the present invention, in the electricity consumption information prediction scenario, the sample quadruple can be expressed as (electricity-consuming area, value attribute of the electricity-consuming unit, total electricity value attribute, timestamp). The first neural network can use various encoding networks to encode the sample quadruple to obtain a sample embedding matrix representation. The specific encoding method is not limited herein. For example, the sample quadruple as a whole can be encoded to obtain a sample embedding matrix representation. Alternatively, the four elements in the sample quadruple can be encoded separately to obtain four encoded vectors, and a sample embedding matrix representation can be obtained by performing matrix transformation on the four encoded vectors.

[0032] In operation S202, the sample embedding matrix representation is input into the second neural network of the electricity consumption information prediction model to obtain the sensitivity feature change rates of the electricity-consuming area, the value attribute of the electricity-consuming unit, and the total electricity value attribute represented by the sample quadruple with respect to the target time represented by the time object.

[0033] According to an embodiment of the present invention, the second neural network can use various feature extraction networks to extract features from the sample embedding matrix representation to obtain the sensitivity feature change rates.

[0034] According to an embodiment of the present invention, the electricity-consuming area, the value attribute of the electricity-consuming unit, and the total electricity value attribute can each have certain sensitivity features with respect to any moment. The sensitivity features can reflect the instantaneous response degree of the electricity-consuming area, the value attribute of the electricity-consuming unit, and the total electricity value attribute with respect to time changes, that is, "how sensitive". The sensitivity feature change rate can characterize the change speed of the sensitivity features of the electricity-consuming area, the value attribute of the electricity-consuming unit, and the total electricity value attribute with respect to time changes, that is, "the change speed of sensitivity".

[0035] In operation S203, the sensitivity feature change rate is input into the third neural network of the electricity consumption information prediction model to calculate the prediction result of the matching degree evaluation value of the prediction information determined for the target entity object in the sample quadruple with respect to the real information represented by the other objects in the sample quadruple. The target entity object represents the head entity object or the tail entity object, and the other objects represent the objects in the sample quadruple except the target entity object. The sample quadruple has a real label of the matching degree evaluation value.

[0036] According to an embodiment of the present invention, the third neural network can use various deep learning networks with learning capabilities. The sample quadruple input into the electricity consumption information prediction model is (electricity consumption area, electricity consumption unit value attribute, total electricity consumption value attribute, timestamp), where the electricity consumption area, electricity consumption unit value attribute, total electricity consumption value attribute, and timestamp all use real information. According to the real information of the electricity consumption area, the real information of the electricity consumption unit value attribute, and the real information of the total electricity consumption value attribute, the true label of the matching degree evaluation value can be calculated.

[0037] According to an embodiment of the present invention, in the processing stage of the third neural network, based on the real information of the target entity object learned from the training set, the prediction information for each prediction of the target entity object can be determined. The training set can represent a set of multiple sample quadruples. Then, for each prediction information, the matching degree between it and the above real information can be calculated to obtain the prediction result of the matching degree evaluation value.

[0038] For example, based on the learned real information of the total electricity consumption value attribute, the predicted information of the total electricity consumption value attribute for the tail entity object in the sample quadruple can be determined. Then, the matching degree between the predicted information of the total electricity consumption value attribute and the real information of the electricity consumption area, the real information of the electricity consumption unit value attribute in this sample quadruple can be calculated to obtain the prediction result of the matching degree evaluation value.

[0039] For example, based on the learned real information of the electricity consumption area, the predicted information of the electricity consumption area for the head entity object in the sample quadruple can be determined. Then, the matching degree between the electricity consumption area and the real information of the electricity consumption unit value attribute, the real information of the total electricity consumption value attribute, and / or the real information of the timestamp in this sample quadruple can be calculated to obtain the prediction result of the matching degree evaluation value.

[0040] It should be noted that in some embodiments, during the process of calculating the prediction result of the matching degree evaluation value, the matching degree between the prediction information and the real information of the timestamp can also be calculated. Furthermore, by combining the matching degree between the prediction information and other real information, the prediction result of the matching degree evaluation value can be obtained, which is not limited herein.

[0041] In operation S204, according to the sample embedding matrix representation, the true label of the matching degree evaluation value, and the prediction result of the matching degree evaluation value, the training loss is calculated.

[0042] In operation S205, the model parameters of the first neural network, the second neural network, and the third neural network are adjusted using the training loss to obtain the trained electricity consumption information prediction model.

[0043] According to an embodiment of the present invention, a regularization loss can be constructed based on the actual business requirements, in combination with the sample embedding matrix representation, the true label of the matching degree evaluation value, and the predicted result of the matching degree evaluation value, and the regularization loss value can be calculated. Thus, the model parameters of each network in the electricity consumption information prediction model can be adjusted using the regularization loss value until the loss converges, and then the trained electricity consumption information prediction model can be obtained.

[0044] Through the above embodiments of the present invention, by combining the sensitivity feature change rate to implement data prediction and model training, the differential idea can be well combined, and the relevant features of each object can be better learned from the perspective of the change rate of the sensitivity feature. The trained electricity consumption information prediction model obtained through this learning process can reasonably complete or predict facts with higher efficiency and in a transparent and interpretable manner, and can more accurately solve the problem of electricity consumption information prediction.

[0045] The following will further illustrate the Figure 2 method shown with specific embodiments.

[0046] According to an embodiment of the present invention, the one-hot encoding can be adopted for the encoding method based on the first neural network. The above operation S201 may include: performing one-hot encoding on the electricity consumption area represented by the head entity object, the electricity consumption unit value attribute represented by the relationship object, the total electricity consumption value attribute represented by the tail entity object, and the timestamp information represented by the time object respectively, to obtain the area binary vector, the unit value attribute binary vector, the total value attribute binary vector, and the time binary vector. Perform matrix conversion on the area binary vector, the unit value attribute binary vector, the total value attribute binary vector, and the time binary vector to obtain the sample embedding matrix representation.

[0047] For example, for the dataset of sample quadruples, one-hot encoding can be used to encode the sample quadruples. Specifically, each entity object in the dataset is represented as dimensional one-hot binary vector, including: making the i-th element of the head entity object and the i-th element of the tail entity object equal to 1, and setting other elements to 0. For each relationship object is represented as an l-dimensional one-hot binary vector, including: making the j-th element of the relationship object equal to 1, and setting the remaining elements to 0. Similarly, the event object can be encoded accordingly to obtain the corresponding one-hot binary vector. Then the encoding process ends. Perform matrix conversion on the one-hot binary vectors of the four objects to obtain the sample embedding matrix representation. For example, it can be represented as . Among them, It can represent a one-hot binary vector related to the electricity consumption area in the sample embedding matrix representation. It can represent a one-hot binary vector related to the value attribute of the electricity consumption unit in the sample embedding matrix representation. It can represent a one-hot binary vector related to the total electricity consumption value attribute in the sample embedding matrix representation. It can represent a one-hot binary vector related to the timestamp information in the sample embedding matrix representation.

[0048] According to an embodiment of the present invention, the above operation S202 may include: performing feature extraction on the sample embedding matrix representation to obtain the sensitivity features exhibited by the electricity consumption area, the value attribute of the electricity consumption unit, and the total electricity consumption value attribute at the target moment. According to the sensitivity features and the target embedding matrix representations of at least one target quadruple representing the target moment obtained from the quadruple dataset, extracting the state change features exhibited by the sample quadruple at the target moment. Performing noise addition processing on the state change features to obtain the sensitivity feature change rate.

[0049] According to an embodiment of the present invention, the quadruple dataset may include the above training set, or may include other datasets with corresponding formats, which are not limited herein. The target embedding matrix representation may have the same or similar technical features as the sample embedding matrix representation, which are not limited herein.

[0050] According to an embodiment of the present invention, according to the real information of the electricity consumption area, the value attribute of the electricity consumption unit, and the total electricity consumption value attribute at any moment, it can reflect the state features of the electricity consumption area, the value attribute of the electricity consumption unit, and the total electricity consumption value attribute relative to the corresponding moment, that is, "what is the real information". The state change features can characterize the change speed of the state features of the electricity consumption area, the value attribute of the electricity consumption unit, and the total electricity consumption value attribute over time, that is, "the change speed of the real information".

[0051] For example, for a given training set, it can first be subjected to sensitivity partitioning and feature interaction through sensitivity-driven stochastic delay differential equations (SDDE) to facilitate finding more accurate answers using the sensitivities of entities and relationships during the final prediction. The specific formula is as shown in formula (1).

[0052] (1)

[0053] Wherein, represents the sensitivity features of the electricity consumption area, the value attribute of the electricity consumption unit, and the total electricity consumption value attribute in the sample quadruple at the moment, It represents the matrix representation of the one-hot binary vectors of the power consumption area, the value attribute of the power consumption unit, and the total value attribute of the power consumption, which have time-sensitive dependence relationships in the sample embedding matrix representation. It represents the sample embedding matrix representation The power consumption area, the value attribute of the power consumption unit, and the total value attribute in the sample quadruple characterized by The state characteristics at the

[0054] By performing the derivative operation in the differential equation on the sample embedding matrix representation, the state change characteristics and the sensitivity characteristic change rate at a specific moment can be obtained using SDDE. The form of SDDE in a continuous time period can be shown in Equation (2) for example.

[0055] (2)

[0056] Among them, represents the starting time point, represents the output time point, represents the sensitivity characteristic at the represents the sensitivity characteristic at the is the drift term, representing the characteristic change between the time intervals and ; is the diffusion term, representing the random change, is the random noise function, represents the infinitesimal increment of the Brownian motion.

[0057] After taking the derivative of Equation (2), the SDDE representation can be obtained as Equation (3).

[0058] (3)

[0059] Among them, represents the target embedding matrix representation of the target quadruple in the quadruple dataset where the time part is at the represents the sensitivity characteristic change rate of the power consumption area, the value attribute of the power consumption unit, and the total value attribute in the continuous time change, represents the state change characteristic of the sample quadruple at the ; represents the noise part.

[0060] According to the business requirements, the noise part in Equation (3) can be selectively used or deleted.

[0061] According to an embodiment of the present invention, the above-mentioned adding noise to the state change feature to obtain the sensitivity feature change rate may include: calculating a random noise vector according to the sensitivity feature, the target time, and the target embedding matrix representation. Using the random noise vector, adding noise to the state change feature to obtain the sensitivity feature change rate.

[0062] In this embodiment, the random noise vector may, for example, use the noise part in formula (3) , and is not limited thereto.

[0063] According to an embodiment of the present invention, other objects may include relationship objects and other entity objects in the sample quadruple except the target entity object. The sensitivity feature change rate may include a relationship sensitivity feature change rate related to the relationship object and an entity sensitivity feature change rate related to other entity objects. The above operation S203 may include: extracting the state feature exhibited by the sample quadruple at the target time according to the sensitivity feature and the sample embedding matrix representation. Calculating a relationship sensitivity matching degree evaluation value between the prediction information and the real information represented by the relationship object according to the prediction information and the relationship sensitivity feature change rate. Calculating an entity sensitivity matching degree evaluation value between the prediction information and the real information represented by other entity objects according to the prediction information and the entity sensitivity feature change rate. Calculating a prediction result of the matching degree evaluation value of the prediction information relative to the real information according to the relationship sensitivity matching degree evaluation value, the entity sensitivity matching degree evaluation value, and the state feature.

[0064] Based on the above formula (1), the state feature can be obtained . On this basis, for example, a scoring function can be constructed to calculate the prediction result of the matching degree evaluation value. The scoring function may depend on the joint embedding and sensitivity features of the power consumption area, the value attribute of the power consumption unit, the total power consumption value attribute, and time, and is dynamically updated with the change of time.

[0065] Based on the target entity object being the tail entity object representing the total power consumption value attribute, other objects may include the head entity object representing the power consumption area and the relationship object representing the value attribute of the power consumption unit. For this embodiment, the scoring function may specifically be as shown in formula (4).

[0066] (4)

[0067] Among them, is a scoring function for calculating the compatibility between the power consumption area, the value attribute of the power consumption unit, and the total power consumption value attribute, is the training round, represents the round. represents the head entity sensitivity feature change rate, Indicates the change rate of the relationship sensitivity feature. and are respectively the processing and application of entity and relationship sensitivity features at a certain moment, Indicates the evaluation value of the entity sensitivity matching degree between the predicted information of the total electricity consumption value attribute and the real information of the electricity consumption area, Indicates the evaluation value of the relationship sensitivity matching degree between the predicted information of the total electricity consumption value attribute and the unit electricity consumption value attribute. 、 are weight coefficients.

[0068] According to business requirements, the 、 and in formula (4) can be selectively used or deleted.

[0069] According to the embodiments of the present invention, there can be multiple prediction results for predicting the target entity object. On this basis, based on the above formula (4), there can be multiple prediction results of the matching degree evaluation value calculated. The above operation S204 may include: performing normalization processing on multiple prediction results of the matching degree evaluation value to obtain the normalized result of the matching degree evaluation value. According to the true label of the matching degree evaluation value and the normalized result of the matching degree evaluation value, calculate the conditional probability of the true label of the matching degree evaluation value under the conditions characterized by multiple normalized results of the matching degree evaluation value. Calculate the training loss according to the conditional probability, the relationship sensitivity matching degree evaluation value, and the entity sensitivity matching degree evaluation value.

[0070] Corresponding to the above embodiment where the target entity object is the tail entity object representing the total electricity consumption value attribute, other objects include the head entity object representing the electricity consumption area and the relationship object representing the unit electricity consumption value attribute. The conditional probability of the total electricity consumption value attribute can be determined by scoring through a scoring function as shown in formula (4). The conditional probability function when predicting the total electricity consumption value attribute can be as shown in formula (5).

[0071] (5)

[0072] Among them, represents the predicted information of the total electricity consumption value attribute, is the set of all predicted information of the total electricity consumption value attribute, and the sum in the denominator is the normalization of all predicted information of the total electricity consumption value attribute. represents the conditional probability of the true information of the total electricity consumption value attribute under the conditions characterized by the normalized result of the matching degree evaluation value of all predicted information of the total electricity consumption value attribute.

[0073] The loss function defined for the electricity consumption information prediction model can be as shown in formula (6).

[0074] (6)

[0075] wherein, represents the training loss. is a quadruple dataset, 、 are regularization parameters.

[0076] It should be noted that, based on the target entity object being the head entity object representing the power consumption area, other objects may include the tail entity object representing the total power consumption value attribute and the relationship object representing the unit power consumption value attribute. For this embodiment, the above formulas (4) to (6) can be correspondingly transformed to facilitate predicting the power consumption area and calculating the matching degree evaluation value between the predicted information of the power consumption area and the true information of the unit power consumption value attribute and the true information of the total power consumption value attribute.

[0077] The model generates the final training loss by using the matching evaluation value between the sensitivity features and the prediction results of the target entity object in the sample quadruple, and this training loss is used as the training objective for optimizing the model.

[0078] According to the embodiment of the present invention, the relationship sensitivity matching degree evaluation value has a relationship sensitivity weight coefficient, and the entity sensitivity matching degree evaluation value has an entity sensitivity weight coefficient. The above operation S205 may include: adjusting the relationship sensitivity weight coefficient and the entity sensitivity weight coefficient by using the training loss.

[0079] For example, for 、 in formula (4), learning and updating can be performed at each time step during the training process.

[0080] It should be noted that the model parameters updated during the training process may not be limited to 、 . In the actual implementation process, it may also include other model parameters in the power consumption information prediction model, which are not limited herein.

[0081] Through the above embodiments of the present invention, a power consumption information prediction model for mine exploitation based on differential equations can be trained. Based on this power consumption information prediction model, a power consumption information prediction system can be constructed to implement power consumption information prediction.

[0082] Figure 3 shows a flowchart of the power consumption information prediction method according to the embodiment of the present invention.

[0083] As Figure 3 shown, this method includes operations S301 to S302.

[0084] In operation S301, obtain an incomplete quadruple composed of a head entity object representing the electricity consumption area, a relationship object representing the value attribute of the electricity consumption unit, a tail entity object representing the total electricity consumption value attribute, and a time object representing the timestamp information. Among them, the information represented by the incomplete object in the incomplete quadruple is empty, and the incomplete object represents the head entity object or the tail entity object.

[0085] In operation S302, input the incomplete quadruple into the electricity consumption information prediction model to obtain a target completed result determined for the incomplete object, and output a completed quadruple including the target completed result.

[0086] For the electricity consumption information prediction system, its main task mainly focuses on the completion in the case of the missing quadruple (head entity, relationship, tail entity, timestamp). The missing information can be represented by the null value NULL, that is, the prediction and completion of the missing situation of the incomplete quadruples (head entity, relationship, NULL, timestamp) and (NULL, relationship, tail entity, timestamp).

[0087] According to an embodiment of the present invention, the above operation S302 may include: inputting the incomplete quadruple into the electricity consumption information prediction model to obtain multiple candidate completed results determined for the incomplete object. Calculate the candidate matching degree evaluation value prediction results of each candidate completed result relative to the real information represented by other objects in the incomplete quadruple to obtain multiple candidate matching degree evaluation value prediction results, and other objects represent the objects in the incomplete quadruple except the incomplete object. Determine the candidate completed result corresponding to the target matching degree evaluation value prediction result with the largest value among the multiple candidate matching degree evaluation value prediction results as the target completed result.

[0088] It should be noted that the incomplete object has the same or similar technical features as the aforementioned target entity object. The candidate completed result has the same or similar technical features as the aforementioned prediction result. The candidate matching degree evaluation value prediction result has the same or similar technical features as the aforementioned matching degree evaluation value prediction result. Details are not described herein again.

[0089] For each incomplete quadruple in the given test set, the trained electricity consumption information prediction model can be used to calculate the scoring function shown in formula (4), and the candidate completed result with the highest calculated score is automatically determined as the target completed result. By traversing all the incomplete quadruples in the test set, the test set calculation and data completion can be completed.

[0090] Through the above embodiments of the present invention, a method for predicting electricity consumption information based on differential equations is realized, which can perform real-time intelligent prediction and completion of the database, and can reasonably predict and present the future facts through intelligent speculation. Compared with other methods, this method can reasonably complete or predict facts with higher efficiency and interpretability. In addition, by introducing a mathematical model into the system application and presenting it in a black box manner, the system interface is simple and there are no technical requirements for operators.

[0091] Figure 4 Fig. shows a model architecture diagram for predicting electricity consumption information by a sensitivity-driven method based on differential equations according to an embodiment of the present invention.

[0092] As Figure 4 shown, the electricity consumption information prediction model 400 includes an embedding processing module 410, a sensitivity perception module 420, and a result prediction module 430.

[0093] During the model training process, the embedding processing module 410 is configured to receive (power consumption area, power consumption unit value attribute, total power consumption value attribute, timestamp) as input and output an embedding matrix representation of (power consumption area, power consumption unit value attribute, total power consumption value attribute, timestamp). The sensitivity perception module 420 is configured to receive the embedding matrix representation as input, perform vector sensitivity division, and output the sensitivity characteristic change rates of the power consumption area, the power consumption unit value attribute, and the total power consumption value attribute with respect to the timestamp respectively. The result prediction module 430 can receive the sensitivity characteristic change rate as input and output a prediction result of the matching degree evaluation value of the prediction information determined for the power consumption area or the total power consumption value attribute in (power consumption area, power consumption unit value attribute, total power consumption value attribute, timestamp) with respect to other real information. Then, the training loss can be calculated in combination with the loss function to adjust the model parameters in the electricity consumption information prediction model 400 to obtain a trained electricity consumption information prediction model.

[0094] During the model application process, the entity and relationship embedding processing module 410 is configured to receive (power consumption area, power consumption unit value attribute, NULL, timestamp) as input and output an embedding matrix representation of (power consumption area, power consumption unit value attribute, NULL, timestamp). The sensitivity perception module 420 is configured to receive the embedding matrix representation as input and output the sensitivity characteristic change rates of the power consumption area and the power consumption unit value attribute with respect to the timestamp respectively. The result prediction module 430 can receive the sensitivity characteristic change rate as input and output the prediction information of the total power consumption value attribute with the highest matching degree evaluation value as the completion result for the NULL value in (power consumption area, power consumption unit value attribute, NULL, timestamp).

[0095] Figure 5AThe functional module diagram of the electricity consumption information prediction system constructed based on the sensitivity-driven method of differential equations according to an embodiment of the present invention is shown.

[0096] As Figure 5A shown, the electricity consumption information prediction system 500 constructed based on the sensitivity-driven method of differential equations includes a database module 510, a data import and processing module 520, a sensitivity-driven prediction module 530, and a system maintenance module 540.

[0097] The functions of the following modules will be described in conjunction with Figure 5B this Figure 5A to illustrate the functions of each module.

[0098] Figure 5B The overall flowchart of electricity consumption information prediction based on the sensitivity-driven method of differential equations according to an embodiment of the present invention is shown.

[0099] As Figure 5B shown, this method includes operations S501 to S509.

[0100] In operation S501, the data set to be completed is split into a training set, a validation set, and a test set.

[0101] Corresponding to this operation, the data set to be completed can be imported and stored in the database module 510, including: clicking "Data Import" on the page, selecting the data set to be imported in the folder, clicking "Upload" to complete the data import operation. After that, the data set can be divided into a training set T, a validation set V, and a test set S according to a ratio of 8:1:1.

[0102] In operation S502, the data set is processed structurally.

[0103] Corresponding to this operation, in combination with the data import and processing module 520, the selected data set in the database module 510 can be read, and the imported data can be organized into a structured form to facilitate subsequent model processing. The specific steps of structuring are as follows: using the Language Technology Platform (LTP) to perform entity extraction, relationship extraction, entity unification, and anaphora resolution on the data set in sequence. Among them: entity extraction, that is, the recognition of entities, including the detection and classification of entities. Relationship extraction can generally be understood as quadruple extraction, that is, a data set can be represented in the form of a quadruple (head entity, relationship, tail entity, timestamp). Entity unification is to unify the entity names of the same entity referred to by different names. Anaphora resolution is to represent the object referred to by a referring noun as the entity name of the object.

[0104] In operation S503, the structured data set is encoded.

[0105] Corresponding to this operation, one-hot encoding can be used for encoding.

[0106] In operation S504, an electricity consumption information prediction model is established, and the model includes two parts: a scoring function and a loss function.

[0107] Corresponding to this operation, the construction of the electricity consumption information prediction model can be realized by establishing a relational interaction block decomposition model based on the sensitivity-driven prediction module 530.

[0108] In operation S505, model hyperparameters are set based on the sensitivity-driven differential equation.

[0109] Corresponding to this operation, the learning rate of the model initial parameters can be set, including: Learning rate, Batchsize, ent_vec_dim, rel_vec_dim, tem_vec_dim, Epoch, etc., and it is not limited to this. As shown in Table 1.

[0110] Table 1:

[0111]

[0112] In operation S506, the model is trained using the training set until the model loss function converges.

[0113] Corresponding to this operation, the model training can be realized based on the above formulas (1) to (6), which will not be elaborated here.

[0114] In operation S507, the trained model is used to complete the test set.

[0115] Corresponding to this operation, the trained electricity consumption information prediction model can be used to complete the test set S, including: inputting the quadruple to be completed into the model, calculating the matching degree evaluation value prediction result based on the scoring function, and the one with the highest evaluation value is the search result.

[0116] In operation S508, the model is evaluated through evaluation metrics.

[0117] Corresponding to this operation, the effect of the above electricity consumption information prediction model based on the differential equation can be evaluated and verified through experiments, including: calculating relevant coefficients such as the mean reciprocal rank (MRR), the proportion of the correct answer in the first result (Hit@1), the proportion of the correct answer in the top 3 results (Hit@3), and the proportion of the correct answer in the top 10 results (Hit@10) of the experimental evaluation metrics, so as to realize the evaluation of the model. A comparative experiment is used to evaluate and verify the calculation results to realize the evaluation of this model.

[0118] The summary table of test evaluation indicators is shown in Table 2, which shows the comparison results between SDDE and a classical model for knowledge graph embedding (Translation Embedding, abbreviated as TransE), a knowledge graph embedding model based on bilinear transformation (DistMult, abbreviated as DisMult), a knowledge graph embedding model (Rotation Embedding, abbreviated as RotatE), a knowledge graph embedding model based on complex vector space (Complex Embeddings, abbreviated as ComplEx), a knowledge graph embedding model based on convolutional neural network (Convolutional Embedding, abbreviated as ConvE), a graph neural network model dedicated to processing multi-relational graph data (Relational Graph Convolutional Network, abbreviated as R-GCN), a temporal knowledge graph reasoning model based on graph convolutional network (Recurrent Evolutionary Graph Convolutional Network, abbreviated as RE-GCN), a model for image processing and computer vision (Generalized Hough Transform, abbreviated as GHT), and an advanced artificial intelligence model combining retrieval-augmented generation technology and language transformation model (Retrieval-Augmented Generation Language Transformer, abbreviated as rGalT).

[0119] Table 2:

[0120]

[0121] In operation S509, the search prediction result is visually displayed.

[0122] During the entire process of the above operations S501 - S509, real-time maintenance can be combined with the system maintenance module 540.

[0123] Figure 6 The block diagram of the training device of the electricity consumption information prediction model according to an embodiment of the present invention is shown.

[0124] As Figure 6 shown, the training device 600 of the electricity consumption information prediction model includes a first neural network module 610, a second neural network module 620, a third neural network module 630, a loss module 640, and a training module 650.

[0125] The first neural network module 610 is configured to input a sample quadruple composed of a head entity object representing the power consumption area, a relationship object representing the value attribute of the power consumption unit, a tail entity object representing the total power consumption value attribute, and a time object representing the timestamp information into the first neural network of the power consumption information prediction model, and obtain a sample embedding matrix representation of the sample quadruple.

[0126] The second neural network module 620 is configured to input the sample embedding matrix representation into the second neural network of the power consumption information prediction model, and obtain the sensitivity feature change rates of the power consumption area, the value attribute of the power consumption unit, and the total power consumption value attribute represented by the sample quadruple with respect to the target time represented by the time object.

[0127] The third neural network module 630 is configured to input the sensitivity feature change rate into the third neural network of the power consumption information prediction model, and calculate a prediction result of the matching degree evaluation value of the prediction information determined for the target entity object in the sample quadruple with respect to the real information represented by other objects in the sample quadruple. The target entity object represents the head entity object or the tail entity object, and other objects represent the objects in the sample quadruple except the target entity object. The sample quadruple has a real label of the matching degree evaluation value.

[0128] The loss module 640 is configured to calculate a training loss according to the sample embedding matrix representation, the real label of the matching degree evaluation value, and the prediction result of the matching degree evaluation value.

[0129] The training module 650 is configured to adjust the model parameters of the first neural network, the second neural network, and the third neural network respectively by using the training loss, and obtain a trained power consumption information prediction model.

[0130] According to an embodiment of the present invention, the second neural network module includes a sensitivity feature extraction unit, a state change feature extraction unit, and a noise adding unit.

[0131] The sensitivity feature extraction unit is configured to perform feature extraction on the sample embedding matrix representation, and obtain the sensitivity features shown by the power consumption area, the value attribute of the power consumption unit, and the total power consumption value attribute at the target time respectively.

[0132] The state change feature extraction unit is configured to extract the state change features shown by the sample quadruple at the target time according to the sensitivity features and the target embedding matrix representations of at least one target quadruple in which the time object obtained from the quadruple dataset represents the target time.

[0133] The noise adding unit is configured to perform noise adding processing on the state change features to obtain the sensitivity feature change rate.

[0134] According to an embodiment of the present invention, the noise adding unit includes a random noise vector calculation sub-unit and a noise adding sub-unit.

[0135] A random noise vector calculation subunit, configured to calculate a random noise vector according to the sensitivity feature, the target time, and the target embedding matrix representation.

[0136] A noise addition subunit, configured to perform noise addition processing on the state change feature by using the random noise vector to obtain a sensitivity feature change rate.

[0137] According to an embodiment of the present invention, other objects include relationship objects and other entity objects in the sample quadruple except the target entity object, and the sensitivity feature change rate includes a relationship sensitivity feature change rate related to the relationship object and an entity sensitivity feature change rate related to the other entity object. The third neural network module includes a state feature extraction unit, a relationship sensitivity matching degree evaluation value calculation unit, an entity sensitivity matching degree evaluation value calculation unit, and a matching degree evaluation value prediction result calculation unit.

[0138] A state feature extraction unit, configured to extract the state feature exhibited by the sample quadruple at the target time according to the sensitivity feature and the sample embedding matrix representation.

[0139] A relationship sensitivity matching degree evaluation value calculation unit, configured to calculate a relationship sensitivity matching degree evaluation value between the prediction information and the real information represented by the relationship object according to the prediction information and the relationship sensitivity feature change rate.

[0140] An entity sensitivity matching degree evaluation value calculation unit, configured to calculate an entity sensitivity matching degree evaluation value between the prediction information and the real information represented by the other entity object according to the prediction information and the entity sensitivity feature change rate.

[0141] A matching degree evaluation value prediction result calculation unit, configured to calculate a matching degree evaluation value prediction result of the prediction information relative to the real information according to the relationship sensitivity matching degree evaluation value, the entity sensitivity matching degree evaluation value, and the state feature.

[0142] According to an embodiment of the present invention, there are multiple matching degree evaluation value prediction results. The loss module includes a normalization unit, a conditional probability calculation unit, and a training loss calculation unit.

[0143] A normalization unit, configured to perform normalization processing on the multiple matching degree evaluation value prediction results to obtain a matching degree evaluation value normalization result.

[0144] A conditional probability calculation unit, configured to calculate the conditional probability of the matching degree evaluation value real label under the condition represented by the multiple matching degree evaluation value normalization results according to the matching degree evaluation value real label and the matching degree evaluation value normalization result.

[0145] A training loss calculation unit, configured to calculate a training loss according to a conditional probability, a relationship sensitivity matching degree evaluation value, and an entity sensitivity matching degree evaluation value.

[0146] According to an embodiment of the present invention, the relationship sensitivity matching degree evaluation value has a relationship sensitivity weight coefficient, and the entity sensitivity matching degree evaluation value has an entity sensitivity weight coefficient. The training module includes an adjustment unit.

[0147] The adjustment unit is configured to adjust the relationship sensitivity weight coefficient and the entity sensitivity weight coefficient by using the training loss.

[0148] According to an embodiment of the present invention, the first neural network module includes an encoding unit and a matrix conversion unit.

[0149] The encoding unit is configured to perform one-hot encoding on the power consumption area represented by the head entity object, the power consumption unit value attribute represented by the relationship object, the total power consumption value attribute represented by the tail entity object, and the timestamp information represented by the time object respectively, to obtain an area binary vector, a unit value attribute binary vector, a total value attribute binary vector, and a time binary vector.

[0150] The matrix conversion unit is configured to perform matrix conversion on the area binary vector, the unit value attribute binary vector, the total value attribute binary vector, and the time binary vector to obtain a sample embedding matrix representation.

[0151] Figure 7 The block diagram of the power consumption information prediction device according to an embodiment of the present invention is shown.

[0152] As Figure 7 shown, the power consumption information prediction device 700 includes an incomplete quadruple acquisition module 710 and a completion module 720.

[0153] The incomplete quadruple acquisition module 710 is configured to acquire an incomplete quadruple composed of a head entity object representing a power consumption area, a relationship object representing a power consumption unit value attribute, a tail entity object representing a total power consumption value attribute, and a time object representing timestamp information, where the information represented by the incomplete object in the incomplete quadruple is empty, and the incomplete object represents a head entity object or a tail entity object.

[0154] The completion module 720 is configured to input the incomplete quadruple into the power consumption information prediction model to obtain a target completed result determined for the incomplete object, and output a completed quadruple including the target completed result, where the power consumption information prediction model is trained by using the training method of the above-mentioned power consumption information prediction model.

[0155] According to an embodiment of the present invention, the completion module includes a candidate completed result obtaining unit, a candidate matching degree evaluation value prediction result calculating unit, and a target completed result determining unit.

[0156] The candidate completed result obtaining unit is configured to input the quadruple to be completed into the electricity consumption information prediction model to obtain a plurality of candidate completed results determined for the object to be completed.

[0157] The candidate matching degree evaluation value prediction result calculating unit is configured to calculate the candidate matching degree evaluation value prediction results of each candidate completed result with respect to the real information represented by other objects in the quadruple to be completed, to obtain a plurality of candidate matching degree evaluation value prediction results, where the other objects represent the objects other than the object to be completed in the quadruple to be completed.

[0158] The target completed result determining unit is configured to determine the candidate completed result corresponding to the target matching degree evaluation value prediction result with the largest value among the plurality of candidate matching degree evaluation value prediction results calculated as the target completed result.

[0159] Any multiple of the modules, units, and subunits according to the embodiments of the present invention, or at least part of the functions of any of them, can be implemented in one module. Any one or more of the modules, units, and subunits according to the embodiments of the present invention can be split into multiple modules for implementation. Any one or more of the modules, units, and subunits according to the embodiments of the present invention can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on substrate, a system on package, an application specific integrated circuit (ASIC), or can be implemented by any other reasonable way of integrating or packaging circuits, in hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in a suitable combination of any several of them. Alternatively, one or more of the modules, units, and subunits according to the embodiments of the present invention can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding functions can be executed.

[0160] For example, any one or more of the first neural network module 610, the second neural network module 620, the third neural network module 630, the loss module 640, and the training module 650, or the incomplete quadruple acquisition module 710 and the completion module 720 can be combined and implemented in one module / unit / sub-unit, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of these modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present invention, at least one of the first neural network module 610, the second neural network module 620, the third neural network module 630, the loss module 640, and the training module 650, or the incomplete quadruple acquisition module 710 and the completion module 720 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or any other reasonable way of integrating or packaging circuits, etc., in hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in a suitable combination of any several of them. Alternatively, at least one of the first neural network module 610, the second neural network module 620, the third neural network module 630, the loss module 640, and the training module 650, or the incomplete quadruple acquisition module 710 and the completion module 720 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding functions can be executed.

[0161] Figure 8 FIG. shows a block diagram of an electronic device suitable for implementing at least one of the training method and the electricity consumption information prediction method of the electricity consumption information prediction model according to an embodiment of the present invention. Figure 8 The shown electronic device is only an example and should not impose any limitation on the functions and the scope of use of the embodiments of the present invention.

[0162] As Figure 8As shown, the electronic device 800 according to an embodiment of the present invention includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage section 808 into a random access memory (RAM) 803. The processor 801 can include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), and so on. The processor 801 can also include on-board memory for caching purposes. The processor 801 can include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0163] In the RAM 803, various programs and data required for the operation of the electronic device 800 are stored. The processor 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. The processor 801 performs various operations of the method flow according to an embodiment of the present invention by executing programs in the ROM 802 and / or the RAM 803. It should be noted that the program can also be stored in one or more memories other than the ROM 802 and the RAM 803. The processor 801 can also perform various operations of the method flow according to an embodiment of the present invention by executing programs stored in the one or more memories.

[0164] According to an embodiment of the present invention, the electronic device 800 can also include an input / output (I / O) interface 805, and the input / output (I / O) interface 805 is also connected to the bus 804. The system 800 can also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed so that a computer program read from it can be installed into the storage section 808 as needed.

[0165] According to an embodiment of the present invention, the method flow according to the embodiment of the present invention can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the processor 801, the above functions defined in the system of the embodiment of the present invention are executed. According to an embodiment of the present invention, the above-described system, device, apparatus, module, unit, etc. can be implemented by computer program modules.

[0166] The present invention also provides a computer-readable storage medium, which can be included in the device / device / system described in the above embodiment; or can exist separately without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0167] According to an embodiment of the present invention, the computer-readable storage medium can be a non-volatile computer-readable storage medium. For example, it can include but is not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present invention, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device.

[0168] For example, according to an embodiment of the present invention, the computer-readable storage medium can include the above-described ROM 802 and / or RAM 803 and / or one or more memories other than ROM 802 and RAM 803.

[0169] An embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method provided by the embodiment of the present invention. When the computer program product runs on an electronic device, the program code is used to cause the electronic device to implement at least one of the training method and the prediction method of the electricity consumption information prediction model provided by the embodiment of the present invention.

[0170] When the computer program is executed by the processor 801, the above functions defined in the system / apparatus of the embodiments of the present invention are executed. According to the embodiments of the present invention, the systems, apparatuses, modules, units, etc. described above can be implemented by computer program modules.

[0171] In one embodiment, the computer program can rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program can also be transmitted and distributed in the form of signals on a network medium, and be downloaded and installed through the communication part 809, and / or be installed from the removable medium 811. The program code included in the computer program can be transmitted by any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0172] The embodiments of the present invention have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present invention is defined by the appended claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and these substitutions and modifications should fall within the scope of the present invention.

Claims

1. A training method for an electricity consumption information prediction model, characterized in that, The method includes: Inputting a sample quadruple composed of a head entity object representing an electricity - consuming area, a relationship object representing the value attribute of an electricity - consuming unit, a tail entity object representing the total electricity - value attribute, and a time object representing timestamp information into the first neural network of the electricity - consumption information prediction model to obtain a sample embedding matrix representation of the sample quadruple; Inputting the sample embedding matrix representation into the second neural network of the electricity - consumption information prediction model to obtain the sensitivity feature change rates of the electricity - consuming area, the value attribute of the electricity - consuming unit, and the total electricity - value attribute represented by the sample quadruple with respect to the target time represented by the time object; Inputting the sensitivity feature change rates into the third neural network of the electricity - consumption information prediction model to calculate a prediction result of the matching degree evaluation value of the prediction information determined for the target entity object in the sample quadruple with respect to the real information represented by the other objects in the sample quadruple, where the target entity object represents the head entity object or the tail entity object, the other objects represent the objects in the sample quadruple except the target entity object, and the sample quadruple has a real label of the matching degree evaluation value; Calculating a training loss according to the sample embedding matrix representation, the real label of the matching degree evaluation value, and the prediction result of the matching degree evaluation value; Adjusting the model parameters of the first neural network, the second neural network, and the third neural network respectively by using the training loss to obtain a trained electricity - consumption information prediction model.

2. The method according to claim 1, characterized in that, The step of inputting the sample embedding matrix representation into the second neural network of the electricity - consumption information prediction model to obtain the sensitivity feature change rates of the electricity - consuming area, the value attribute of the electricity - consuming unit, and the total electricity - value attribute represented by the sample quadruple with respect to the target time represented by the time object includes: Performing feature extraction on the sample embedding matrix representation to obtain the sensitivity features of the electricity - consuming area, the value attribute of the electricity - consuming unit, and the total electricity - value attribute respectively at the target time; Extracting the state change features of the sample quadruple at the target time according to the sensitivity features and the target embedding matrix representations of at least one target quadruple in the quadruple dataset, where the time object represents the target time; Performing noise - adding processing on the state change features to obtain the sensitivity feature change rates.

3. The method according to claim 2, wherein The step of performing noise - adding processing on the state change features to obtain the sensitivity feature change rates includes: Calculating a random noise vector according to the sensitivity features, the target time, and the target embedding matrix representation; Using the random noise vector to perform noise - adding processing on the state change features to obtain the sensitivity feature change rates.

4. The method according to claim 2, characterized in that, The other objects include the relationship object and the other entity objects in the sample quadruple except the target entity object, and the sensitivity feature change rates include a relationship sensitivity feature change rate related to the relationship object and an entity sensitivity feature change rate related to the other entity objects; Inputting the change rate of the sensitivity feature into the third neural network of the electricity consumption information prediction model, and calculating the prediction result of the matching degree evaluation value of the prediction information determined for the target entity object in the sample quadruple relative to the true information represented by other objects in the sample quadruple includes: Extracting the state features shown by the sample quadruple at the target moment according to the sensitivity feature and the sample embedding matrix representation; Calculating the relationship sensitivity matching degree evaluation value between the prediction information and the true information represented by the relationship object according to the prediction information and the change rate of the relationship sensitivity feature; Calculating the entity sensitivity matching degree evaluation value between the prediction information and the true information represented by the other entity objects according to the prediction information and the change rate of the entity sensitivity feature; Calculating the prediction result of the matching degree evaluation value of the prediction information relative to the true information according to the relationship sensitivity matching degree evaluation value, the entity sensitivity matching degree evaluation value and the state features; 5. The method according to claim 4, characterized in that, There are multiple prediction results of the matching degree evaluation value; Calculating the training loss according to the sample embedding matrix representation, the true label of the matching degree evaluation value and the prediction result of the matching degree evaluation value includes: Performing normalization processing on multiple prediction results of the matching degree evaluation value to obtain the normalized result of the matching degree evaluation value; Calculating the conditional probability of the true label of the matching degree evaluation value under the conditions represented by multiple normalized results of the matching degree evaluation value according to the true label of the matching degree evaluation value and the normalized result of the matching degree evaluation value; Calculating the training loss according to the conditional probability, the relationship sensitivity matching degree evaluation value and the entity sensitivity matching degree evaluation value; 6. The method according to claim 4, wherein The relationship sensitivity matching degree evaluation value has a relationship sensitivity weight coefficient, and the entity sensitivity matching degree evaluation value has an entity sensitivity weight coefficient; Adjusting the model parameters of the first neural network, the second neural network and the third neural network respectively by using the training loss includes: Adjusting the relationship sensitivity weight coefficient and the entity sensitivity weight coefficient by using the training loss; 7. The method according to claim 1, wherein Inputting the sample quadruple composed of the head entity object representing the electricity consumption area, the relationship object representing the value attribute of the electricity consumption unit, the tail entity object representing the total value attribute of the electricity consumption, and the time object representing the timestamp information into the first neural network of the electricity consumption information prediction model, and obtaining the sample embedding matrix representation of the sample quadruple includes: Performing one-hot encoding on the electricity consumption area represented by the head entity object, the value attribute of the electricity consumption unit represented by the relationship object, the total value attribute of the electricity consumption represented by the tail entity object, and the timestamp information represented by the time object respectively to obtain the area binary vector, the unit value attribute binary vector, the total value attribute binary vector, and the time binary vector; Performing matrix transformation on the area binary vector, the unit value attribute binary vector, the total value attribute binary vector, and the time binary vector to obtain the sample embedding matrix representation; 8. A method for predicting power consumption information, characterized in that, The method includes: Obtain an incomplete quadruple composed of a head entity object representing the power consumption area, a relationship object representing the value attribute of the power consumption unit, a tail entity object representing the total power consumption value attribute, and a time object representing the timestamp information, wherein the information represented by the incomplete object in the incomplete quadruple is empty, and the incomplete object represents the head entity object or the tail entity object; Input the incomplete quadruple into the power consumption information prediction model to obtain a target completed result determined for the incomplete object, and output a completed quadruple including the target completed result, wherein the power consumption information prediction model is trained by using the method described in any one of claims 1-7.

9. The method according to claim 8, characterized in that The inputting the incomplete quadruple into the power consumption information prediction model to obtain a target completed result determined for the incomplete object includes: Input the incomplete quadruple into the power consumption information prediction model to obtain multiple candidate completed results determined for the incomplete object; Calculate the predicted results of the candidate matching degree evaluation values of each candidate completed result relative to the real information represented by other objects in the incomplete quadruple, and obtain multiple predicted results of the candidate matching degree evaluation values, where the other objects represent the objects in the incomplete quadruple except the incomplete object; Determine the candidate completed result corresponding to the target matching degree evaluation value prediction result with the largest value among the multiple predicted results of the candidate matching degree evaluation values as the target completed result.

10. A training device for an electricity consumption information prediction model, characterized in that, The device includes: A first neural network module for inputting a sample quadruple composed of a head entity object representing the power consumption area, a relationship object representing the value attribute of the power consumption unit, a tail entity object representing the total power consumption value attribute, and a time object representing the timestamp information into the first neural network of the power consumption information prediction model to obtain a sample embedding matrix representation of the sample quadruple; A second neural network module for inputting the sample embedding matrix representation into the second neural network of the power consumption information prediction model to obtain the sensitivity feature change rates of the power consumption area, the value attribute of the power consumption unit, and the total power consumption value attribute represented by the sample quadruple relative to the target time represented by the time object; A third neural network module for inputting the sensitivity feature change rate into the third neural network of the power consumption information prediction model to calculate the predicted result of the matching degree evaluation value of the predicted information determined for the target entity object in the sample quadruple relative to the real information represented by other objects in the sample quadruple, where the target entity object represents the head entity object or the tail entity object, the other objects represent the objects in the sample quadruple except the target entity object, and the sample quadruple has a real label of the matching degree evaluation value; A loss module for calculating a training loss according to the sample embedding matrix representation, the real label of the matching degree evaluation value, and the predicted result of the matching degree evaluation value; A training module, configured to adjust the model parameters of the first neural network, the second neural network, and the third neural network respectively by using the training loss, so as to obtain a trained electricity consumption information prediction model.

Images