Flexible load node power analysis method and apparatus, and electronic device

Abstract

The invention discloses a flexible load node power analysis method, a flexible load node power analysis device and electronic equipment. The method comprises the steps that firstly, a target data set corresponding to a power grid is obtained, the target data set at least comprises node data of L nodes in a preset period, the node data are used for representing voltage information and current information of the nodes, then the target data set is input into a target model, and a target power set is obtained, the target power set at least comprises power information of L nodes in a future period, the target model is a model obtained by training a ZIP model and a neural network model according to node data of the L nodes in different historical periods, and the ZIP model is used for describing physical static characteristics of the nodes; the neural network model is used for predicting the power of the nodes according to the voltage fluctuation information of the nodes. The technical problem of low accuracy of predicting the power of the flexible load node in the power grid in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the technical fields of electrical technology and new energy technology. Specifically, it relates to a method, device, and electronic device for analyzing the power of flexible load nodes. Background Art

[0002] With the rapid development of new energy technology and the advancement of smart grids, there are many controllable loads in the power grid that can actively adjust their own electricity consumption behaviors according to incentives. These loads are called flexible loads. The fluctuations in the power of flexible load nodes will affect the stability of the power grid. Therefore, technicians need to adjust the power grid based on the change amount of the power of flexible load nodes. In the prior art, the power of flexible load nodes is usually predicted based on traditional physical models. However, traditional physical models cannot capture the dynamic characteristics of flexible loads participating in voltage regulation. Therefore, it is difficult for traditional physical models in the prior art to effectively simulate the changes of flexible loads under transient conditions such as voltage fluctuations, that is, the prediction accuracy of traditional physical models is limited, resulting in a low accuracy problem in predicting the power of flexible load nodes in the power grid.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] The present application provides a method, device, and electronic device for analyzing the power of flexible load nodes, so as to at least solve the technical problem of low accuracy in predicting the power of flexible load nodes in the power grid in the prior art.

[0005] According to one aspect of the present application, a method for analyzing flexible loads in a power grid is provided, including: obtaining a target data set corresponding to the power grid, where the target data set at least includes node data of L nodes within a preset period, L is a positive integer, and the L nodes are L flexible load devices in the power grid, and the node data is used to characterize the voltage information and current information of the nodes; inputting the target data set into a target model to obtain a target power set, where the target power set at least includes the power information of L nodes in a future period, and the target model is obtained by training a ZIP model and a neural network model based on the node data of L nodes in different historical periods. The ZIP model describes the physical static characteristics of the nodes, and the neural network model predicts the power of the nodes based on the voltage fluctuation information of the nodes.

[0006] Optionally, obtain a target data set corresponding to the power grid, including: collecting current data and voltage data of each of the L nodes every other power cycle, where the current data includes at least the current amplitude and the current phase angle value of the node, and the voltage data includes at least the voltage amplitude and the voltage phase angle value of the node, and the power cycle is used to characterize the frequency of the power grid; storing the current data and voltage data of each of the L nodes in the database based on the acquisition time sequence; generating an original data set according to the current data and voltage data collected in a preset period in the database; and preprocessing the original data set to obtain the target data set.

[0007] Optionally, preprocess the original data set to obtain the target data set, including: obtaining first data and / or second data in the original data set, where the first data is blank current data and / or voltage data in the original data set, and the second data is current data and / or voltage data with abnormal values in the original data set; reconstructing the first data according to the bidirectional Markov interpolation method and the weight optimization method to obtain supplementary data; replacing the first data in the original data set with the supplementary data to obtain a first data set; and deleting the second data in the first data set to obtain the target data set.

[0008] Optionally, the target model is trained through the following steps: generating a training set and a test set according to the node data of the L nodes in different historical periods, where the training set includes the node data of P nodes in different historical periods, the test set includes the node data of Q nodes in different historical periods, P and Q are both positive integers, and L is equal to the sum of P and Q; iteratively training the ZIP model and the neural network model according to the node data of the P nodes included in the training set to obtain an initial model; determining the performance data of the initial model according to the node data of the Q nodes included in the test set; iteratively updating the initial model according to the objective function value corresponding to the initial model obtained each time and the performance data, where the objective function value is used to characterize the error between the power information of the node predicted by the initial model and the power information of the node actually measured; and using the initial model obtained by the last update as the target model.

[0009] Optionally, the model parameters corresponding to the initial model include at least a first parameter, a second parameter, and a third parameter, where the first parameter is used to characterize the physical parameters in the ZIP model, the second parameter is used to characterize the weights in the neural network model, and the third parameter is used to determine the proportion of the ZIP model and the neural network model in the initial model.

[0010] Optionally, iteratively train the ZIP model and the neural network model based on the node data of the P nodes included in the training set to obtain an initial model, including: fusing the ZIP model and the neural network model to obtain a neuro-physical semi-ZIP model; iteratively training the neuro-physical semi-ZIP model based on the node data of the P nodes included in the training set to obtain a first parameter, a second parameter, and a third parameter; determining the initial model according to the first parameter, the second parameter, and the third parameter.

[0011] Optionally, before iteratively updating the initial model according to the objective function value and performance data corresponding to the initial model obtained by each training, it includes: determining a first function value according to the first parameter, where the first function value is used to characterize the error between the power information of the nodes predicted by the ZIP model and the power information of the nodes obtained by actual measurement; determining a second function value according to the second parameter, where the second function value is used to characterize the error between the power information of the nodes predicted by the neural network model and the power information of the nodes obtained by actual measurement; determining the objective function value according to the first function value, the second function value, and the third parameter.

[0012] According to another aspect of the present application, there is provided an analysis device for the power of flexible load nodes, including: an acquisition unit, configured to acquire a target data set corresponding to a power grid, where the target data set includes at least the node data of L nodes within a preset period, L is a positive integer, and the L nodes are L flexible load devices in the power grid, and the node data is used to characterize the voltage information and current information of the nodes; an input unit, configured to input the target data set into a target model to obtain a target power set, where the target power set includes at least the power information of L nodes in a future period, and the target model is trained based on the node data of L nodes in different historical periods for the ZIP model and the neural network model, the ZIP model describes the physical static characteristics of the nodes, and the neural network model predicts the power of the nodes based on the voltage fluctuation information of the nodes.

[0013] According to still another aspect of the present application, there is provided a computer program product, which includes a computer program. When the computer program runs, it controls the computer program product to execute the above-mentioned analysis method for the power of flexible load nodes.

[0014] According to the last aspect of the present application, there is provided an electronic device, including one or more processors and a memory. The memory is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the above-mentioned analysis method for the power of flexible load nodes.

[0015] In this application, first, a target data set corresponding to the power grid is obtained. The target data set includes at least the node data of L nodes within a preset period, where L is a positive integer, and the L nodes are L flexible load devices in the power grid. The node data is used to represent the voltage information and current information of the nodes. Then, this application inputs the target data set into the target model to obtain a target power set. The target power set includes at least the power information of L nodes in a future period. The target model is a model obtained by training the ZIP model and the neural network model based on the node data of L nodes in different historical periods. The ZIP model is used to describe the physical static characteristics of the nodes, and the neural network model is used to predict the power of the nodes based on the voltage fluctuation information of the nodes.

[0016] As can be seen from the above, this application predicts the power of flexible load nodes through a semi-ZIP model (i.e., the target model) obtained by fusing the ZIP model and the neural network model. The semi-ZIP model is a fusion model obtained by training the ZIP model and the neural network model based on the node data of different flexible nodes in different historical periods. Among them, the ZIP model is used to describe the physical static characteristics of flexible load nodes, and the introduced neural network model makes up for the deficiency that the ZIP model cannot capture the voltage fluctuations of flexible load nodes, thus achieving the purpose of bidirectionally capturing the static characteristics and voltage fluctuation characteristics of nodes through the target model, and further improving the accuracy of the predicted power of flexible load nodes.

[0017] Thus, this application realizes the technical effect of improving the accuracy of the predicted power of flexible load nodes by bidirectionally capturing the static characteristics and voltage fluctuation characteristics of nodes through the target model, and further solves the technical problem of low accuracy in predicting the power of flexible load nodes in the power grid in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of this application and form a part of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0019] Figure 1 is a flowchart of an optional method for analyzing flexible loads in a power grid according to an embodiment of this application;

[0020] Figure 2 is a topological structure diagram of flexible load devices in a distribution network according to an embodiment of this application;

[0021] Figure 3 is a schematic diagram of data measurement of a flexible load node according to an embodiment of this application;

[0022] Figure 4 It is a structural diagram of an optional neurophysical semi-ZIP model according to an embodiment of the present application;

[0023] Figure 5 It is a schematic diagram of an application scenario of an optional flexible load node power prediction method according to an embodiment of the present application;

[0024] Figure 6 It is a schematic diagram of an optional analysis device for flexible load node power according to an embodiment of the present application;

[0025] Figure 7 It is a schematic diagram of an optional electronic device according to an embodiment of the present application. Detailed implementation manners

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] It should also be noted that the relevant information (including but not limited to voltage information and current information) and data (including but not limited to the data for display and the data for analysis) involved in the present application are all information and data authorized by the user or fully authorized by all parties. For example, an interface is set between the present system and the relevant users or institutions. Before obtaining the relevant information, a request for obtaining needs to be sent to the aforementioned users or institutions through the interface, and after receiving the consent information feedback from the aforementioned users or institutions, the relevant information can be obtained.

[0029] In addition, in the process of collecting, storing, using, processing, transmitting, providing, disclosing, and applying relevant information and relevant data involved in this application, all comply with relevant laws, regulations, and standards in the relevant regions, and necessary confidentiality measures are taken, without violating public order and good customs. In addition, this application provides corresponding operation entrances for users to choose to agree to authorize or refuse to authorize. If the user chooses to refuse to authorize, the corresponding expert decision-making process will be entered.

[0030] In an alternative embodiment, with the rapid development of renewable energy and the advancement of smart grids, the power supply structure of the power system is constantly reformed, and the penetration rate of new energy sources such as wind power in the power system is continuously increasing. Its randomness and volatility pose severe challenges to the supply-demand balance of the power system. At the same time, the proportion of traditional regulating units, such as thermal power units, is continuously decreasing. To ensure the safe and stable operation of the power system, it is required that the power system has more flexibility to absorb the power imbalance caused by uncertainties. Compared with traditional loads, there are many controllable loads in the distribution network that can actively adjust their own electricity consumption behaviors according to incentives. These load devices are called flexible load devices.

[0031] Flexible load devices can be flexibly adjusted when the power system fluctuates, providing a means to balance the load for the power system, thereby improving the stability and security of the power system. Flexible loads include many types, such as temperature-controlled loads with thermal inertia, electric vehicles, and some energy storage devices. Flexible loads can transfer loads in different time periods or adjust the electricity consumption within a certain range. Therefore, the flexible load resources connected to the distribution network have spatio-temporal dependence characteristics and are closely related to the location of the access node and time.

[0032] According to an embodiment of the present application, an embodiment of a method for analyzing flexible loads in a power grid is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0033] The present application provides a power grid flexible load analysis system (hereinafter referred to as the analysis system) for executing the power grid flexible load analysis method in the present application. Figure 1 is a flowchart of an alternative method for analyzing flexible loads in a power grid according to an embodiment of the present application, as Figure 1 shown, the method includes the following steps:

[0034] Step S101, obtain the target data set corresponding to the power grid.

[0035] In step S101, the target data set includes at least the node data of L nodes within a preset period, where L is a positive integer, and the L nodes are L flexible load devices in the power grid. The node data is used to characterize the voltage information and current information of the nodes.

[0036] Optionally, the target data set refers to the node data corresponding to all nodes in the power grid obtained by the analysis system through controlling the data acquisition system. Among them, the data acquisition system can be set as a PMU (Phasor Measurement Unit).

[0037] Optionally, the node data includes the voltage information and current information measured by the flexible load nodes in the power grid within a preset period. The flexible load device refers to a device that can actively adjust its power consumption or power demand. Figure 2 It is an optional topological structure diagram of flexible load devices in a distribution network according to an embodiment of the present application.

[0038] Step S102, input the target data set into the target model to obtain a target power set.

[0039] In step S102, the target power set includes at least the power information of L nodes in a future period. The target model is a model obtained by training the ZIP model and the neural network model based on the node data of L nodes in different historical periods. The ZIP model is used to describe the physical static characteristics of the nodes, and the neural network model is used to predict the power of the nodes based on the voltage fluctuation information of the nodes.

[0040] Optionally, the target power set contains the power information of L nodes in a future period, that is, the active power and reactive power that these nodes will consume predicted by the model.

[0041] Optionally, the target model is a fusion model obtained by iteratively training the physical ZIP model and the neural network model at the same time. The target model is used to predict the power information of flexible load devices in a future period.

[0042] Optionally, the historical period is the period when voltage fluctuations or equipment start-stop occur in L nodes, so as to ensure that the target model has sufficient information to learn the fluctuation characteristics of flexible load devices during the training stage, thereby improving the accuracy of the target model trained to predict the power of flexible load nodes.

[0043] Optionally, the constant impedance - current - power ZIP model can describe the voltage static characteristics of the load. Under the condition of ensuring voltage quality, the ZIP model reduces the feeder voltage and thus the load power consumption. The ZIP model is usually used to model flexible loads without a clearly specified category. However, the traditional ZIP model in the prior art cannot reflect the dynamic characteristics of flexible load devices participating in voltage regulation, that is, the existing ZIP model cannot capture complex and diverse flexible load information, and the dynamic response ability of the existing ZIP model to flexible loads is limited. It is difficult to effectively simulate the changes of flexible load devices under transient conditions such as voltage fluctuations, device startup or shutdown. The application of the existing ZIP model lacks flexibility and adaptability.

[0044] Optionally, DNN (Deep Neural Network) is one of the important technologies in the field of machine learning. Through the connection and training of multiple layers of neurons, the DNN model can be used to process complex non - linear problems. The DNN model consists of an input layer, a hidden layer, and an output layer. Among them, the input layer is responsible for receiving input data, the hidden layer is located between the input layer and the output layer, and is used to map the input data to a higher - level feature space to capture complex patterns and relationships in the input data (such as the dynamic characteristics of voltage fluctuations of flexible load nodes captured in this application), and the output layer is responsible for generating the final output result. DNN learns and identifies patterns in data through multi - level abstraction to perform prediction tasks. The neural network model in this application uses a four - layer neural network with 20 neurons in each layer.

[0045] As can be seen from the content recorded in step S102, the analysis system uses the target data set generated in step S101 as the input data of the target model to predict the power consumption of the flexible load device in the future period. Among them, the target model is obtained by fusing the ZIP model and the neural network model. Compared with the existing ZIP model, the target model can not only describe the static characteristics of nodes according to physical laws, but also capture the dynamic change trend of nodes through machine learning, thus improving the accuracy and reliability of the prediction results.

[0046] In the above - mentioned embodiment, steps S101 and S102 together constitute the basic process of the analysis system from data acquisition to power prediction. Among them, step S101 is responsible for acquiring the voltage and current data of the flexible load device, and step S102 analyzes the acquired data through a fusion model including physical characteristics and machine learning to predict the power consumption of the flexible load node in the future period. The above process combines physical modeling and data - driven analysis techniques, thus realizing a more accurate and comprehensive power prediction for flexible load devices.

[0047] As can be seen from the above, in this application, a semi-ZIP model (i.e., the target model) obtained by fusing a ZIP model and a neural network model is used to predict the power of flexible load nodes. The semi-ZIP model is a fusion model obtained by training the ZIP model and the neural network model based on the node data of different flexible nodes in different historical periods. Among them, the ZIP model is used to describe the physical static characteristics of flexible load nodes, and the introduced neural network model makes up for the deficiency that the ZIP model cannot capture the voltage fluctuations of flexible load nodes, thereby achieving the purpose of bidirectionally capturing the static characteristics and voltage fluctuation characteristics of nodes through the target model, and further improving the accuracy of the predicted power of flexible load nodes.

[0048] It can be seen that in this application, through the way of bidirectionally capturing the static characteristics and voltage fluctuation characteristics of nodes by the target model, the technical effect of improving the accuracy of the predicted power of flexible load nodes is achieved, and further the technical problem of low accuracy in predicting the power of flexible load nodes in the power grid in the prior art is solved.

[0049] In an optional embodiment, in the process of obtaining the target data set corresponding to the power grid, the analysis system collects the current data and voltage data of each of the L nodes every other power cycle. Among them, the current data at least includes the current amplitude and the current phase angle value of the node, and the voltage data at least includes the voltage amplitude and the voltage phase angle value of the node. The power cycle is used to characterize the frequency of the power grid. After that, the analysis system stores the current data and voltage data of each of the L nodes into the database based on the acquisition time sequence. Then, the analysis system generates the original data set according to the current data and voltage data collected in the preset period in the database. Finally, the analysis system preprocesses the original data set to obtain the target data set.

[0050] Optionally, the power cycle refers to the time required for the alternating current in the power grid to complete a complete waveform change. The power cycle is closely related to the power grid frequency. For example, when measuring data in a power grid with a frequency of 50 Hz or 60 Hz, the power cycle is 1 / 50 second or 1 / 60 second.

[0051] Optionally, in the process of obtaining the target data set corresponding to the power grid, the analysis system first selects the power cycle to be measured. Every other power cycle time interval, the PMU measurement device is used to measure the current information and voltage information of the flexible load node in real time. The current information at least includes the current amplitude and the current phase angle value, and the voltage information at least includes the voltage amplitude and the voltage phase angle value.

[0052] Optionally, the analysis system can use the DAS (Data Acquisition System) to store the measured data in the database.

[0053] Optionally, Figure 3 is a schematic diagram of data measurement of a flexible load node according to an embodiment of the present application, as Figure 3 shown, the voltage data acquisition information of the flexible load node is shown in the following formula (1).

[0054]

[0055] As Figure 3 shown, the current data acquisition information of the flexible load node is shown in the following formula (2).

[0056]

[0057] In the above formulas (1) and (2), V(t) is the voltage of the flexible node at time t, I(t) is the current of the flexible node at time t, ∠θ v (t) represents the phase angle value of the voltage at time t, ∠θ I (t) represents the phase angle value of the current at time t.

[0058] Optionally, the analysis system stores the current data and voltage data of each of the L nodes in the database based on the acquisition time sequence, ensuring the integrity and time correlation of the collected data, facilitating the consideration of the time series attributes of the data during subsequent analysis, such as the periodic load changes and voltage fluctuations of the power grid. When storing the data, each node's data will generate an identifier corresponding to that node according to the specific time point of its acquisition, so as to facilitate retrieval in chronological order, thereby achieving the purpose of ensuring the continuity and time dependence of the data, and further improving the data processing efficiency.

[0059] In an optional embodiment, in the process of preprocessing the original data set to obtain the target data set, the analysis system first obtains the first data and / or the second data in the original data set, where the first data is the blank current data and / or voltage data in the original data set, and the second data is the abnormally valued current data and / or voltage data in the original data set. Then, the analysis system reconstructs the first data based on the bidirectional Markov interpolation method and the weight optimization method to obtain supplementary data. Then, the analysis system replaces the first data in the original data set with the supplementary data to obtain the first data set. Finally, the analysis system deletes the second data in the first data set to obtain the target data set.

[0060] Optionally, the missing data (i.e., the first data) in the node data may be caused by equipment failures, data recording errors, or signal interruptions, etc., while the abnormal data (i.e., the second data) may be caused by measurement instrument errors or environmental interference.

[0061] Optionally, the bidirectional Markov interpolation method is a time series data interpolation technique. This algorithm utilizes the Markov property of the sequence and bidirectional prediction (i.e., prediction from the front to the back and from the back to the front) to generate missing values. The weight optimization rule adjusts the weight parameters in the interpolation process based on the relevance and importance of the surrounding data points, further improving the accuracy of interpolation. By applying the bidirectional Markov interpolation method and the weight optimization method, the analysis system achieves the purpose of filling in the missing current data and voltage data in the original data set, improving the integrity of the subsequent obtained target data set. Generate supplementary data to complete the original data set.

[0062] Optionally, before training the ZIP model and the neural network model based on the node data of L nodes in different historical periods, the analysis system also needs to perform preprocessing operations on the node data of L nodes in different historical periods. During the preprocessing process, not only the integrity of the model training data is enhanced, but also the reliability of the model training data is improved through techniques such as weight optimization and Markov interpolation, enabling the trained target model to more accurately learn the characteristics of flexible load devices, thereby improving the accuracy and stability of power prediction.

[0063] In an alternative embodiment, the target model is trained through the following steps: First, generate a training set and a test set based on the node data of L nodes in different historical periods. Among them, the training set includes the node data of P nodes in different historical periods, and the test set includes the node data of Q nodes in different historical periods. Both P and Q are positive integers, and L is equal to the sum of P and Q. Then, iteratively train the ZIP model and the neural network model based on the node data of the P nodes included in the training set to obtain an initial model. Next, determine the performance data of the initial model based on the node data of the Q nodes included in the test set. Secondly, iteratively update the initial model based on the objective function value corresponding to the initial model obtained from each training and the performance data. Among them, the objective function value is used to characterize the error between the power information of the nodes predicted by the initial model and the power information of the nodes measured actually. Finally, use the initial model obtained from the last update as the target model.

[0064] Optionally, the objective function value reflects the degree of difference between the model prediction result and the true value and is the goal of model optimization. Through the calculation of the objective function value and combined with the test results corresponding to the test set, the analysis system can evaluate the prediction ability of the model.

[0065] Optionally, the DPO (Differentially Private Optimization) method is used during the process of iterative updating of the model. Through this method, the parameters of the ZIP model and the weights of the neural network are adjusted to minimize the objective function value, which helps the model gradually improve its performance in the power prediction task.

[0066] Optionally, after preprocessing the node data of L nodes in different historical periods, a historical dataset is obtained. The analysis system can use 80% of the node data in the historical dataset as the training set and the remaining 20% of the node data in the historical dataset as the test set. The purpose of this division is to ensure that the model can be trained on a part of the data and then tested on another part of the data that has not participated in the training, so as to evaluate the generalization ability of the trained model.

[0067] Optionally, the model parameters corresponding to the initial model at least include a first parameter, a second parameter, and a third parameter. Among them, the first parameter is used to characterize the physical parameters in the ZIP model, the second parameter is used to characterize the weights in the neural network model, and the third parameter is used to determine the proportion of the ZIP model and the neural network model in the initial model.

[0068] Optionally, the objective function value can be calculated through the following formula (3).

[0069]

[0070] In the above formula (3), the active power P(t,k) and the reactive power Q(t,k) are expressed as a convex combination of the objective functions of the ZIP model and the neural network model. Among them, is the active power calculated by the ZIP model, is the reactive power calculated by the ZIP model, is the active power calculated by the neural network model, is the reactive power calculated by the neural network model. The parameters a and b are scalars (i.e., the third parameter) that determine the output ratio based on the ZIP model and the neural network model. When a = 1 and b = 1, the target model is a pure physical ZIP model. When a = 0 and b = 0, the target model is a pure neural network model. The structural diagram of the neuro-physical semi-ZIP model (i.e., the target model) obtained by fusing the physical ZIP model and the neural network model is as shown in Figure 4 shown.

[0071] Optionally, the calculation formulas for the active power and reactive power calculated by the ZIP model are as shown in the following formula (4).

[0072]

[0073] In the above formula (4), the power calculated by the ZIP model consists of the power consumed by the constant impedance, the power corresponding to the constant current load, and the constant power component. The ZIP model represents the calculated active power and reactive power as second-order polynomial functions of voltage, where, is the initial load active power, is the initial load reactive power, V(t0) is the initial voltage, and α p , α i , α z , β p , β i , β z are the polynomial model coefficients (i.e., the first parameters) in the ZIP model.

[0074] Optionally, the calculation formulas for the active power and reactive power calculated by the neural network model are as shown in the following formula (5).

[0075]

[0076] In the above formula (5), θ in the parameter π θ is the weight of the neural network model (i.e., the second parameter).

[0077] Optionally, the analysis system sets the number of iterations of the current neurophysical semi-ZIP model to m, where m ∈ [1, M max , and M max is the maximum number of iterations. Randomly and non-repeatedly extract sample node data from the current sample set and input it into the initial model. The input is the feature vector set {(V(t, k), θ V (t, k)): t = 1, 2, ……, K} corresponding to the flexible load node data, and the model output is the power information {(P(t, k), Q(t, k)): t = 1, 2, ……, K} of the predicted flexible load node.

[0078] Optionally, during the training process, it is necessary to calculate the loss function and use the loss function for penalty constraints. The expression of the loss function is as shown in the following formula (6).

[0079] L con = Q g ||ReLU((x i , η i )|| l + Q h ||h(x i , η i )|| l

[0080]

[0081] L = L con + L obj (6)

[0082] In the above formula (6), l represents the norm type, Q g and Q h represent the weight factors, and the data samples are composed of the actual power measurement values η i ={P * (t,k), Q * (t,k)}.

[0083] Optionally, the gradient of the calculated value of the loss function each time is used to guide the update of the model parameters, so as to optimize the physical parameters and the neural network weights, making the value of the loss function reach the minimum. Among them, the analysis system can perform gradient optimization through the AdamW (Adaptive Moment Estimation) solver, and the learning rate of the solver can be set to 0.01, so as to obtain a flexible load attribute prediction model (i.e., the target model) of the distribution network that meets the constraint requirements.

[0084] Specifically, the polynomial model coefficients (i.e., the first parameter), the weights of the neural network model (i.e., the second parameter), and the scalar of the output ratio of the ZIP model and the neural network model (i.e., the third parameter) in the ZIP model can be solved by determining the least squares constraint through the DPO method. The general solution method is shown in the following formula (7).

[0085]

[0086] s.t. g(x i , η i ) ≤ 0

[0087] h(x i , η i ) = 0

[0088] x i = π Θ (η i )

[0089]

[0090] In the above formula (7), Ξ represents the sampling data set, i represents the i-th batch of sampling data, x i represents the variable to be optimized, and θ represents the solution of the constrained optimization problem.

[0091] Specifically, the specific solution methods for the first parameter, the second parameter, and the third parameter in this application are shown in the following formula (8).

[0092]

[0093] such that α p + α i + α z = 1

[0094] β p + β i + β z = 1

[0095] α p , α i , α z ≥ 0

[0096] β p , β i , β z ≥ 0

[0097] 0 ≤ a ≤ 1, 0 ≤ b ≤ 1 (8)

[0098] In the above formula (8), K represents the set of time - series period trajectories k, T is the time step of each trajectory k, and the measured active power P * (t, k) and the measured reactive power Q * (t, k) can be calculated from the measured values of voltage and current at the flexible load node.

[0099] Specifically, the calculation formulas for the measured active power P * (t, k) and the measured reactive power Q * (t, k) are as shown in the following formula (9).

[0100] P * (t, k)=V(t, k)I(t, k)cos(θ V (t, k)-θI(t, k))

[0101] Q * (t, k)=V(t, k)I(t, k)sin(θ V (t, k)-θ I (t, k)) (9)

[0102] Optionally, the above embodiments are intended to construct a high-precision flexible load voltage analysis model through fine data partitioning, model training, performance evaluation, and parameter optimization. The overall process of the above embodiments ensures that the target model not only learns the characteristics of flexible load devices in the distribution network during the training process but also verifies the prediction ability of the target model on an independent test set. Through the iterative update mechanism, the accuracy of the model is significantly improved, ensuring that the finally trained target model can effectively predict the power information of nodes, contributing to the stable operation and intelligent scheduling of the power system, thereby improving the efficiency of managing flexible load nodes in the power grid.

[0103] In an alternative embodiment, during the process of iteratively training the ZIP model and the neural network model based on the node data of P nodes included in the training set to obtain an initial model, the analysis system first fuses the ZIP model and the neural network model to obtain a neuro-physical semi-ZIP model. Then, the analysis system iteratively trains the neuro-physical semi-ZIP model based on the node data of P nodes included in the training set to obtain the first parameter, the second parameter, and the third parameter. Finally, the analysis system determines the initial model based on the first parameter, the second parameter, and the third parameter.

[0104] Optionally, the analysis system combines the commonly used physical ZIP model of flexible loads with a neural network, which can increase the flexibility and adaptability of the model. By selecting a differentiable parameter optimization method, the problem of predicting the theoretical attributes of flexible loads is abstracted into a least squares problem in mathematics. By minimizing the sum of squared residuals, the physical parameters and neural network weights in the model are optimized, and then the optimal model parameters are obtained. Under the satisfied imposed constraints, the prediction accuracy can be greatly improved.

[0105] In an alternative embodiment, before iteratively updating the initial model based on the objective function value and performance data corresponding to the initial model obtained from each training, the analysis system first determines a first function value based on the first parameter, where the first function value is used to characterize the error between the power information of the nodes predicted by the ZIP model and the power information of the nodes obtained by actual measurement. Then, the analysis system determines a second function value based on the second parameter, where the second function value is used to characterize the error between the power information of the nodes predicted by the neural network model and the power information of the nodes obtained by actual measurement. Then, the analysis system determines the objective function value based on the first function value, the second function value, and the third parameter.

[0106] Specifically, after the analysis system obtains the first function value and the second function value, the target function value can be calculated by weighted summing the first function value and the second function value based on a third parameter. The first function value can be determined according to the difference between the calculation results of the above formulas (4) and (9), and the second function value can be determined according to the difference between the calculation results of the above formulas (5) and (9).

[0107] According to an aspect of the embodiments of the present application, there is also provided a flexible load node power prediction method Figure 5 which is a schematic diagram of an application scenario of an optional flexible load node power prediction method according to the embodiments of the present application. As Figure 5 shown, the method includes three major stages: data generation, generation of a prediction model, and theoretical attribute prediction.

[0108] Specifically, the specific execution steps of the data generation stage include: using a PMU to measure flexible load node data in real time, using a DAS to store the measured data in a database, processing outliers and missing values in the stored data, and generating an original dataset Dataset suitable for the physical neural semi-ZIP model.

[0109] Specifically, the specific execution steps of the generation of the prediction model stage include: inputting 80% of the data in the original dataset Dataset as training data into the initially constructed physical neural semi-ZIP model, thereby performing iterative training on the initially constructed physical neural semi-ZIP model. The DPO method is selected to optimize the parameters during the training process, so as to obtain a semi-ZIP model that meets the constraint requirements. When the loss function converges or reaches the maximum number of iterative training times, a theoretical prediction model Neuro-ZIPModel (i.e., the target model) is obtained.

[0110] Specifically, the specific execution steps of the theoretical attribute prediction stage include: using the distribution network flexible load attribute prediction model Neuro-ZIP Model to complete the prediction of the theoretical voltage of different flexible load nodes in the distribution network during different power cycles.

[0111] As can be seen from the above, the present application predicts the flexible load node power through a semi-ZIP model (i.e., the target model) obtained by fusing a ZIP model and a neural network model. The semi-ZIP model is a fusion model obtained by training the ZIP model and the neural network model based on the node data of different flexible nodes in different historical cycles. Among them, the ZIP model is used to describe the physical static characteristics of flexible load nodes, and the introduced neural network model makes up for the deficiency that the ZIP model cannot capture the voltage fluctuations of flexible load nodes, thereby achieving the purpose of bidirectionally capturing the static characteristics and voltage fluctuation characteristics of nodes through the target model, and further improving the accuracy of the predicted flexible load node power.

[0112] It can be seen that, through the method of bidirectionally capturing the static characteristics and voltage fluctuation characteristics of nodes by the target model in this application, the technical effect of improving the accuracy of the predicted power of flexible load nodes is achieved, and further the technical problem of low accuracy in predicting the power of flexible load nodes in the power grid in the prior art is solved.

[0113] According to another aspect of the embodiments of the present application, there is also provided an analysis device for the power of flexible load nodes. Figure 6 It is a schematic diagram of an optional analysis device for the power of flexible load nodes according to the embodiments of the present application. As Figure 6 shown, the analysis device for the power of flexible load nodes includes: an acquisition unit 601 and an input unit 602.

[0114] Optionally, the acquisition unit is configured to acquire a target data set corresponding to the power grid, where the target data set at least includes node data of L nodes within a preset period, L is a positive integer, and the L nodes are L flexible load devices in the power grid, and the node data is used to characterize the voltage information and current information of the nodes; the input unit is configured to input the target data set into the target model to obtain a target power set, where the target power set at least includes the power information of L nodes within a future period, and the target model is a model obtained by training a ZIP model and a neural network model based on the node data of L nodes in different historical periods. The ZIP model is used to describe the physical static characteristics of the nodes, and the neural network model is used to predict the power of the nodes based on the voltage fluctuation information of the nodes.

[0115] In an optional embodiment, the acquisition unit includes: a collection subunit, a storage subunit, a generation subunit, and a preprocessing subunit.

[0116] Optionally, the collection subunit is configured to collect the current data and voltage data of each of the L nodes every other power cycle, where the current data at least includes the current amplitude and current phase angle value of the node, and the voltage data at least includes the voltage amplitude and voltage phase angle value of the node. The power cycle is used to characterize the frequency of the power grid; the storage subunit is configured to store the current data and voltage data of each of the L nodes in the database based on the acquisition time sequence; the generation subunit is configured to generate an original data set based on the current data and voltage data collected in the preset period in the database; the preprocessing subunit is configured to preprocess the original data set to obtain a target data set.

[0117] In an optional embodiment, the preprocessing subunit includes: an acquisition module, a reconstruction module, a replacement module, and a deletion module.

[0118] Optionally, an acquisition module is configured to acquire first data and / or second data in an original dataset, where the first data is blank current data and / or voltage data in the original dataset, and the second data is current data and / or voltage data with abnormal values in the original dataset; a reconstruction module is configured to reconstruct the first data according to a bidirectional Markov interpolation method and a weight optimization method to obtain supplementary data; a replacement module is configured to replace the first data in the original dataset with the supplementary data to obtain a first dataset; a deletion module is configured to delete the second data in the first dataset to obtain a target dataset.

[0119] In an alternative embodiment, the flexible load node power analysis device further includes: a generation unit, iterative training, a first determination unit, an iterative update unit, and a second determination unit.

[0120] Optionally, a generation unit is configured to generate a training set and a test set according to node data of L nodes in different historical periods, where the training set includes node data of P nodes in different historical periods, the test set includes node data of Q nodes in different historical periods, P and Q are both positive integers, and L is equal to the sum of P and Q; an iterative training unit is configured to iteratively train a ZIP model and a neural network model according to the node data of the P nodes included in the training set to obtain an initial model; a first determination unit is configured to determine performance data of the initial model according to the node data of the Q nodes included in the test set; an iterative update unit is configured to iteratively update the initial model according to the objective function value and the performance data corresponding to the initial model obtained by each training, where the objective function value is used to characterize the error between the power information of the nodes predicted by the initial model and the power information of the nodes actually measured; a second determination unit is configured to use the initial model obtained by the last update as the target model.

[0121] In an alternative embodiment, the model parameters corresponding to the initial model in the flexible load node power analysis device at least include a first parameter, a second parameter, and a third parameter, where the first parameter is used to characterize the physical parameters in the ZIP model, the second parameter is used to characterize the weights in the neural network model, and the third parameter is used to determine the proportion of the ZIP model and the neural network model in the initial model.

[0122] In an alternative embodiment, the iterative training unit includes: a fusion subunit, a first determination subunit, and a second determination subunit.

[0123] Optionally, a fusion subunit is configured to fuse the ZIP model and the neural network model to obtain a neuro-physical semi-ZIP model; a first determination subunit is configured to iteratively train the neuro-physical semi-ZIP model according to the node data of P nodes included in the training set to obtain a first parameter, a second parameter, and a third parameter; a second determination subunit is configured to determine an initial model according to the first parameter, the second parameter, and the third parameter.

[0124] In an alternative embodiment, the analysis device for the flexible load node power further includes: a third determination unit, a fourth determination unit, and a fifth determination unit.

[0125] Optionally, the third determination unit is configured to determine a first function value according to the first parameter, where the first function value is used to characterize the error between the power information of the node predicted by the ZIP model and the power information of the node obtained by actual measurement; the fourth determination unit is configured to determine a second function value according to the second parameter, where the second function value is used to characterize the error between the power information of the node predicted by the neural network model and the power information of the node obtained by actual measurement; the fifth determination unit is configured to determine an objective function value according to the first function value, the second function value, and the third parameter.

[0126] As can be seen from the above, the present application predicts the flexible load node power through a semi-ZIP model (i.e., the target model) obtained by fusing the ZIP model and the neural network model. The semi-ZIP model is a fusion model obtained by training the ZIP model and the neural network model according to the node data of different flexible nodes in different historical periods. Among them, the ZIP model is used to describe the physical static characteristics of the flexible load node, and the introduced neural network model makes up for the deficiency that the ZIP model cannot capture the voltage fluctuation of the flexible load node, thereby achieving the purpose of bidirectionally capturing the static characteristics and voltage fluctuation characteristics of the node through the target model, and further improving the accuracy of the predicted flexible load node power.

[0127] It can be seen that the present application realizes the technical effect of improving the accuracy of the predicted flexible load node power by bidirectionally capturing the static characteristics and voltage fluctuation characteristics of the node through the target model, and further solves the technical problem of low accuracy in predicting the power of flexible load nodes in the power grid in the prior art.

[0128] According to another aspect of the embodiments of the present application, there is also provided a computer program product, where the computer program product includes a stored computer program, and when the computer program runs, it controls the computer program product to execute the regulation method of the integrated energy system in any one of the above.

[0129] According to another aspect of the embodiments of the present application, there is also provided an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the adjustment method of the integrated energy system in any one of the above by executing the executable instructions.

[0130] Optionally, Figure 7 is a schematic diagram of an optional electronic device according to an embodiment of the present application. As Figure 7 shown, an embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program stored on the memory and executable on the processor. When the processor executes the program, it implements the adjustment method of the integrated energy system in any one of the above.

[0131] The above-described embodiments or examples disclosed in the present application are not exhaustive. They are only schematic representations of some embodiments or examples and do not constitute specific limitations on the protection scope disclosed in the present application. Without contradiction, each step in a certain embodiment or example in the present application can be implemented as an independent embodiment, and the steps can be combined arbitrarily. For example, the solution after removing some steps in a certain embodiment or example can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment or example can be arbitrarily exchanged. Additionally, the optional ways or optional examples in a certain embodiment or example can be combined arbitrarily; furthermore, the various embodiments or examples can be combined arbitrarily. For example, some or all of the steps of different embodiments or examples can be combined arbitrarily, and a certain embodiment or example can be combined arbitrarily with the optional ways or optional examples of other embodiments or examples.

[0132] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0133] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division, and there can 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 displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the units or modules can be in electrical or other forms.

[0134] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] In addition, each functional unit in various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0136] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs and other various media that can store program codes.

[0137] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for analyzing the power of a flexible load node, characterized in that: include: Acquire a target data set corresponding to the power grid, wherein the target data set includes at least node data of L nodes within a preset period, L is a positive integer, the L nodes are L flexible load devices in the power grid, and the node data is used to characterize voltage information and current information of the nodes; The target data set is input into the target model to obtain a target power set, wherein the target power set includes at least the power information of the L nodes in a future period, and the target model is a model obtained by training a ZIP model and a neural network model based on the node data of the L nodes in different historical periods, the ZIP model is used to describe the physical static characteristics of the node, and the neural network model is used to predict the power of the node based on the voltage fluctuation information of the node.

2. The method for analyzing the power of a flexible load node according to claim 1, characterized in that: Obtain the target data set corresponding to the power grid, including: The current data and voltage data of each of the L nodes are collected at intervals of one power cycle, wherein the current data at least includes the current amplitude and the current phase angle value of the node, and the voltage data at least includes the voltage amplitude and the voltage phase angle value of the node, and the power cycle is used to characterize the frequency of the power grid; storing the current data and voltage data of each of the L nodes in a database in order of acquisition time; Generate an original data set based on the current data and voltage data collected in the database within the preset period; The original data set is preprocessed to obtain the target data set.

3. The method for analyzing the power of a flexible load node according to claim 2, characterized in that: Preprocessing the original data set to obtain the target data set includes: Acquire first data and / or second data in the original data set, wherein the first data is blank current data and / or voltage data in the original data set, and the second data is current data and / or voltage data with abnormal values ​​in the original data set; Reconstructing the first data according to a bidirectional Markov interpolation method and a weight optimization method to obtain supplementary data; Replacing the first data in the original data set with the supplementary data to obtain a first data set; The second data in the first data set is deleted to obtain the target data set.

4. The method for analyzing the power of a flexible load node according to claim 1, characterized in that: The target model is trained by the following steps: Generate a training set and a test set based on the node data of the L nodes in different historical periods, wherein the training set includes the node data of P nodes in different historical periods, and the test set includes the node data of Q nodes in different historical periods, P and Q are both positive integers, and L is equal to the sum of P and Q; Iteratively training the ZIP model and the neural network model according to the node data of the P nodes included in the training set to obtain an initial model; Determining performance data of the initial model according to node data of the Q nodes included in the test set; Iteratively updating the initial model according to the objective function value and performance data corresponding to the initial model obtained in each training, wherein the objective function value is used to characterize the error between the power information of the node predicted by the initial model and the power information of the node actually measured; The initial model obtained by the last update is used as the target model.

5. The method for analyzing the power of a flexible load node according to claim 4, characterized in that: The model parameters corresponding to the initial model include at least a first parameter, a second parameter and a third parameter, wherein the first parameter is used to characterize the physical parameters in the ZIP model, the second parameter is used to characterize the weights in the neural network model, and the third parameter is used to determine the proportion of the ZIP model and the neural network model in the initial model.

6. The method for analyzing the power of a flexible load node according to claim 5, characterized in that: The ZIP model and the neural network model are iteratively trained according to the node data of the P nodes included in the training set to obtain an initial model, including: The ZIP model and the neural network model are integrated to obtain a neural physical semi-ZIP model; Iteratively training the neurophysical semi-ZIP model according to the node data of the P nodes included in the training set to obtain a first parameter, a second parameter, and a third parameter; The initial model is determined according to the first parameter, the second parameter and the third parameter.

7. The method for analyzing the power of a flexible load node according to claim 5, characterized in that: Before iteratively updating the initial model according to the objective function value and performance data corresponding to the initial model obtained in each training, it includes: Determine a first function value according to the first parameter, wherein the first function value is used to characterize an error between power information of the node predicted by the ZIP model and power information of the node actually measured; Determine a second function value according to the second parameter, wherein the second function value is used to characterize the error between the power information of the node predicted by the neural network model and the power information of the node actually measured; The objective function value is determined according to the first function value, the second function value and a third parameter.

8. A flexible load node power analysis device, characterized in that: include: An acquisition unit is used to acquire a target data set corresponding to the power grid, wherein the target data set includes at least node data of L nodes within a preset period, L is a positive integer, the L nodes are L flexible load devices in the power grid, and the node data is used to characterize voltage information and current information of the nodes; An input unit is used to input the target data set into a target model to obtain a target power set, wherein the target power set includes at least the power information of the L nodes in a future period, and the target model is a model obtained by training a ZIP model and a neural network model based on the node data of the L nodes in different historical periods, the ZIP model is used to describe the physical static characteristics of the node, and the neural network model is used to predict the power of the node based on the voltage fluctuation information of the node.

9. A computer program product, characterized in that The computer program product comprises a computer program, wherein when the computer program is run, the computer program product is controlled to execute the flexible load node power analysis method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the flexible load node power analysis method described in any one of claims 1 to 7.

Images