Microgrid prediction auxiliary state estimation method considering unknown input

Abstract

The invention relates to the field of artificial intelligence technology application, in particular to a micro-grid prediction auxiliary state estimation method considering unknown input, and the method comprises the following steps: collecting historical data from a micro-grid, including observation EDS variables and unknown EDS variables, and carrying out the preprocessing of the historical data to obtain preprocessed data; designing a neural network model for predicting a parameter relationship between the observation EDS variable and the unknown EDS variable; training the neural network model on the preprocessed data to obtain a state estimation model; and deploying the state estimation model in a power system of the micro-grid, and processing a data stream containing the unknown EDS variable in the power system through the state estimation model, so as to predict the unknown EDS variable in the data stream, and obtain a state estimation result of the micro-grid. According to the method, state estimation with relatively low errors is realized while the data observability is ensured, and the method has relatively good adaptability.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology applications, and particularly to a method for predicting assisted state estimation of a microgrid considering unknown inputs. Background Art

[0002] Predictive assisted state estimation (FASE) is an important technology based on model and measurement data in modern power systems for predicting and estimating voltages and currents. As a smaller-scale power system, the control system, protection system, stability analysis, etc. of a microgrid benefit from the accurate and rapid response of voltages and currents.

[0003] However, for the predictive assisted state estimation of a power system, it is usually necessary to have observability of the entire power system, which means that if some buses are unobservable, it will lead to ineffective or incomplete estimation. Especially in a microgrid, due to cost issues, it is often impossible to install enough sensors, resulting in poor state estimation performance of the microgrid.

[0004] For the above problems, the prior art uses pseudo-measurement technology and a joint estimation method for system state-unknown input to optimize. Among them, the pseudo-measurement technology predicts based on the historical data of the unknown signal and uses it as a signal. The joint estimation method regards the unmeasurable signal as an unknown input and augments it to estimate the system state, and designs an advanced estimation algorithm to jointly estimate the system state and the unknown input.

[0005] However, the disadvantages of the pseudo-measurement technology method are mainly that the data-driven signal prediction is still based on structures such as long short-term memory neural networks and recurrent neural networks. The long-term prediction effects of these methods are extremely poor, and their cumulative errors will reduce the accuracy of microgrid state estimation. Secondly, such data-driven prediction methods usually predict hourly data and have poor prediction effects on the microsecond-level data of state estimation.

[0006] The research on augmenting the unknown input to the system state in the joint estimation of system state-unknown input is in its infancy. In particular, there is still little calculation and discussion on the observability related to this method. At present, this method cannot solve the problem of the observability criterion of a power system containing unknown inputs. Summary of the Invention

[0007] The present invention aims to solve the problems of poor prediction effect and unreliable observability existing in the prior art for the method of predictive assisted state estimation of a microgrid.

[0008] To solve the above technical problems, the present invention provides a method for predicting assisted state estimation of a microgrid considering unknown inputs, including the following steps:

[0009] Collect historical data from the microgrid, where the historical data includes observed EDS variables and unknown EDS variables, and preprocess the historical data to obtain preprocessed data;

[0010] Design a neural network model for predicting the parameter relationship between the observed EDS variables and the unknown EDS variables;

[0011] Train the neural network model on the preprocessed data to obtain a state estimation model;

[0012] Deploy the state estimation model in the power system of the microgrid, and process the data stream containing the unknown EDS variables in the power system through the state estimation model to predict the unknown EDS variables in the data stream, and obtain the microgrid state estimation result.

[0013] Furthermore, the neural network model includes a first input layer, a second input layer, a first LSTM network, a second LSTM network, a first FNN network, a second FNN network, a third FNN network and an output layer. Among them, the first input layer is fully connected to the input end of the first LSTM network, and the second input layer is fully connected to the input end of the second LSTM network, and the input layer is used to receive input data;

[0014] The output end of the first LSTM network is connected to the input end of the first FNN network, and the output end of the second LSTM network is connected to the input end of the first FNN network. The first LSTM network and the second LSTM network are used to process the timing information and long-term dependence relationship in the input data;

[0015] The output ends of the first FNN network and the second FNN network are both connected to the input end of the third FNN network. The first FNN network and the second FNN network are used to perform non-linear transformation and feature extraction on the data output by the first LSTM network and the second LSTM network;

[0016] The output end of the third FNN network is connected to the output layer;

[0017] The output layer is used to output the result of state estimation.

[0018] Furthermore, the method for auxiliary state estimation of microgrid prediction considering unknown inputs further includes the following steps:

[0019] Optimize the hyperparameters of the state estimation model according to the microgrid state estimation result.

[0020] Further, the hyperparameters in the first LSTM network and the second LSTM network are as follows: the number of input layers is 2, the number of hidden layers is 256, the number of hidden layer levels is 4, the length of the input data sequence is 5, and the learning rate is 1 e-4 。

[0021] Further, the hyperparameters in the first FNN network and the second FNN network are as follows: the number of hidden layers is 256, the number of output layers is 1, the number of hidden layer levels is 3, and the learning rate is 1 e-4 。

[0022] Further, the method for preprocessing the historical data is to use the z-score normalization method for processing.

[0023] The beneficial effects achieved by the present invention are as follows: a method for microgrid prediction-assisted state estimation considering unknown inputs based on a neural network is proposed. The neural network used in this method learns from the historical data of the microgrid to predict the unknown EDS variables, and can achieve state estimation with low error while ensuring data observability. Moreover, the neural network used in this method is easy to be deployed in the existing power system and has good adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a schematic flowchart of the steps of the method for microgrid prediction-assisted state estimation considering unknown inputs provided by an embodiment of the present invention;

[0025] Figure 2 is a schematic structural diagram of the neural network model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0027] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of the steps of the method for microgrid prediction-assisted state estimation considering unknown inputs provided by an embodiment of the present invention, and includes the following steps:

[0028] S1. Collect historical data from the microgrid. The historical data includes observed EDS variables and unknown EDS variables, and preprocess the historical data to obtain preprocessed data;

[0029] S2. Design a neural network model for predicting the parameter relationship between the observed EDS variables and the unknown EDS variables;

[0030] S3. Train the neural network model on the preprocessed data to obtain a state estimation model;

[0031] S4. Deploy the state estimation model in the power system of the microgrid, and process the data stream containing the unknown EDS variables in the power system through the state estimation model to predict the unknown EDS variables in the data stream, so as to obtain a microgrid state estimation result.

[0032] The historical data refers to the past circuit performance data in the microgrid, which may also include state estimation data obtained by other methods.

[0033] Furthermore, please refer to Figure 2 , Figure 2 FIG. is a schematic structural diagram of the neural network model provided by an embodiment of the present invention. The neural network model includes a first input layer, a second input layer, a first LSTM network, a second LSTM network, a first FNN network, a second FNN network, a third FNN network, and an output layer. Among them, the first input layer is fully connected to the input end of the first LSTM network, the second input layer is fully connected to the input end of the second LSTM network, and the input layer is used to receive input data;

[0034] The output end of the first LSTM network is connected to the input end of the first FNN network, the output end of the second LSTM network is connected to the input end of the first FNN network, and the first LSTM network and the second LSTM network are used to process the timing information and long-term dependence relationship in the input data;

[0035] The output ends of the first FNN network and the second FNN network are both connected to the input end of the third FNN network, and the first FNN network and the second FNN network are used to perform nonlinear transformation and feature extraction on the data output by the first LSTM network and the second LSTM network;

[0036] The output end of the third FNN network is connected to the output layer;

[0037] The output layer is used to output the result of state estimation.

[0038] Specifically, the neural network model includes two input layers, which respectively receive input data;

[0039] During the implementation process, different input layers are used to input two ringing signals of the power system in the historical data, and the signals are represented as two exponentially damped sine signal sequences.

[0040] In a defined power system, x, y, and u represent the system state, algebraic variables, and system inputs respectively; A, B, and C are the system matrix, input matrix, and measurement matrix respectively. The fluctuations of grid variables, EDS, are modeled as exponentially damped sine curves and satisfy the following relationships:

[0041]

[0042] y = Cx;

[0043] To estimate the voltage and current of the power grid, the predictive auxiliary state estimation method assumes that the power system is quasi - stable and is mathematically modeled using the double exponential smoothing method (Holt):

[0044]

[0045] F K-1 = α(1 + β)I;

[0046] g K-1 = (1 + β)(1 - α)X K-1 - βa k-2 +(1 - β)b K-2 ;

[0047]

[0048] b K-1 = β(a K-1 - a K-2 )+(1 - β)b K-2 ;

[0049] where x is the state vector, F K-1 is the Jacobian matrix of the system function at time step k, α and β are parameters ranging from 0 to 1; g is calculated based on the double exponential smoothing method, a and b depend on the system model (C, V, and λ) and the initial state Z(0) of the oscillation mode, I is the identity matrix, and

[0050] Based on modal analysis theory, in the free motion of the power system, the free oscillation of any signal is the time - domain solution of the following equation, that is, the p - th measurement value can be obtained through the following equation:

[0051]

[0052] Then, it can be obtained that:

[0053]

[0054] λ i = σ i + jω i ;

[0055] a i +jb i =CV i z i (0);

[0056] where Vi is the first row of the left eigenvector of matrix A, Z i (0) is the initial value of the i-th oscillation mode, λi is the eigenvalue of the i-th mode, y i (t) is the i-th mode of the oscillation signal y(t) written in the form of exponentially damped sine wave EDS, λ i is the i-th mode of the power system, σ i , ω i and Z i are the damping, frequency and the initial value of the i-th mode. It can be seen that estimating one EDS based on one EDS is a compensation for the amplitude and initial phase. Therefore, the unknown EDS variable can be estimated through the observed EDS variable. In an embodiment of the present invention, the estimation process is implemented through the neural network model, so as to find a set of parameters that minimize the error θ M between the observed EDS variable and the unknown EDS variable. This process satisfies:

[0057]

[0058] where θ M is the neural network parameter, y p is the measured EDS signal, y m is the unknown EDS signal; fnn is the use of the neural network method. Based on the above expression, in an embodiment of the present invention, the neural network algorithm is used to estimate the unknown EDS variable through the observed EDS variable.

[0059] There are multiple hidden layers between the input layer and the output layer, including two LTSM networks (Long Short-Term Memory networks), respectively labeled as LSTM1 and LSTM2. LSTM is a special recurrent neural network for processing sequence data and can capture long-term dependencies in the data; there are also multiple FNN networks (Feed-Forward Neural Networks) in the hidden layer, respectively labeled as FNN1, FNN2 and FNN3. FNN is a simple neural network structure where information propagates unidirectionally from the input layer to the output layer without feedback connections;

[0060] The output layer is used to generate the final output result.

[0061] It can be understood that the neural network model combines the advantages of LSTM and FNN. LSTM can process time series data and capture the dynamic characteristics and long-term dependencies of various electrical quantities in the microgrid over time, such as the periodicity and volatility of distributed power generation output. FNN, on the other hand, can perform complex non-linear transformations and feature extractions on the input data and can well adapt to the non-linear electrical relationships in the microgrid, such as the non-linear relationships between power, voltage, and current. That is, the state estimation process performed by the neural network model can ensure the observability of the data.

[0062] Furthermore, the method for microgrid prediction-assisted state estimation considering unknown inputs further includes the following steps:

[0063] Optimize the hyperparameters of the state estimation model according to the microgrid state estimation results. Hyperparameters are parameters set by humans and are used to control the training process and structure of the model. In the embodiments of the present invention, for the convenience of processing EDS data in the microgrid, the hyperparameters in the first LSTM network and the second LSTM network are as follows: the number of input layers is 2, the number of hidden layers is 256, the number of hidden layer levels is 4, the length of the input data sequence is 5, and the learning rate is 1 e-4 ;

[0064] The hyperparameters in the first FNN network and the second FNN network are as follows: the number of hidden layers is 256, the number of output layers is 1, the number of hidden layer levels is 3, and the learning rate is 1 e-4 。

[0065] It can be understood that the number of hidden layers in the LSTM layer is configurable to adapt to different data processing requirements and complexities.

[0066] In a possible implementation, the neural network model further includes a module for preprocessing the input data. The preprocessing module is used to perform normalization processing on the input data so that the numerical range of the input data adapts to the input requirements of the neural network model, thereby improving the training efficiency and performance of the neural network model.

[0067] Furthermore, the method for preprocessing the historical data is to use the z-score standardization method for processing.

[0068] During the training process of the neural network model, use the preprocessed data as the training set to train the neural network model. Preferably, adjust the weights and parameters of the network through optimization methods such as the backpropagation algorithm to make the output of the network as close as possible to the actual microgrid state value. During the training process, use the validation set to monitor the performance of the model to prevent overfitting. When the error on the validation set no longer decreases, stop training to obtain the state estimation model.

[0069] The beneficial effects achieved by the present invention lie in proposing a method for microgrid prediction-assisted state estimation based on a neural network that considers unknown inputs. The neural network used in this method learns from the historical data of the microgrid to predict the unknown EDS variables, enabling state estimation with low errors while ensuring data observability. Moreover, the neural network used in this method is easily deployable in existing power systems and has good adaptability.

[0070] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0071] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or device. Without further limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article, or device including that element.

[0072] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0073] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. What is disclosed is only the preferred embodiments of the present invention. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative rather than restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many equivalent changes in form without departing from the spirit of the present invention and the scope protected by the claims, and all of them fall within the protection scope of the present invention.

Claims

1. A method for predictive assisted state estimation of a microgrid considering unknown inputs, characterized in that: The following steps are involved: Collect historical data from the microgrid, the historical data including observed EDS variables and unknown EDS variables, and preprocess the historical data to obtain preprocessed data; Designing a neural network model for predicting the parameter relationship between the observed EDS variables and the unknown EDS variables; Training the neural network model on the preprocessed data to obtain a state estimation model; The state estimation model is deployed in the power system of the microgrid, and the data stream containing the unknown EDS variables in the power system is processed by the state estimation model to predict the unknown EDS variables in the data stream to obtain a microgrid state estimation result.

2. The method for predictive auxiliary state estimation of a microgrid considering unknown input according to claim 1, characterized in that: The neural network model includes a first input layer, a second input layer, a first LSTM network, a second LSTM network, a first FNN network, a second FNN network, a third FNN network and an output layer, wherein the first input layer is fully connected to the input end of the first LSTM network, the second input layer is fully connected to the input end of the second LSTM network, and the input layer is used to receive input data; The output end of the first LSTM network is connected to the input end of the first FNN network, the output end of the second LSTM network is connected to the input end of the first FNN network, and the first LSTM network and the second LSTM network are used to process the timing information and long-term dependency in the input data; The output ends of the first FNN network and the second FNN network are connected to the input end of the third FNN network, and the first FNN network and the second FNN network are used to perform nonlinear transformation and feature extraction on the data output by the first LSTM network and the second LSTM network; The output end of the third FNN network is connected to the output layer; The output layer is used to output the result of state estimation.

3. The method for predictive auxiliary state estimation of a microgrid considering unknown input according to claim 2, characterized in that: The method for predictive auxiliary state estimation of a microgrid considering unknown inputs also includes the following steps: The hyperparameters of the state estimation model are optimized according to the microgrid state estimation result.

4. The method for predictive auxiliary state estimation of a microgrid considering unknown input according to claim 3, characterized in that: The hyperparameters in the first LSTM network and the second LSTM network are: the number of input layers is 2, the number of hidden layers is 256, the number of hidden layers is 4, the length of the input data sequence is 5, and the learning rate is 1. e-4 .

5. The method for predictive auxiliary state estimation of a microgrid considering unknown input according to claim 3, characterized in that: The hyperparameters in the first FNN network and the second FNN network are: the number of hidden layers is 256, the number of output layers is 1, the number of hidden layers is 3, and the learning rate is 1. e-4 .

6. The method for predictive auxiliary state estimation of a microgrid considering unknown input according to claim 3, characterized in that: The method for preprocessing the historical data is to use the z-score standardization method for processing.

Images