CN117236380A - Power system fault prediction method, system, electronic equipment and medium - Google Patents
Power system fault prediction method, system, electronic equipment and medium Download PDFInfo
- Publication number
- CN117236380A CN117236380A CN202311348865.2A CN202311348865A CN117236380A CN 117236380 A CN117236380 A CN 117236380A CN 202311348865 A CN202311348865 A CN 202311348865A CN 117236380 A CN117236380 A CN 117236380A
- Authority
- CN
- China
- Prior art keywords
- power system
- model
- operation data
- fault
- power
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000012549 training Methods 0.000 claims abstract description 46
- 238000013139 quantization Methods 0.000 claims abstract description 35
- 238000013138 pruning Methods 0.000 claims abstract description 16
- 230000008447 perception Effects 0.000 claims abstract description 15
- 238000010606 normalization Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 9
- 238000004140 cleaning Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000005457 optimization Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 4
- 239000013598 vector Substances 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 5
- 230000007774 longterm Effects 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000013341 scale-up Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a power system fault prediction method, a power system fault prediction system, electronic equipment and a medium, and relates to the technical field of fault prediction. The method comprises the following steps: acquiring power system operation data at the current moment; the power system operation data comprise loads, voltages, currents and power of all nodes of the power grid; inputting the operation data of the power system of the power grid at the current moment into a fault prediction model to obtain the fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training the multi-scale-level LSTM network by taking the operation data of the power system at the historical moment as input and the fault state corresponding to the operation data of each power system at the historical moment as output. The invention can improve the accuracy of the fault prediction result, thereby improving the safety and reliability of the power system.
Description
Technical Field
The present invention relates to the field of fault prediction technologies, and in particular, to a method, a system, an electronic device, and a medium for predicting a power system fault.
Background
The power system is used as an indispensable infrastructure of the modern society and is responsible for supplying electric energy to meet the electricity demand of people. The power system comprises links of power generation, power transmission, power distribution and the like, wherein a power station converts various energy sources into electric energy and transmits the electric energy to a user terminal, so that the stable operation of a power grid is ensured. However, as economies develop and the process of urbanization advances, the power system continues to scale up, with increased fluctuations and uncertainties in the power load, resulting in power systems facing increasing challenges.
Conventional power system fault prediction methods typically employ rule-based static strategies, such as empirical formulas or optimization models, to implement fault prediction of the power system. However, these conventional methods are often based on fixed assumptions and rules, and cannot adapt to complex and variable power system environments, and cannot cope with emergencies and abnormal situations. Thus, conventional approaches have limitations in terms of complexity and uncertainty, nonlinearity and time-varying and manual intervention in terms of power system failure prediction.
In terms of complexity and uncertainty, a power system is a highly complex dynamic system involving interactions between multiple components. Factors such as load fluctuation, weather change, equipment faults and the like cause the uncertainty of the system state to be increased, and the traditional method is difficult to accurately model and predict.
In terms of nonlinearity and time variability, the nonlinearity and time variability of a power system make the operation rule of the power system difficult to describe by a simple linear model, and the traditional fault prediction method is difficult to adapt to the dynamic change of the operation of the system.
In terms of manual intervention, conventional fault prediction methods typically rely on manual scheduling and control, the efficiency and accuracy of which are limited by personnel experience and level of operation.
In summary, the conventional power system fault prediction method results in low accuracy of the fault prediction result, thereby reducing safety and reliability of the power system.
Disclosure of Invention
The invention aims to provide a power system fault prediction method, a system, electronic equipment and a medium, which can improve the accuracy of a fault prediction result and further improve the safety and reliability of a power system.
In order to achieve the above object, the present invention provides the following solutions:
a power system fault prediction method, comprising:
acquiring power system operation data at the current moment; the power system operation data comprise loads of all nodes of the power grid, voltages of all nodes of the power grid, currents of all nodes of the power grid and power of all nodes of the power grid;
inputting the power system operation data of the power grid at the current moment into a fault prediction model to obtain a fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking power system operation data at a historical moment as input and taking fault states corresponding to the power system operation data at the historical moment as output; the fault state is faulty or has no fault; the model parameters include weights and biases.
Optionally, the determining process of the fault prediction model is as follows:
acquiring power system operation data at a historical moment and fault states corresponding to the power system operation data at the historical moment;
taking the operation data of the power system at the historical moment as input, taking the fault state corresponding to the operation data of each power system at the historical moment as output, and training the multi-scale-level LSTM network by adopting a Seq2Seq learning strategy to obtain a trained model;
quantizing model parameters in the trained model to obtain a quantized model;
and sequentially carrying out quantitative perception training and pruning operation on the quantitative model to obtain a fault prediction model.
Optionally, before the step of training the multi-scale level LSTM network by using the Seq2Seq learning strategy to obtain a trained model, using the power system operation data at the historical moment as input and using the fault state corresponding to each power system operation data at the historical moment as output, the method further includes:
and carrying out data cleaning, normalization operation and structuring treatment on the power system operation data at the historical moment in sequence.
Optionally, the multi-scale level LSTM network includes a nonce layer and a plurality of parallel LSTM layers; the output end of each LSTM layer is connected with the input end of the nonce layer, and the LSTM layer comprises a plurality of LSTM units which are connected in series.
A power system fault prediction system, comprising:
the data acquisition module is used for acquiring the operation data of the power system at the current moment; the power system operation data comprise loads of all nodes of the power grid, voltages of all nodes of the power grid, currents of all nodes of the power grid and power of all nodes of the power grid;
the prediction module is used for inputting the power system operation data of the power grid at the current moment into a fault prediction model to obtain the fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking power system operation data at a historical moment as input and taking fault states corresponding to the power system operation data at the historical moment as output; the fault state is faulty or has no fault; the model parameters include weights and biases.
Optionally, the power system fault prediction system further includes:
the training set acquisition module is used for acquiring the power system operation data at the historical moment and the fault states corresponding to the power system operation data at the historical moment;
the training module is used for taking the power system operation data at the historical moment as input, taking the fault state corresponding to each power system operation data at the historical moment as output, and training the multi-scale-level LSTM network by adopting a Seq2Seq learning strategy to obtain a trained model;
the quantization module is used for quantizing the model parameters in the trained model to obtain a quantized model;
and the optimization module is used for sequentially carrying out quantitative perception training and pruning operation on the quantitative model to obtain a fault prediction model.
Optionally, the power system fault prediction system further includes:
and the preprocessing module is used for sequentially carrying out data cleaning, normalization operation and structuring processing on the power system operation data at the historical moment.
Optionally, the multi-scale level LSTM network includes a nonce layer and a plurality of parallel LSTM layers; the output end of each LSTM layer is connected with the input end of the nonce layer, and the LSTM layer comprises a plurality of LSTM units which are connected in series.
An electronic device, comprising:
the power system fault prediction system comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor runs the computer program to enable the electronic equipment to execute the power system fault prediction method.
A computer readable storage medium storing a computer program which, when executed by a processor, implements a power system fault prediction method as described above.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention inputs the operation data of the power system of the power grid at the current moment into a fault prediction model to obtain the fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking the operation data of the power system at the historical moment as input and the fault state corresponding to the operation data of each power system at the historical moment as output, and the multi-scale level LSTM network is adopted, so that the characteristic extraction on different time scales can be focused, the comprehensive capture of the complex dynamic characteristics of the power system is realized, the accuracy of a fault prediction result can be improved, and the safety and the reliability of the power system are further improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a power system fault prediction method according to an embodiment of the present invention;
fig. 2 is a diagram of a multi-scale level LSTM network according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
The deep learning algorithm can automatically learn features and modes from large-scale data, and has stronger characterization capability. Therefore, the introduction of the deep learning technology brings breakthrough progress to the field of power system safety and improves the system safety and reliability, and the invention aims to provide a method for predicting the power system faults by using the deep learning technology to improve the power system safety.
Step 101: and acquiring the operation data of the power system at the current moment. The power system operation data comprises loads of all nodes of the power grid, voltages of all nodes of the power grid, currents of all nodes of the power grid and power of all nodes of the power grid.
Step 102: and inputting the power system operation data of the power grid at the current moment into a fault prediction model to obtain the fault state of the power grid at the next moment. The fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking power system operation data at a historical moment as input and taking fault states corresponding to the power system operation data at the historical moment as output; the fault state is faulty or has no fault; the model parameters include weights and biases.
In practical application, the determining process of the fault prediction model is as follows:
and acquiring power system operation data at the historical moment and fault states corresponding to the power system operation data at the historical moment.
And taking the power system operation data at the historical moment as input, taking the fault state corresponding to each power system operation data at the historical moment as output, and training the multi-scale-level LSTM network by adopting a Seq2Seq learning strategy to obtain a trained model.
And quantizing model parameters in the trained model to obtain a quantized model.
And sequentially carrying out quantitative perception training and pruning operation on the quantitative model to obtain a fault prediction model.
In practical application, before taking the power system operation data at the historical moment as input and the fault state corresponding to each power system operation data at the historical moment as output, training the multi-scale level LSTM network by adopting the Seq2Seq learning strategy to obtain a trained model, the method further comprises the following steps:
and carrying out data cleaning, normalization operation and structuring treatment on the power system operation data at the historical moment in sequence.
In practical application, the multi-scale level LSTM network comprises a nonce layer and a plurality of parallel LSTM layers; the output end of each LSTM layer is connected with the input end of the nonce layer, and the LSTM layer comprises a plurality of LSTM units which are connected in series.
The invention provides a more specific embodiment for describing the method in detail, which comprises the following steps:
step 1: acquisition and processing of power system operational data
Step 1.1: acquisition of power system operation data
And acquiring operation data of the power system, such as load, voltage, current, power and the like of each node of the power grid, wherein the operation data source of the power system is divided into two parts, and the system and the historical data record are monitored in real time. Historical data is derived from power system operational data over a period of time, which is stored in a database. The real-time monitoring system collects power system operation data in real time from existing power system monitoring equipment.
For a certain node, the power of the node is calculated by adopting the following balance:
wherein P is i Is the power of node i, V i And V j Is the voltage of node i and node j, G ij And B ij Is the real and imaginary parts of the impedance matrix of the power system, θ ij Is the phase angle between nodes i and j.
Step 1.2: preprocessing of data
The collected data needs to be preprocessed, including cleaning, normalization, and structuring.
Data cleaning is mainly to remove missing values, outliers, repeated values, and the like. In this embodiment, the missing values are processed using interpolation, and for outliers or noise data, smoothing techniques may be employed or replaced with reasonable values at neighboring time points.
The data normalization operation is to eliminate the influence caused by the dimension and numerical value differences among different features. One common normalization method is max-min normalization, which is formulated as follows:
wherein X is original data, X min And X max Respectively minimum and maximum values of data, X normalized Is normalized data.
The data structuring process is the processing of data into a form suitable for model input. The present embodiment organizes time series data into a sliding window form as an input of a model. Assume that there is a time-series data x= [ X ] of length T 1 ，x 2 ，...，x T ]It can be processed by sliding window method into multiple sequences with length t,i.e. [ x ] i-t+1 ，x i-t+2 ，...，x i ](i=t, t+1,) T, as input to the model.
Step 1.3: grouping and tagging of data
Firstly, sequencing the operation data of the power system collected at the historical moment according to a time sequence. The data is then divided into training, validation and test sets according to the intended tasks and actual requirements. The training set is used for parameter learning of the model, the verification set is used for adjusting the super parameters of the model, and the test set is used for evaluating the performance of the model.
Second, the tag of the data refers to marking the data sample with a corresponding flag to indicate the status or event of the power system. The labels of the data may be defined differently depending on the particular task. In this embodiment, a data sample for a period of time in which a fault occurs is marked as a "1" indicating a positive sample (i.e., a sample in which a fault occurs), and a data sample for the remaining time is marked as a "0" indicating a negative sample (i.e., a sample in which no fault occurs).
Step 2: construction and training of LSTM model
The multi-scale LSTM structure (multi-scale level LSTM network) aims to effectively solve the safety problem of the power system. In the structure, the complex dynamic characteristics of the power system are comprehensively captured and modeled by carrying out multi-layer parallel design on a long-short-term memory (LSTM) neural network and data processing on different time scales.
Step 2.1: construction of multi-scale hierarchical LSTM networks
The operation of an electrical power system is a highly dynamic and complex process involving variations on various time scales. For example, power systems may experience rapid load and voltage fluctuations in the short term, and slow changes in equipment aging and potential failure accumulation in the long term. Conventional LSTM models may have difficulty adequately capturing these multi-scale dynamics because they are primarily processing data on a single time scale, and multi-scale LSTM structures contain multiple parallel LSTM layers, each layer processing data on a particular time scale. As shown in FIG. 2, the multi-scale LSTM structure is shown in FIG. 2, and comprises a shallower LSTM layer short_LSTM for handling rapid changes in the short term, a deeper LSTM layer longtime_LSTM for handling slow changes in the long term, and a middle LSTM layer middletime_LSTM for sensing short-term information and deep-term information. In this way, different levels of LSTM networks can focus on feature extraction and modeling on different time scales, enabling the model to more fully understand the evolution process of the power system.
The LSTM unit is composed of three main gating units, which are respectively: input gate (i) t ) Forgetting gate (forget gate, o) t ) And an output gate (f t ). These gating units allow the LSTM unit to selectively receive, forget, and output information at each time step.
In a multi-scale LSTM structure, each LSTM layer is composed of a plurality of LSTM cells, each LSTM layer has its own hidden state and memory cell, processes input data on a respective time scale, and outputs a respective state representation. In order to integrate these different scale information, an efficient fusion method is employed, where the output of each LSTM layer is weighted and summed, and the weights are dynamically adjusted according to the respective time scale and importance. In this way, a comprehensive system state representation is obtained, which contains characteristic information on different time scales.
Step 2.2: training of multiscale hierarchical LSTM networks
To better train and optimize the multiscale LSTM structure, this embodiment introduces a sequence-to-sequence (Seq 2 Seq) learning strategy. This strategy allows the model to take into account the entire sequence of information, not just the current time of day input, when predicting the future state of the power system. By the method, the model can more accurately predict the evolution condition of the power system on different time scales and timely discover possible safety problems.
The following is a detailed description of the Seq2Seq learning strategy:
1. encoder (Encoder):
the encoder is an LSTM that is responsible for processing the input sequence. Each element of the input sequence is processed step by step through the encoder network and then a "context vector" or "encoding vector" is generated at the last time step of the encoder. This encoded vector plays a key role in compressing the entire input sequence information into a fixed length vector representation. For each time scale, a separate encoder network is constructed. The input of the encoder is a sub-sequence of time-series data, which is output as a corresponding encoded vector. These encoded vectors are used as inputs to a multi-scale hierarchical LSTM network.
2. Decoder (Decoder):
the decoder is also an LSTM, which is responsible for generating the output sequence. At each time step, the decoder uses the output of the last time step (typically initially using a special start tag) and the encoded vector generated by the encoder as inputs. The goal of the decoder is to generate an element from these inputs until the entire output sequence is generated or a special termination mark is encountered. For each time scale of a multi-scale hierarchical LSTM network, a separate decoder network is built. Each decoder generates elements of the output sequence from the encoded vector of the corresponding time scale and the prediction result of the previous time step.
3. Teacher force (Teacher force):
during training, teacher enforcement is typically used to accelerate model convergence. Teacher forcing is a technique that uses the true target output sequence as input in the training of the decoder, rather than using the prediction results of the model itself. Doing so may reduce the error propagation problem in training, but during actual reasoning, cumulative errors may occur.
4. Loss function:
we use a cross entropy loss function for measuring the difference between the model generated sequence and the real target sequence. The output sequence of the decoder is compared element by element with the target sequence and the loss is calculated. Optimization algorithms typically use back-propagation algorithms to minimize this loss.
5. Attention mechanism (Attention Mechanism):
for longer input sequences, the encoded vectors produced by the encoder may not capture all important information. To solve this problem, attention mechanisms have been introduced. The attention mechanism allows the decoder to "notice" different portions of the encoder output as each output element is generated, thereby making better use of the information of the input sequence.
Step 3: feedback optimization method of model
Step 3.1: model quantization and real-time feedback system construction
Weights and biases in a multi-scale hierarchical LSTM network are first quantized and converted from a floating point representation to 8-bit integers to preserve model performance as much as possible while reducing storage space and computational complexity. After the quantization parameters, a quantization perception training is performed. During training, quantization errors are introduced and incorporated into the loss function for optimization. Therefore, the model can still keep good prediction performance after quantization, and the performance loss caused by quantization is reduced as much as possible. And further optimizing the quantized LSTM model by combining a model pruning technology. The model scale is further reduced by pruning some unimportant parameters or channels, so that the quantization effect is more remarkable, and meanwhile, the storage requirement and the calculation cost of the model are reduced.
The embodiment of the invention also provides a power system fault prediction system corresponding to the embodiment of the method, which comprises the following steps:
the data acquisition module is used for acquiring the operation data of the power system at the current moment; the power system operation data comprises loads of all nodes of the power grid, voltages of all nodes of the power grid, currents of all nodes of the power grid and power of all nodes of the power grid.
The prediction module is used for inputting the power system operation data of the power grid at the current moment into a fault prediction model to obtain the fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking power system operation data at a historical moment as input and taking fault states corresponding to the power system operation data at the historical moment as output; the fault state is faulty or has no fault; the model parameters include weights and biases.
In practical application, the power system fault prediction system further includes:
the training set acquisition module is used for acquiring the power system operation data at the historical moment and the fault states corresponding to the power system operation data at the historical moment.
The training module is used for taking the power system operation data at the historical moment as input, taking the fault state corresponding to each power system operation data at the historical moment as output, and training the multi-scale-level LSTM network by adopting a Seq2Seq learning strategy to obtain a trained model.
And the quantization module is used for quantizing the model parameters in the trained model to obtain a quantized model.
And the optimization module is used for sequentially carrying out quantitative perception training and pruning operation on the quantitative model to obtain a fault prediction model.
In practical application, the power system fault prediction system further includes:
and the preprocessing module is used for sequentially carrying out data cleaning, normalization operation and structuring processing on the power system operation data at the historical moment.
In practical application, the multi-scale level LSTM network comprises a nonce layer and a plurality of parallel LSTM layers; the output end of each LSTM layer is connected with the input end of the nonce layer, and the LSTM layer comprises a plurality of LSTM units which are connected in series.
The embodiment of the invention also provides electronic equipment, which comprises:
a memory for storing a computer program, and a processor that runs the computer program to cause the electronic device to execute the power system failure prediction method according to the above embodiment.
The embodiment of the invention also provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the power system fault prediction method according to the above embodiment.
The following technical effects are achieved in the power system by applying the multi-scale hierarchical LSTM network:
1. modeling long-term dependency relationship: multi-scale hierarchical LSTM networks are capable of capturing long-term dependencies in power systems, which is difficult to achieve with conventional approaches. Through the memory state of the LSTM, the model can effectively track the historical state and the change trend of the power system, so that the future power system state can be predicted better.
2. High accuracy prediction: a multi-scale hierarchical LSTM network has advantages in processing sequence data that can more accurately predict fault conditions of a power system, including load changes, voltage fluctuations, and the like. The method is favorable for early finding possible safety problems, and corresponding measures are adopted in advance to ensure the stable operation of the power system.
3. Safety improvement of the power system: the fault prediction method provided by the invention can better predict the fault state of the power system, help power system operators to discover and solve problems in time, and improve the safety and stability of the power system.
4. Resource efficiency: through the technologies of model quantization and the like, the storage requirement and the computational complexity of an LSTM model can be remarkably reduced, so that the method is deployed more efficiently in an environment with limited resources, such as mobile equipment, edge computing equipment and the like.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (10)
1. A power system fault prediction method, comprising:
acquiring power system operation data at the current moment; the power system operation data comprise loads of all nodes of the power grid, voltages of all nodes of the power grid, currents of all nodes of the power grid and power of all nodes of the power grid;
inputting the power system operation data of the power grid at the current moment into a fault prediction model to obtain a fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking power system operation data at a historical moment as input and taking fault states corresponding to the power system operation data at the historical moment as output; the fault state is faulty or has no fault; the model parameters include weights and biases.
2. The power system fault prediction method according to claim 1, wherein the determining process of the fault prediction model is:
acquiring power system operation data at a historical moment and fault states corresponding to the power system operation data at the historical moment;
taking the operation data of the power system at the historical moment as input, taking the fault state corresponding to the operation data of each power system at the historical moment as output, and training the multi-scale-level LSTM network by adopting a Seq2Seq learning strategy to obtain a trained model;
quantizing model parameters in the trained model to obtain a quantized model;
and sequentially carrying out quantitative perception training and pruning operation on the quantitative model to obtain a fault prediction model.
3. The power system fault prediction method according to claim 2, wherein before taking power system operation data at a historical moment as input and a fault state corresponding to each power system operation data at the historical moment as output, training the multi-scale level LSTM network by using a Seq2Seq learning strategy to obtain a trained model, further comprises:
and carrying out data cleaning, normalization operation and structuring treatment on the power system operation data at the historical moment in sequence.
4. The power system fault prediction method of claim 1, wherein the multi-scale level LSTM network comprises a nonce layer and a plurality of parallel LSTM layers; the output end of each LSTM layer is connected with the input end of the nonce layer, and the LSTM layer comprises a plurality of LSTM units which are connected in series.
5. A power system fault prediction system, comprising:
the data acquisition module is used for acquiring the operation data of the power system at the current moment; the power system operation data comprise loads of all nodes of the power grid, voltages of all nodes of the power grid, currents of all nodes of the power grid and power of all nodes of the power grid;
the prediction module is used for inputting the power system operation data of the power grid at the current moment into a fault prediction model to obtain the fault state of the power grid at the next moment; the fault prediction model is obtained by carrying out quantization perception training and pruning operation on a quantization model, and the quantization model is obtained by quantizing model parameters in the trained model; the trained model is obtained by training a multi-scale level LSTM network by taking power system operation data at a historical moment as input and taking fault states corresponding to the power system operation data at the historical moment as output; the fault state is faulty or has no fault; the model parameters include weights and biases.
6. The power system fault prediction system according to claim 5, further comprising:
the training set acquisition module is used for acquiring the power system operation data at the historical moment and the fault states corresponding to the power system operation data at the historical moment;
the training module is used for taking the power system operation data at the historical moment as input, taking the fault state corresponding to each power system operation data at the historical moment as output, and training the multi-scale-level LSTM network by adopting a Seq2Seq learning strategy to obtain a trained model;
the quantization module is used for quantizing the model parameters in the trained model to obtain a quantized model;
and the optimization module is used for sequentially carrying out quantitative perception training and pruning operation on the quantitative model to obtain a fault prediction model.
7. The power system fault prediction system according to claim 6, further comprising:
and the preprocessing module is used for sequentially carrying out data cleaning, normalization operation and structuring processing on the power system operation data at the historical moment.
8. The power system fault prediction system of claim 5, wherein the multi-scale level LSTM network comprises a nonce layer and a plurality of parallel LSTM layers; the output end of each LSTM layer is connected with the input end of the nonce layer, and the LSTM layer comprises a plurality of LSTM units which are connected in series.
9. An electronic device, comprising:
a memory for storing a computer program, and a processor that runs the computer program to cause the electronic device to execute the power system failure prediction method according to any one of claims 1 to 4.
10. A computer-readable storage medium, characterized in that it stores a computer program which, when executed by a processor, implements the power system failure prediction method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311348865.2A CN117236380A (en) | 2023-10-17 | 2023-10-17 | Power system fault prediction method, system, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311348865.2A CN117236380A (en) | 2023-10-17 | 2023-10-17 | Power system fault prediction method, system, electronic equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117236380A true CN117236380A (en) | 2023-12-15 |
Family
ID=89091135
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311348865.2A Pending CN117236380A (en) | 2023-10-17 | 2023-10-17 | Power system fault prediction method, system, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117236380A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117572159A (en) * | 2024-01-17 | 2024-02-20 | 成都英华科技有限公司 | Power failure detection method and system based on big data analysis |
-
2023
- 2023-10-17 CN CN202311348865.2A patent/CN117236380A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117572159A (en) * | 2024-01-17 | 2024-02-20 | 成都英华科技有限公司 | Power failure detection method and system based on big data analysis |
| CN117572159B (en) * | 2024-01-17 | 2024-03-26 | 成都英华科技有限公司 | Power failure detection method and system based on big data analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11409347B2 (en) | Method, system and storage medium for predicting power load probability density based on deep learning | |
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| Bhusal et al. | Deep ensemble learning-based approach to real-time power system state estimation | |
| CN105740975A (en) | Data association relationship-based equipment defect assessment and prediction method | |
| CN107944612B (en) | Bus net load prediction method based on ARIMA and phase space reconstruction SVR | |
| CN116245033B (en) | Artificial intelligent driven power system analysis method and intelligent software platform | |
| CN117236380A (en) | Power system fault prediction method, system, electronic equipment and medium | |
| Hua et al. | Multi-timescale lifespan prediction for PEMFC systems under dynamic operating conditions | |
| KR20200128232A (en) | Apparatus for predicting electricity demand and method thereof | |
| Sheng et al. | Short-term load forecasting based on SARIMAX-LSTM | |
| Lee et al. | Energy consumption prediction system based on deep learning with edge computing | |
| CN113743674A (en) | Energy storage output prediction method, system, equipment and medium based on deep learning | |
| CN112803398A (en) | Load prediction method and system based on empirical mode decomposition and deep neural network | |
| CN114638425B (en) | Historical data-based large-user monthly electricity consumption prediction method and system | |
| Yang et al. | Real-time resilience optimization combining an AI agent with online hard optimization | |
| KR102656115B1 (en) | Remaining useful life prediction method of fuel cell system and digital twin device performing the same | |
| CN108108837B (en) | Regional new energy power supply structure optimization prediction method and system | |
| CN117200190A (en) | Electric load prediction method for electric Internet of things | |
| CN115204362A (en) | Method for predicting residual life interval of mechanical equipment | |
| Sun et al. | Reliability prediction of distribution network based on PCA-GA-BP neural network | |
| Da Rosa et al. | Modern computing environment for power system reliability assessment | |
| Mohammadian et al. | Learning solutions for intertemporal power systems optimization with recurrent neural networks | |
| CN113807027A (en) | Health state evaluation model, method and system for wind turbine generator | |
| Gupta | Artificial Intelligence (AI) Applications and Techniques in Smart Grid | |
| CN118132997A (en) | Power grid load prediction method and device, electronic equipment and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |