Fault early warning method and system for multi-source energy storage station and multi-source energy storage station

Abstract

The invention discloses a fault early warning method and system for a multi-source energy storage station and the multi-source energy storage station, and relates to the technical field of multi-source energy storage stations. According to the method, the operation parameter data of the multi-source energy storage station are obtained, then the operation parameter data are input into a fault early warning model to obtain a fault prediction value output by the fault early warning model, finally the fault prediction value is compared with a preset early warning threshold value, and the fault prediction value exceeding the preset early warning threshold value is used as early warning information. The fault early warning model is an LSTM (Long Short Term Memory) model trained based on a fault feature discrimination result, and the fault feature discrimination result is an association relationship between historical operation parameter data of the multi-source energy storage station and a fault type obtained by inputting historical fault data of the multi-source energy storage station into a fault discrimination classification model; accurate early warning of the fault of the multi-source energy storage station is realized. Therefore, the problem that the accuracy of fault early warning is affected due to the fact that key fault precursor information is ignored when an existing analysis method is adopted is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-source energy storage stations, and more specifically, to a fault warning method and system for a multi-source energy storage station and a multi-source energy storage station. Background Art

[0002] As the proportion of renewable energy continues to increase worldwide, energy storage technology has become increasingly critical as an important bridge connecting unstable power sources to the power grid. Multi-source energy storage power station systems, especially facilities containing large-scale lithium battery arrays, have become an indispensable part of modern smart grids, assuming multiple responsibilities such as peak-to-valley shifting, energy transfer, and frequency regulation. However, due to the complexity of the battery itself, coupled with the diversity and unpredictability of the operating environment, the operation and maintenance of multi-source energy storage sites face a series of challenges.

[0003] At present, the research on fault warning of multi-source energy storage stations mainly focuses on fault warning of single battery systems, wind turbines or photovoltaic stations. The fault warning models currently used in academia integrate other machine learning algorithms, such as support vector machines (SVM), random forests, etc., as well as statistical analysis methods such as principal component analysis (PCA) and time series analysis, to enhance the model's predictive ability and understanding of the state of complex systems. At the same time, in order to improve the reliability of warnings, researchers are also exploring how to optimize the model's training data set, including data preprocessing, feature selection and engineering, hyperparameter tuning, etc., as well as how to combine domain knowledge and expert experience to explain and verify the model output.

[0004] In recent years, with the advancement of artificial intelligence technology, especially the Long Short-Term Memory Networks (LSTM) in the field of deep learning, due to its excellent time series prediction ability and the ability to capture long-term dependencies, it has begun to be applied to fault warning of energy storage systems, showing significant advantages.

[0005] However, the existing technology analyzes a single or only specific parameters in the electrochemical battery and supercapacitor hybrid energy storage system to diagnose the fault, and fails to conduct overall fault analysis and judgment for more than two multi-source systems including pumped storage, flywheel energy storage, electrochemical battery, flow battery, etc. The fault of the multi-source energy storage system is usually the product of the combined effect of multiple parameters. The existing analysis method often leads to the neglect of key fault precursor information, thus affecting the accuracy of fault warning. Summary of the invention

[0006] The purpose of the present invention is to provide a fault warning method, system and multi-source energy storage station for a multi-source energy storage station, so as to solve the problem that the existing analysis method may cause key fault precursor information to be ignored, thereby affecting the accuracy of fault warning.

[0007] In a first aspect, the present invention provides a fault warning method for a multi-source energy storage station, comprising:

[0008] Obtain operating parameter data of multi-source energy storage stations;

[0009] Inputting the operating parameter data into a fault warning model to obtain a fault prediction value output by the fault warning model;

[0010] Comparing the fault prediction value with a preset warning threshold, and taking the fault prediction value exceeding the preset warning threshold as warning information;

[0011] Among them, the fault warning model is an LSTM model trained based on the fault feature discrimination result, and the fault feature discrimination result is the correlation relationship between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model.

[0012] In a preferred solution, the preset warning threshold is a warning threshold determined based on residual distribution;

[0013] The residual distribution is determined based on the difference between the fault prediction value obtained by inputting the fault feature discrimination result into the fault warning model and the actual value.

[0014] In a preferred embodiment, the residual distribution is represented by a kernel density estimation method.

[0015] In a preferred solution, the fault discrimination classification model is a decision tree model constructed based on the fault characteristics and fault types of the multi-source energy storage station.

[0016] In a preferred solution, the construction of the fault discrimination classification model includes:

[0017] Acquire pre-selected historical fault data of the multi-source energy storage station, wherein the pre-selected historical fault data includes fault characteristic data and fault type data;

[0018] Preprocessing the preselected historical fault data to obtain the historical fault data, wherein the preprocessing includes: cleaning, filling missing values, processing abnormal values, and normalization;

[0019] Taking each of the fault type data in the historical fault data as a root node, a decision tree model is constructed;

[0020] The decision tree model is used as the fault discrimination classification model.

[0021] In a second aspect, the present invention further provides a fault warning system for a multi-source energy storage station, comprising:

[0022] A collection unit, used to obtain operating parameter data of a multi-source energy storage station;

[0023] A prediction unit, used for inputting the operating parameter data into a fault warning model to obtain a fault prediction value output by the fault warning model;

[0024] A comparison unit, configured to compare the fault prediction value with a preset warning threshold, and use the fault prediction value exceeding the preset warning threshold as warning information;

[0025] Among them, the fault warning model is an LSTM model trained based on the fault feature discrimination result, and the fault feature discrimination result is the correlation relationship between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model.

[0026] In a preferred solution, the preset warning threshold is a warning threshold determined based on residual distribution;

[0027] The residual distribution is determined based on the difference between the fault prediction value obtained by inputting the fault feature discrimination result into the fault warning model and the actual value.

[0028] In a preferred embodiment, the residual distribution is represented by a kernel density estimation method.

[0029] In a preferred solution, the fault discrimination classification model is a decision tree model constructed based on the fault characteristics and fault types of the multi-source energy storage station.

[0030] In a third aspect, the present invention provides a multi-source energy storage station, including a multi-source energy storage station body and a fault warning system for the multi-source energy storage station provided in the second aspect of the present invention.

[0031] In order to achieve the above-mentioned purpose, the fault warning method, system and multi-source energy storage station provided by the present invention obtain the operating parameter data of the multi-source energy storage station, and then input the operating parameter data into the fault warning model to obtain the fault prediction value output by the fault warning model, and finally compare the fault prediction value with the preset warning threshold value to use the fault prediction value exceeding the preset warning threshold value as the warning information. Among them, the LSTM model trained based on the fault feature discrimination result is used as the fault warning model to determine the fault prediction value, and the fault feature discrimination result is the correlation relationship between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model, thereby revealing the implicit connection between the parameters from the multi-dimensional parameter distribution, and predicting the fault based on the LSTM model, on the one hand, enhancing the comprehensiveness and sensitivity of the fault warning, realizing the safety monitoring of the multi-source energy storage station, and on the other hand, realizing the accurate prediction of potential faults, reducing unplanned downtime by accurately providing warning signals, thereby improving the overall reliability and economy of the multi-source energy storage station. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flowchart of a fault warning method for a multi-source energy storage station provided in an embodiment of the present invention.

[0033] Figure 2 A flowchart of a method for training an LSTM model.

[0034] Figure 3 A flowchart of a method for training a fault discrimination classification model provided by an embodiment of the present invention.

[0035] Figure 4 A schematic diagram of the structure of an example of a decision tree model provided by an embodiment of the present invention.

[0036] Figure 5 A schematic structural diagram of another example of a decision tree model provided by an embodiment of the present invention.

[0037] Figure 6 A comparison chart of the fault warning accuracy of the fault warning models built using the BP neural network and the LSTM model provided in the embodiment of the present invention.

[0038] Figure 7 A structural diagram of a fault warning system for a multi-source energy storage station provided in an embodiment of the present invention.

[0039] Figure 8 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention is further described in detail below through the accompanying drawings and specific embodiments.

[0041] The fault warning method for a multi-source energy storage station provided by the present invention is executed on an electronic device, which may be an intelligent terminal device such as a laptop, a personal computer, and a tablet computer, or may be a server on the network side.

[0042] like Figure 1 As shown, the fault warning method of a multi-source energy storage station provided by an embodiment of the present invention mainly includes the following steps:

[0043] 110. Obtain operating parameter data of multi-source energy storage stations.

[0044] Specifically, when a multi-source energy storage station fails or is about to fail, the operating parameter data will change. That is, the change in the operating parameter data can reflect the operating status of the multi-source energy storage station.

[0045] More specifically, considering the processing efficiency and computing power, the acquired operating parameter data should be parameters related to the multi-source energy storage station fault. Therefore, based on the fault type of the multi-source energy storage station, the most distinguishing characteristic value in the operating parameter can be determined to analyze the fault characteristics of the multi-source energy storage station.

[0046] In some possible embodiments, the acquired operating parameter data of the multi-source energy storage station include: battery voltage, current, bearing temperature, rotation speed, frame structure status, rotor bearing status, frame structure status, thrust bearing status, current-passing component status, and electric generator status.

[0047] 120. Input the operating parameter data into the fault warning model to obtain the fault prediction value output by the fault warning model.

[0048] Specifically, the change of the operating parameter data can reflect the operating conditions of the multi-source energy storage station. Based on this, by analyzing the regularity of the operating parameter data of the multi-source energy storage station, the failure risk of the multi-source energy storage station can be determined.

[0049] Based on this, in this embodiment, a trained fault warning model is used to obtain a fault prediction value.

[0050] Furthermore, by setting the fault warning model as an LSTM model trained based on the fault feature discrimination result, and the fault feature discrimination result is the correlation between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model. Thus, on the one hand, the LSTM model is used to realize the time series characteristics of the operating data of the multi-source energy storage station, accurately predict potential faults, accurately provide warning signals, reduce unplanned downtime, and improve the overall reliability and economy of the system. On the other hand, by using the correlation between the historical operating parameter data of the multi-source energy storage station and the fault type as the fault feature discrimination result for the training of the LSTM model, the trained fault warning model can capture the complex nonlinear relationship of the data in the fault factors of each energy storage unit in the multi-source energy storage station including pumped storage, flywheel energy storage, electrochemical batteries, etc., which helps to improve the comprehensiveness and accuracy of fault discrimination.

[0051] In some possible embodiments, the Figure 2 The training process shown trains the LSTM model, that is, after inputting the data, first extract the data within the preset time period as a time series, then divide the time series into a training set and a test set according to a preset ratio, and then use the training data in the training set to train the LSTM model. After that, compare and analyze the test data in the test set with the predicted value obtained by LSTM, continuously update the model parameters, and finally realize LSTM modeling.

[0052] Specifically, because the input of each neuron of the LSTM model includes the current moment input and the previous step information, and the current moment input and the previous step information jointly determine the hidden layer output at the current moment. Therefore, the LSTM model can establish a sequence relationship of data, that is, it can predict future information based on historical information. When it is used for fault prediction of multi-source energy storage stations, it can be achieved by monitoring the dynamic changes of the operating parameter data of the multi-source energy storage station in a longer time series to determine the fault prediction value, thereby improving the accuracy of the prediction results.

[0053] It can be understood that a decision tree is a prediction model that uses tree branches to show how data is affected by various variables. It constructs classification rules based on the different effects on the target variable and is generally used in the analysis of customer data. It can be seen that by using the operating parameter data of the multi-source energy storage station as the node of the decision tree, the association between the operating parameter data and the fault type can be obtained.

[0054] Based on this, in an optional embodiment, the fault warning method of the multi-source energy storage station also includes: the fault identification classification model is a decision tree model constructed based on the fault characteristics and fault types of the multi-source energy storage station.

[0055] Furthermore, the construction of the fault identification and classification model includes:

[0056] Obtaining pre-selected historical fault data of a multi-source energy storage station, the pre-selected historical fault data including fault characteristic data and fault type data;

[0057] Preprocess the preselected historical fault data to obtain historical fault data. The preprocessing includes: cleaning, filling missing values, processing outliers, and normalization.

[0058] Taking each fault type data in the historical fault data as the root node, a decision tree model is constructed;

[0059] The decision tree model is used as the fault discrimination classification model.

[0060] Specifically, the decision tree can be used to analyze classification and regression problems. Therefore, energy storage faults can be classified based on the decision tree.

[0061] More specifically, in the training of the fault discrimination classification model, i.e., the decision tree model used to discriminate the faults of the multi-source energy storage station, such as Figure 3 As shown, it is first necessary to determine the pre-selected historical fault data, that is, the historical real data, based on the historical monitoring data set, and then generate a training data set with the fault characteristics obtained through the historical real data as input, and train the decision tree classification model according to the input-output relationship, and finally obtain the discrimination result of the fault characteristics.

[0062] In some optional embodiments, it is assumed that the fault characteristics include: battery voltage v, current i, bearing temperature T b , rotation speed ω, frame structure state T s , rotor bearing condition r b , rack structure status s f , thrust bearing state p t , the state of the current-passing component c c , the motor generator state E, thus obtaining an input vector X=[v,i,T b ,ω,T s ,r b ,s f ,p t ,c c ,E]. Then the trained fault discrimination classification model can be expressed as:

[0063] j(i)=tree c (v(i),i(i),T b (i),ω(i),T s (i),r b (i),s f (i),p t(i),c c (i),E(i)) (1)

[0064] Where J(i) is the fault discrimination result obtained according to the fault discrimination classification model; v(i), i(i), T b (i),ω(i),T s (i),r b (i),s f (i),p t (i),c c (i), E(i) are the actual data of battery voltage, current, bearing temperature, rotation speed, frame structure status, rotor bearing status, frame structure status, thrust bearing status, current-passing component status, and motor generator status at time t.

[0065] In a specific example, the process of building a fault discrimination classification model may include: collecting data on energy storage faults, including symptoms of various faults (input features) and fault types (target categories). According to the fault symptoms, select relevant features, such as: battery voltage, current, bearing temperature, rotation speed, frame structure status, rotor bearing status, frame structure status, thrust bearing status, current-passing component status, electric generator status, etc. The collected data is cleaned, missing values ​​are filled, outliers are processed, and normalized to make the data easier to use. A decision tree model is built using the selected features and target categories. The root node is defined as the entire data set, and for each feature, the information gain ratio of each possible value on the feature is calculated, such as information entropy. The feature with the highest information gain ratio is selected as the feature of the next branch. For each possible value of the selected feature, a new child node is recursively created, and the above steps are repeated until the stop condition is met.

[0066] Specific process example:

[0067] 1. Fault symptom establishment:

[0068] According to the fault analysis of each component of energy storage, the fault is divided into N categories C N (fault type in Table 1 below), from which i fault parameters A are extracted i (The main characteristics are listed in the table below, such as reduced speed, large current value, etc.).

[0069] Take the failure of the flywheel energy storage part as an example:

[0070] Table 1 Typical faults and characteristics of flywheel energy storage units

[0071]

[0072] Then we have:

[0073] A=[A1,A2,...,A i ]=[v,i,T b ,....,ω,T s ,r b ,s f ,p t ,c c ,E] (2)

[0074] A i The value is determined according to the following rules:

[0075]

[0076]

[0077] 2. Build a decision tree model using the selected features and target categories.

[0078] The root node is defined as the entire data set. For each feature, the information gain ratio (e.g., information entropy) of each possible value on the feature is calculated, as shown in J in Table 2 below. C(Ai) :

[0079] Table 2 Information entropy of the main characteristics of typical faults of flywheel energy storage units

[0080] <![CDATA[A1]]> <![CDATA[A2]]> <![CDATA[A3]]> <![CDATA[A4]]> <![CDATA[A5]]> <![CDATA[A6]]> <![CDATA[A7]]> <![CDATA[A8]]> <![CDATA[A9]]> <![CDATA[A 10 ]]> <![CDATA[A 11 ]]> <![CDATA[C1]]> normal normal normal abnormal normal Overrun normal normal abnormal abnormal abnormal <![CDATA[C2]]> normal continued abnormal normal normal decline normal normal abnormal normal normal <![CDATA[C3]]> normal normal abnormal normal normal decline normal normal abnormal normal normal <![CDATA[C4]]> normal fast abnormal normal abnormal decline normal normal abnormal normal normal <![CDATA[C5]]> normal fast abnormal normal normal decline normal normal abnormal abnormal normal <![CDATA[C6]]> normal continued abnormal normal normal decline normal normal normal normal normal … … … … … … … … … … … … <![CDATA[J C(Ai) ]]> 0.56 0.828 0.887 0.559 0.559 0.988 0.559 0.426 0.966 0.828 0.932

[0081] 3. Select the feature with the highest information gain ratio as the feature of the next branch. For each possible value of the selected feature, recursively create a new child node and repeat the above steps until the preset stop condition is met. The resulting decision tree is as follows: Figure 4 and Figure 5 As shown, the fault is specifically predicted and classified according to the fault characteristics (such as speed drop, current is too large, etc.) (the output is a matrix).

[0082] 130. Compare the fault prediction value with a preset warning threshold, and use the fault prediction value exceeding the preset warning threshold as warning information.

[0083] Specifically, by setting a reasonable preset warning threshold, it is possible to accurately determine whether a multi-source energy storage station has a failure risk by comparing the fault prediction value with the preset warning threshold.

[0084] In an optional embodiment, the preset warning threshold is a warning threshold determined based on the residual distribution;

[0085] The residual distribution is determined based on the difference between the fault prediction value and the actual value obtained by inputting the fault feature discrimination result into the fault warning model.

[0086] Further, in an optional embodiment, the residual distribution is represented by a kernel density estimation method.

[0087] Specifically, by utilizing the fault data of multi-source energy storage stations, the kernel density estimation method is used to mine the distribution information from the fault discrimination output data of the fault discrimination classification model, and the preset warning threshold can be estimated.

[0088] The specific implementation method can be as follows:

[0089] Let K(x) represent the kernel function, and n p-dimensional data constitute the sample set Xn×p={Xi, i=1,2,…,n}, where each Xi is a p-dimensional vector, containing the fault classification output of the feature combination of multi-source energy storage voltage, current, temperature, rotor state, generator state, etc., then its probability distribution density can be written as:

[0090]

[0091] Among them, H is the bandwidth parameter matrix of the kernel function.

[0092] Also available:

[0093]

[0094] Where x is a random variable, y is the mean of the distribution center, K(x,y) is the value of the Gaussian kernel function centered on y at x, and σ is the variance.

[0095] It can be seen from formula (3) that the kernel function and bandwidth can determine the probability distribution density, so the selection of this parameter is crucial.

[0096] Furthermore, the accuracy of kernel density estimation is mainly related to the bandwidth parameter matrix. For the bandwidth h, it plays a role in local smoothing of f(x). Since the multi-source energy storage station collects more parameters, the bandwidth matrix H=h with fewer parameters is selected. 2 I, which is expressed as follows:

[0097]

[0098] Where K(x) is the bandwidth h j The kernel density function, n is the number of data samples. From formula (4), we can see that for each point j, there is a corresponding bandwidth h j , that is, the bandwidth is adaptively variable. α is the sensitivity factor, and its value range is 0≤α≤1. It should be noted that, usually when α is 0.5 or 0, the adaptive bandwidth kernel density estimation becomes a fixed bandwidth kernel density estimation.

[0099] The following simulation case specifically illustrates the accuracy of fault warning using the fault warning method of the multi-source energy storage station provided by the above embodiment of the present invention:

[0100] First, the historical operating parameter data of the multi-source energy storage station is collected, including: battery voltage, current, bearing temperature, rotation speed, frame structure status, rotor bearing status, frame structure status, thrust bearing status, current component status, and actual status of the electric generator. First, a fault discrimination classification model based on a decision tree is trained, and the training results are shown in Table 3.

[0101] Table 3 Training results of fault discrimination classification model based on decision tree

[0102]

[0103] As can be seen from Table 3, as the number of samples increases, the training time also increases, but the training can still be completed within ten seconds, and when the number of samples reaches 20,000, the accuracy can reach more than 99%.

[0104] Then the fault warning models built using BP neural network (method 1) and LSTM model (method 2) are compared. The fault warning accuracy of the two is compared. Figure 6 shown.

[0105] At the same time, because the mean average percentage error (MAPE) shown in formula (5) is usually used to evaluate the accuracy of time series prediction models, especially when it comes to small values ​​in time series data, the closer the MAPE value is to 0, the more accurate the model prediction is. When its value is greater than or equal to 1, the model prediction effect will be very poor.

[0106]

[0107] where y(k) is the actual value of the i-th observation, is the predicted value for the ith observation, and n is the total number of observations.

[0108] The root mean square error (RMSE) shown in formula (6) is used as an indicator to measure the overall accuracy of the model. The smaller the RMSE, the closer the model and the predicted results are to the true value, and the better the effect.

[0109]

[0110] where y(k) is the actual value of the i-th observation, is the mean of the ith observation, and n is the total number of observations. 2 The coefficient of determination indicates the percentage of the variation of the outcome variable that the predictor variable can explain. It ranges from 0 to 1. The closer the value is to 1, the better the fit of the model, as shown in formula (7).

[0111]

[0112] Therefore, the MAPE, RMSE, and R of the fault warning model built by method 1 and method 2 are calculated based on formulas (5) to (7) respectively. 2 , the results are shown in Table 4 below.

[0113] Table 4 Fault errors of each model

[0114]

[0115] By Figure 6 It can be seen from Table 4 that compared with the fault warning model built with BP neural network, the three effective evaluation indicators of the fault warning model built with LSTM are better than those of the fault warning model built with BP neural network.

[0116] In summary, the fault warning method for a multi-source energy storage station provided by the above-mentioned embodiment of the present invention, on the one hand, for a multi-source energy storage station including pumped storage, flywheel energy storage, electrochemical battery, flow battery, etc., by fully considering the fault factors of each energy storage unit in the multi-source energy storage station, capturing the complex nonlinear relationship of the data in the fault factors, and constructing a decision tree model (fault discrimination classification model) that simulates the human decision-making process, realizes fault discrimination, thereby helping to improve the comprehensiveness and accuracy of fault discrimination. On the other hand, by introducing the kernel density estimation (KDE) technology into the calculation of the fault warning threshold of the multi-source energy storage station. Compared with a single or hybrid energy storage system for electrochemical batteries and supercapacitors, the dynamic warning threshold based on KDE can be adjusted according to the real-time changes in the operating status of the multi-source energy storage station, and can adapt to different operating conditions more flexibly, thereby improving the sensitivity and accuracy of the fault warning. On the other hand, based on the time series characteristics of the operating parameter data of multi-source energy storage stations, fault warning is performed by applying the LSTM model (fault warning model), thereby achieving accurate prediction of potential faults and accurately providing early warning signals, thereby reducing unplanned downtime and improving the overall reliability and economy of the station.

[0117] In summary, the fault warning method for a multi-source energy storage station provided in the embodiment of the present invention provides a complete solution for intelligent operation and maintenance and fault warning of energy storage stations, has significant practical value and broad application prospects, and provides strong technical support for the efficient and stable operation of new energy power systems.

[0118] A fault warning system for a multi-source energy storage station provided in an embodiment of the present invention is introduced below. The fault warning system for a multi-source energy storage station described below can be considered as a module architecture for implementing a fault warning method for a multi-source energy storage station provided in an embodiment of the present invention; the content described below can be cross-referenced with the above.

[0119] Optional, see Figure 7 , Figure 7 This is a structural block diagram of a fault warning system for a multi-source energy storage station provided by an embodiment of the present invention. The system may include:

[0120] A collection unit 10 is used to obtain operating parameter data of a multi-source energy storage station;

[0121] The prediction unit 20 is used to input the operating parameter data into the fault warning model to obtain the fault prediction value output by the fault warning model;

[0122] A comparison unit 30, used to compare the fault prediction value with a preset warning threshold, and use the fault prediction value exceeding the preset warning threshold as warning information;

[0123] Among them, the fault warning model is an LSTM model trained based on the fault feature discrimination results. The fault feature discrimination results are the correlation between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model.

[0124] Optionally, the preset warning threshold is a warning threshold determined based on residual distribution;

[0125] The residual distribution is determined based on the difference between the fault prediction value and the actual value obtained by inputting the fault feature discrimination result into the fault warning model.

[0126] Optionally, the residual distribution is represented by kernel density estimation.

[0127] Optionally, the fault identification and classification model is a decision tree model constructed based on the fault characteristics and fault types of the multi-source energy storage station.

[0128] Optional, construction of a fault discrimination classification model, including:

[0129] Obtaining pre-selected historical fault data of a multi-source energy storage station, the pre-selected historical fault data including fault characteristic data and fault type data;

[0130] Preprocess the preselected historical fault data to obtain historical fault data. The preprocessing includes: cleaning, filling missing values, processing outliers, and normalization.

[0131] Taking each fault type data in the historical fault data as the root node, a decision tree model is constructed;

[0132] The decision tree model is used as the fault discrimination classification model.

[0133] Optionally, an embodiment of the present invention further provides a multi-source energy storage station, which includes a multi-source energy storage station body and a fault warning system for the multi-source energy storage station provided in any of the above embodiments.

[0134] Below, reference Figure 8 To describe the electronic device provided by the embodiment of the present application, the electronic device provided by the embodiment may include: at least one processor 100, at least one communication interface 200, at least one memory 300 and at least one communication bus 400;

[0135] In the embodiment of the present invention, the number of the processor 100, the communication interface 200, the memory 300, and the communication bus 400 is at least one, and the processor 100, the communication interface 200, and the memory 300 communicate with each other through the communication bus 400; obviously, Figure 8 The communication connections shown for the processor 100, the communication interface 200, the memory 300, and the communication bus 400 are merely optional;

[0136] Optionally, the communication interface 200 may be an interface of a communication module, such as an interface of a GSM module; the processor 100 may be a central processing unit CPU, or an application specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0137] The memory 300 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0138] The processor 100 is specifically used to execute the application program in the memory to implement the steps of the above-mentioned crane hoisting wire rope installation control method.

[0139] The above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A fault warning method for a multi-source energy storage station, characterized in that: include: Obtain operating parameter data of multi-source energy storage stations; Inputting the operating parameter data into a fault warning model to obtain a fault prediction value output by the fault warning model; Comparing the fault prediction value with a preset warning threshold, and taking the fault prediction value exceeding the preset warning threshold as warning information; Among them, the fault warning model is an LSTM model trained based on the fault feature discrimination result, and the fault feature discrimination result is the correlation relationship between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model.

2. The fault warning method for a multi-source energy storage station according to claim 1, characterized in that: The preset warning threshold is a warning threshold determined based on the residual distribution; The residual distribution is determined based on the difference between the fault prediction value obtained by inputting the fault feature discrimination result into the fault warning model and the actual value.

3. The fault warning method for a multi-source energy storage station according to claim 2, characterized in that: The residual distribution is represented by kernel density estimation.

4. The fault warning method for a multi-source energy storage station according to claim 1, characterized in that: The fault discrimination classification model is a decision tree model constructed based on the fault characteristics and fault types of the multi-source energy storage station.

5. The fault warning method for a multi-source energy storage station according to claim 4 is characterized in that: The construction of the fault discrimination classification model includes: Acquire pre-selected historical fault data of the multi-source energy storage station, wherein the pre-selected historical fault data includes fault characteristic data and fault type data; Preprocessing the preselected historical fault data to obtain the historical fault data, wherein the preprocessing includes: cleaning, filling missing values, processing abnormal values, and normalization; Taking each of the fault type data in the historical fault data as a root node, a decision tree model is constructed; The decision tree model is used as the fault discrimination classification model.

6. A fault warning system for a multi-source energy storage station, characterized in that: include: A collection unit, used to obtain operating parameter data of a multi-source energy storage station; A prediction unit, used for inputting the operating parameter data into a fault warning model to obtain a fault prediction value output by the fault warning model; A comparison unit, configured to compare the fault prediction value with a preset warning threshold, and use the fault prediction value exceeding the preset warning threshold as warning information; Among them, the fault warning model is an LSTM model trained based on the fault feature discrimination result, and the fault feature discrimination result is the correlation relationship between the historical operating parameter data of the multi-source energy storage station and the fault type obtained by inputting the historical fault data of the multi-source energy storage station into the fault discrimination classification model.

7. The fault warning system for a multi-source energy storage station according to claim 6, characterized in that: The preset warning threshold is a warning threshold determined based on the residual distribution; The residual distribution is determined based on the difference between the fault prediction value obtained by inputting the fault feature discrimination result into the fault warning model and the actual value.

8. The fault warning system for a multi-source energy storage station according to claim 7, characterized in that: The residual distribution is represented by kernel density estimation.

9. The fault warning method for a multi-source energy storage station according to claim 6, characterized in that: The fault discrimination classification model is a decision tree model constructed based on the fault characteristics and fault types of the multi-source energy storage station.

10. A multi-source energy storage station, characterized in that: It comprises a multi-source energy storage station body and a fault warning system for the multi-source energy storage station as claimed in any one of claims 6 to 9.

Images