Power plant operation status monitoring system and method based on multivariate time series prediction

Abstract

The present invention belongs to the technical field of status monitoring, and discloses a power plant operation status monitoring system and method based on multivariate time series prediction; m groups of historical power plant characteristic data are collected, and the historical power plant characteristic data include power plant data, power grid data and environmental data; according to the historical power plant characteristic data, corresponding historical characteristic coefficients are calculated, and corresponding historical characteristic coefficient time series are constructed; according to the historical characteristic coefficient time series, a coefficient prediction model for predicting the historical characteristic coefficients at future time points is trained; the historical characteristic coefficient time series is input into the coefficient prediction model to obtain the characteristic coefficients at future time points, and marked as monitoring data; according to the monitoring data, it is determined whether to generate a fault instruction; according to the fault instruction, adjustments are made, and it is determined whether to generate an early warning instruction; the stability and reliability of power plant operation are improved, and the monitoring and management level of the power plant operation status is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of condition monitoring, and more specifically, to a power plant operation condition monitoring system and method based on multivariate time series prediction. Background Art

[0002] With the development of society and the acceleration of industrialization, electricity, as a key energy resource, plays a vital role in maintaining the normal operation of modern society. As the main production unit of electricity, the stability and reliability of the operation status of the power station are crucial to the operation of the entire power system. Therefore, in order to ensure the efficient operation of the power station and timely detect potential faults, the power station operation status monitoring system has gradually become a research direction that has attracted much attention in the power field.

[0003] Traditional power station monitoring methods mainly rely on discrete manual inspections and regular maintenance. This method has the disadvantages of long detection cycles and difficulty in discovering potential problems. Of course, there are also intelligent detection methods. For example, the patent with publication number CN116961226A discloses a monitoring method and system for transformers in substations, including obtaining the operating data of each transformer in the substation, evaluating the operating status of the transformer based on the substation status evaluation system, and obtaining the operating status evaluation results of each transformer; using a time series model to predict the monitoring data of each transformer in the substation, and obtaining the predicted value of the monitoring data of each transformer; allowing operation and maintenance personnel to judge the operating status of the transformer in advance and make Take corresponding measures to maintain the safe and stable operation of the transformer; for example, the patent with publication number CN116539994A discloses a method for detecting the operating status of the main equipment of a substation based on multi-source time series data, including obtaining multiple time series characteristics that are highly correlated with the abnormal operation of the substation equipment as target indicators; inputting multiple historical target indicators into the prediction model in sequence according to the preset time step to obtain the target indicators of multiple future time series of the substation operation status; using the density-based noise application spatial clustering algorithm, clustering the target indicators of the current and future time series of the substation equipment operation status to filter out abnormal data, and warning that the equipment operation status may be abnormal in the future;

[0004] However, the above technology uses multiple data as the output of an LSTM model, which increases the difficulty and computational cost of model training, thereby reducing the prediction effect. In addition, the monitoring of the power plant operation status is only achieved by directly comparing the predicted data with the threshold to determine whether it is abnormal, which has certain limitations. At the same time, due to various random events in the operation of the power plant, misdiagnosis may occur.

[0005] In view of this, the present invention proposes a power plant operation status monitoring system and method based on multivariate time series prediction to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a power plant operation status monitoring method based on multivariate time series prediction, comprising:

[0007] Collect m groups of historical power plant characteristic data, the historical power plant characteristic data including power plant data, power grid data and environmental data, where m is an integer greater than 1;

[0008] According to the historical power plant characteristic data, the corresponding historical characteristic coefficients are calculated, and the corresponding historical characteristic coefficient time series are constructed; the historical characteristic coefficients include power plant coefficients, power grid coefficients and environmental coefficients; the historical characteristic coefficient time series includes power plant coefficient time series, power grid coefficient time series and environmental coefficient time series;

[0009] Based on the power plant coefficient time series, a first coefficient prediction model for predicting the power plant coefficient at a future time point is trained; based on the power grid coefficient time series, a second coefficient prediction model for predicting the power grid coefficient at a future time point is trained; based on the environmental coefficient time series, a third coefficient prediction model for predicting the environmental coefficient at a future time point is trained;

[0010] Input the historical characteristic coefficient time series into the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model respectively, obtain the power station coefficient, the power grid coefficient and the environmental coefficient at the future time point, and mark them as monitoring data; determine whether to generate a fault instruction according to the monitoring data;

[0011] The power plant data corresponding to the power plant coefficient at a future time point is adjusted according to the fault instruction, and it is determined whether to generate an early warning instruction.

[0012] Further, the m groups of historical power station characteristic data correspond to m time points; the power station data include speed, power, temperature and electromagnetic radiation; the electromagnetic radiation includes electromagnetic radiation frequency and electromagnetic radiation intensity; the speed is the speed of the engine in the power station, the power is the output power of the engine in the power station, the temperature is the temperature of the engine in the power station, and the electromagnetic radiation is the electromagnetic radiation of the engine in the power station; the power grid data includes power grid voltage, power grid current and power grid frequency;

[0013] The environmental data includes first environmental data and second environmental data. The first environmental data includes a first environmental temperature and a first environmental humidity. The second environmental data includes a second environmental temperature and a second environmental humidity.

[0014] Furthermore, the calculation method of the power station coefficient includes:

[0015]

[0016] Where DZX i is the power station coefficient, ZSi is the speed, GL i is power, DW i is the temperature, DP i is the electromagnetic radiation frequency, DQ i is the electromagnetic radiation intensity, ω1, ω2, ω3 are preset weights, i is the i-th time point, i∈[1,m], m is the number of time points;

[0017] The calculation method of the power grid coefficient includes:

[0018]

[0019] Where DWX i is the power grid coefficient, DY i Grid voltage, DL i Grid current, PL i Grid frequency, ω4 and ω5 are preset weights;

[0020] The calculation method of the environmental factor includes:

[0021]

[0022] In the formula, HJXi is the environmental coefficient, is the first ambient temperature, is the first ambient humidity, is the second ambient temperature, is the second ambient humidity, ω6 and ω7 are preset weights, and A is a preset constant.

[0023] Furthermore, the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model are all LSTM models in the recurrent neural network model, and the training process is consistent;

[0024] The training process of the first coefficient prediction model includes:

[0025] Define the architecture of the coefficient prediction model, including input state, hidden state, number of output units, and hyperparameters; use the deep learning framework to create the coefficient prediction model network layer, which includes an input layer, multiple LSTM layers, and an output layer, where each LSTM layer consists of an input gate, a forget gate, an output gate, and a cell state; connect the created LSTM layers in sequence to form a complete coefficient prediction model, ensure that the input dimension matches the feature dimension of the data, and set the number of units in the output layer to accommodate the required prediction results;

[0026] The power plant coefficient time series is used as the input of the first coefficient prediction model. The first coefficient prediction model uses the power plant coefficient at a future time point as the output, and the actual power plant coefficient is used as the prediction target. The first coefficient prediction model is trained, and the prediction result is evaluated for model accuracy using the mean absolute percentage error MAPE. When the calculated MAPE is less than the preset MAPE, the training of the first coefficient prediction model is completed; wherein, the calculation formula of MAPE is: Where K is the Kth time point, α K is the actual power station coefficient corresponding to the Kth time point, μ K is the predicted power plant coefficient corresponding to the Kth time point, and R is the number of predicted power plant coefficients.

[0027] Further, the fault instruction includes a preliminary fault instruction, a first fault instruction, a second fault instruction and a third fault instruction; the preliminary fault instruction includes a preliminary first fault instruction, a preliminary second fault instruction and a preliminary third fault instruction;

[0028] The method for determining whether to generate a first fault instruction and a preliminary fault instruction includes:

[0029] According to m power station coefficients, calculate the corresponding power station coefficient mean and power station coefficient mean square error;

[0030] The calculation process of the mean power plant coefficient includes:

[0031]

[0032] In the formula, JZ is the mean value of the power station coefficient, and m is the number of power station coefficients;

[0033] The calculation process of the mean square error of the power station coefficient includes:

[0034]

[0035] Where, JFC is the mean square error of power plant coefficients, and m is the number of power plant coefficients;

[0036] Preset constants Z and C, C>Z; subtract the power station coefficient mean from the predicted future time point power station coefficient to obtain the power station coefficient difference, multiply the power station coefficient mean square error by Z to obtain the first product, multiply the power station coefficient mean square error by C to obtain the second product, and compare the absolute value of the power station coefficient difference with the first product and the second product respectively;

[0037] If the absolute value of the power station coefficient difference is less than or equal to the first product, no preliminary fault instruction is generated;

[0038] If the absolute value of the power station coefficient difference is greater than the first product and less than or equal to the second product, a preliminary first fault instruction and a preliminary second fault instruction are generated;

[0039] If the absolute value of the power station coefficient difference is greater than the second product, a preliminary first fault instruction and a preliminary third fault instruction are generated;

[0040] If a preliminary first fault instruction is generated, the time point corresponding to the preliminary first fault instruction is marked as a suspected fault point, and a set of suspected fault points is set. If the time points corresponding to multiple suspected fault points are continuous time points, the multiple continuous suspected fault points are added to the set of suspected fault points in the order of time points. If the times corresponding to multiple suspected fault points are not continuous time points, the suspected fault points are not added to the set of suspected fault points. The time point corresponding to the last suspected fault point in the set of suspected fault points is subtracted from the time point corresponding to the first suspected fault point to obtain the time point difference, and the time point difference is marked as the duration.

[0041] comparing the duration to a preset duration threshold;

[0042] If the duration is greater than or equal to the duration threshold, generating a first fault instruction;

[0043] If the duration is less than the duration threshold, no first fault instruction is generated;

[0044] If the first fault instruction is generated, the set of suspected fault points is cleared.

[0045] Further, the method for determining whether to generate the second fault instruction includes:

[0046] If a preliminary second fault instruction is generated, the predicted environmental coefficient at a future time point is subtracted from the environmental coefficient corresponding to the time point before the time point to obtain the difference in environmental coefficients;

[0047] Compare the absolute value of the environmental coefficient difference with a preset environmental difference threshold;

[0048] If the environmental coefficient difference is less than or equal to the environmental difference threshold, the second fault instruction is not generated;

[0049] If the environmental coefficient difference is greater than the environmental difference threshold, a second fault instruction is generated.

[0050] Further, the method for determining whether to generate the third fault instruction includes:

[0051] If a preliminary third fault instruction is generated, the grid coefficient difference is obtained by subtracting the grid coefficient corresponding to the previous time point from the predicted future time point;

[0052] Comparing the absolute value of the grid coefficient difference with a preset grid difference threshold;

[0053] If the grid coefficient difference is less than or equal to the grid difference threshold, the third fault instruction is not generated;

[0054] If the grid coefficient difference is greater than the grid difference threshold, a third fault instruction is generated.

[0055] Furthermore, the method for adjusting the power plant data corresponding to the power plant coefficient at a future time point includes:

[0056] If the first fault instruction is generated, the rotation speed and power in the power station data corresponding to the future time point are adjusted;

[0057] The first product plus the mean of the power station coefficient is taken as the maximum power station coefficient, and the difference obtained by subtracting the maximum power station coefficient from the power station coefficient at the future time point is marked as the power station coefficient adjustment amount;

[0058] The power station coefficient adjustment amount is used as the input of the fuzzy controller, the speed adjustment amount and the power adjustment amount are used as the output of the fuzzy controller, and the range of the speed adjustment amount and the power adjustment amount is defined according to the speed range value and the power range value;

[0059] According to the relationship among the speed regulation, power regulation and power station coefficient regulation, the fuzzy control rule is designed, and the fuzzy control rule is expressed in the form of IF-THEN; the relationship among the speed regulation, power regulation and power station coefficient regulation is obtained according to the calculation formula of the power station coefficient;

[0060] The speed regulation variable is converted into a fuzzy set through fuzzification technology;

[0061] According to the fuzzified speed regulation value and fuzzy control rules, the fuzzy output is obtained;

[0062] The fuzzy output is converted into speed regulation quantity and power regulation quantity by using defuzzification technology;

[0063] The speed and power in the power station data corresponding to the future time point are adjusted according to the speed adjustment amount and the power adjustment amount.

[0064] Further, the method for determining whether to generate an early warning instruction includes:

[0065] If the second fault instruction or the third fault instruction is generated, a warning instruction is generated.

[0066] A power plant operation status monitoring system based on multivariate time series prediction, implementing the power plant operation status monitoring method based on multivariate time series prediction, comprising:

[0067] A data collection module collects m groups of historical power plant characteristic data, the historical power plant characteristic data includes power plant data, power grid data and environmental data;

[0068] The data processing module calculates the corresponding historical characteristic coefficients according to the historical power plant characteristic data, and constructs the corresponding historical characteristic coefficient time series; the historical characteristic coefficients include power plant coefficients, power grid coefficients and environmental coefficients; the historical characteristic coefficient time series includes power plant coefficient time series, power grid coefficient time series and environmental coefficient time series;

[0069] The model training module trains a first coefficient prediction model for predicting the power plant coefficient at a future time point based on the power plant coefficient time series; trains a second coefficient prediction model for predicting the power grid coefficient at a future time point based on the power grid coefficient time series; and trains a third coefficient prediction model for predicting the environmental coefficient at a future time point based on the environmental coefficient time series;

[0070] The state monitoring module inputs the historical characteristic coefficient time series into the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model respectively, obtains the power station coefficient, the power grid coefficient and the environmental coefficient at the future time point, and marks them as monitoring data; determines whether to generate a fault instruction according to the monitoring data;

[0071] The adjustment module adjusts the power plant data corresponding to the power plant coefficient at a future time point according to the fault instruction, and determines whether to generate an early warning instruction.

[0072] The technical effects and advantages of the power plant operation status monitoring system and method based on multivariate time series prediction of the present invention are as follows:

[0073] By collecting and analyzing historical power plant characteristic data, the historical characteristic coefficient time series of the power plant coefficient, grid coefficient and environmental coefficient are obtained, and multiple accurate prediction models are trained based on these sequences, so that the power plant, grid and environmental coefficients at future time points can be predicted; these prediction models are used for monitoring data. Once an abnormal situation is detected, the monitoring system can generate fault instructions and early warning instructions in a timely manner, prompting operation and maintenance personnel to make adjustments and maintenance, thereby improving the stability and reliability of power plant operation, reducing potential failure risks, and improving the monitoring and management level of power plant operation status. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 This is a schematic diagram of a power plant operation status monitoring system based on multivariate time series prediction according to Embodiment 1 of the present invention;

[0075] Figure 2 This is a schematic diagram of the edge division of a power station according to Embodiment 1 of the present invention;

[0076] Figure 3 This is a schematic diagram of a power plant operation status monitoring method based on multivariate time series prediction according to Embodiment 2 of the present invention;

[0077] Figure 4This is a schematic diagram of the structure of an electronic device according to Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0078] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0079] Example 1

[0080] See also Figure 1 As shown, the power plant operation status monitoring system based on multivariate time series prediction described in this embodiment includes a data acquisition module, a data processing module, a model training module, a status monitoring module and an adjustment module; each module is connected by wired and / or wireless means to realize data transmission between modules;

[0081] A data collection module collects m groups of historical power station characteristic data, where m is an integer greater than 1;

[0082] The historical power plant characteristic data includes power plant data, power grid data and environmental data; a collection interval is preset, and m groups of historical power plant characteristic data are collected according to the collection interval, that is, the m groups of historical power plant characteristic data correspond to m time points;

[0083] Power station data includes speed, power, temperature and electromagnetic radiation; electromagnetic radiation includes electromagnetic radiation frequency and electromagnetic radiation intensity;

[0084] The speed is the speed of the engine in the power station, which is obtained by the speed sensor installed at the engine rotor bearing; the abnormal engine speed indicates that the mechanical parts of the engine are faulty, which indicates that the power station is operating abnormally; the power is the output power of the engine in the power station, which is obtained by the power sensor installed on the generator output cable; the temperature is the temperature of the engine in the power station, which is obtained by the temperature sensor installed in the engine; the greater the degree of abnormality of the engine power and the engine temperature, the greater the probability of engine failure, which indicates that the power station is operating abnormally; the electromagnetic radiation is the electromagnetic radiation of the engine in the power station, and the electromagnetic radiation frequency and electromagnetic radiation intensity are obtained by the electromagnetic radiation sensor installed inside the engine cover; the deviation of the electromagnetic radiation frequency and electromagnetic radiation intensity from the normal value range indicates that the engine is operating abnormally, which indicates that the power station is operating abnormally;

[0085] Grid data includes grid voltage, grid current and grid frequency; the grid voltage, grid current and grid frequency are all obtained by the grid monitoring system. Significant fluctuations in grid voltage, i.e., excessively high grid voltage indicates equipment failure or improper regulation and control, and excessively low grid voltage indicates excessive load or grid failure, both of which will affect the operation status of the power station and cause abnormal operation of the power station; continuous fluctuations in grid frequency indicate that the power system is unstable, thus causing unstable operation of the power station;

[0086] Environmental data includes first environmental data and second environmental data. The first environmental data includes first environmental temperature and first environmental humidity, and the second environmental data includes second environmental temperature and second environmental humidity. The first environmental temperature and second environmental temperature are both external environmental temperature, and the first environmental humidity and second environmental humidity are both external environmental humidity. The external environment is the environment of the area where the power station is located. The first environmental data is obtained through the meteorological station, which will announce the meteorological conditions of the area where the power station is located. The meteorological conditions include the first environmental temperature and the first environmental humidity. The second environmental data is obtained through sensors. The edge of the power station is divided into Q sections according to the sensor collection range, and a temperature sensor and a humidity sensor are installed in each section. The second environmental temperature is obtained by the temperature sensor installed at the edge of the power station, and the second environmental humidity is obtained by the humidity sensor installed at the edge of the power station. For details, please refer to Figure 2 As shown; since a power station is usually a large facility, multiple temperature sensors and humidity sensors are required; the temperatures collected by Q temperature sensors are collected, and the average temperature is used as the second ambient temperature; the humidity collected by Q humidity sensors is collected, and the average humidity is used as the second ambient humidity; the environmental data will affect the operation of the power equipment in the power station, thereby affecting the operation status of the power station;

[0087] It should be understood that the reason for collecting the first environmental data and the second environmental data is that the first environmental data is only the environmental data of the area where the power station is located, and the second environmental data is only the environmental data around the power station. The environmental data is affected during the operation of the power station. If only one of the environmental data is considered, it is impossible to fully express the impact of the environmental data on the operating status of the power station. Therefore, it is necessary to collect two types of environmental data;

[0088] It should be noted that the collection interval is pre-set by the staff according to the historical power plant operation status, and the historical power plant characteristic data is collected according to the appropriate collection interval;

[0089] The data processing module calculates the corresponding historical characteristic coefficients according to the historical power station characteristic data, and constructs the corresponding historical characteristic coefficient time series;

[0090] The historical characteristic coefficients include power plant coefficients, power grid coefficients and environmental coefficients; the historical characteristic coefficient time series include power plant coefficient time series, power grid coefficient time series and environmental coefficient time series; wherein, the power plant coefficient corresponds to the power plant coefficient time series, the power grid coefficient corresponds to the power grid coefficient time series, and the environmental coefficient corresponds to the environmental coefficient time series;

[0091] The calculation methods of power station coefficient include:

[0092]

[0093] Where DZX i is the power station coefficient, ZS i is the speed, GL i is power, DW i is the temperature, DP i is the electromagnetic radiation frequency, DQ i is the electromagnetic radiation intensity, ω1, ω2, ω3 are preset weights, i is the i-th time point, i∈[1,m], m is the number of time points;

[0094] The preset weights in the formula are collected by a technician in this field, and corresponding weights are set for each set of comprehensive parameters. The preset weights and the collected comprehensive parameters are substituted into the formula. Any three formulas constitute a three-variable linear equation system. The calculated weights are screened and averaged to obtain the values ​​of ω1, ω2, and ω3.

[0095] The calculation method of the grid coefficient includes:

[0096]

[0097] Where DWX i is the power grid coefficient, DY i Grid voltage, DL i Grid current, PL i Grid frequency, ω4 and ω5 are preset weights;

[0098] The preset weights in the formula are collected by a technician in this field, and corresponding weights are set for each set of comprehensive parameters. The preset weights and the collected comprehensive parameters are substituted into the formula. Any two formulas constitute a set of binary linear equations. The calculated weights are screened and averaged to obtain the values ​​of ω4 and ω5.

[0099] The calculation method of environmental factor includes:

[0100]

[0101] In the formula, HJXi is the environmental coefficient, is the first ambient temperature, is the first ambient humidity, is the second ambient temperature, is the second ambient humidity, ω6 and ω7 are preset weights, and A is a preset constant;

[0102] The preset weights and preset constants in the formula are collected by a technician in this field. Multiple groups of comprehensive parameters are collected, and corresponding weights and constants are set for each group of comprehensive parameters. The preset weights and constants and the collected comprehensive parameters are substituted into the formula. Any three formulas constitute a three-variable linear equation group. The calculated weights are screened and averaged to obtain the values ​​of ω6, ω7, and A.

[0103] According to the historical power plant characteristic data, the corresponding m power plant coefficients, m power grid coefficients and m environmental coefficients are calculated; since the m groups of historical power plant characteristic data correspond to m time points, the m power plant coefficients, m power grid coefficients and m environmental coefficients also correspond to the m time points one by one, and the corresponding power plant coefficient time series, power grid coefficient time series and environmental coefficient time series are constructed according to the m time points;

[0104] It should be noted that the above calculations of historical characteristic coefficients are all dimensionless calculations because the historical characteristic coefficients are only used to detect the operating status of the power station, and there is no need to consider the relationship between units;

[0105] The model training module trains a first coefficient prediction model for predicting the power plant coefficient at a future time point based on the power plant coefficient time series; trains a second coefficient prediction model for predicting the power grid coefficient at a future time point based on the power grid coefficient time series; and trains a third coefficient prediction model for predicting the environmental coefficient at a future time point based on the environmental coefficient time series;

[0106] The first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model are all LSTM models in the recurrent neural network model, and the training process is the same; therefore, the training process of the first coefficient prediction model is specifically described, and the training process of the second coefficient prediction model and the third coefficient prediction model is the same as that of the first coefficient prediction model, and no further description is given here;

[0107] The training process of the first coefficient prediction model includes:

[0108] Define the architecture of the coefficient prediction model, including input state, hidden state, number of output units, and hyperparameters, such as learning rate, number of iterations, etc.; Use a deep learning framework (such as TensorFlow, PyTorch, etc.) to create a coefficient prediction model network layer, which includes an input layer, multiple LSTM layers, and an output layer, where each LSTM layer consists of an input gate, a forget gate, an output gate, and a cell state; Connect the created LSTM layers in sequence to form a complete coefficient prediction model, ensure that the input dimension matches the feature dimension of the data, and set the number of units in the output layer to accommodate the required prediction results;

[0109] The power plant coefficient time series is used as the input of the first coefficient prediction model. The first coefficient prediction model uses the power plant coefficient at a future time point as the output, and the actual power plant coefficient is used as the prediction target. The first coefficient prediction model is trained, and the prediction result is evaluated for model accuracy using the mean absolute percentage error MAPE. When the calculated MAPE is less than the preset MAPE, the training of the first coefficient prediction model is completed; wherein, the calculation formula of MAPE is: Where K is the Kth time point, α K is the actual power station coefficient corresponding to the Kth time point, μ K is the predicted power plant coefficient corresponding to the Kth time point, and R is the number of predicted power plant coefficients;

[0110] Exemplarily, the power plant coefficient time series is constructed from the power plant coefficients from 1 to 10 o'clock today, the power plant coefficient time series is input into the first coefficient prediction model, the power plant coefficient at 11 o'clock is predicted, the power plant coefficient time series is constructed from the power plant coefficients from 2 to 11 o'clock, the power plant coefficient at 12 o'clock is predicted, and so on to obtain the power plant coefficients at all time points of the day;

[0111] It should be noted that the preset MAPE is pre-set by the staff according to the accuracy required by the model;

[0112] The state monitoring module inputs the historical characteristic coefficient time series into the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model respectively, obtains the power station coefficient, the power grid coefficient and the environmental coefficient at the future time point, and marks them as monitoring data; determines whether to generate a fault instruction according to the monitoring data;

[0113] The fault instructions include a preliminary fault instruction, a first fault instruction, a second fault instruction and a third fault instruction; the preliminary fault instructions include a preliminary first fault instruction, a preliminary second fault instruction and a preliminary third fault instruction;

[0114] The method for determining whether to generate a first fault instruction and a preliminary fault instruction includes:

[0115] According to m power station coefficients, calculate the corresponding power station coefficient mean and power station coefficient mean square error;

[0116] The calculation process of the mean power plant coefficient includes:

[0117]

[0118] In the formula, JZ is the mean value of the power station coefficient, and m is the number of power station coefficients;

[0119] The calculation process of the mean square error of the power station coefficient includes:

[0120]

[0121] Where, JFC is the mean square error of power plant coefficients, and m is the number of power plant coefficients;

[0122] Preset constants Z and C, C>Z; subtract the power station coefficient mean from the predicted future time point power station coefficient to obtain the power station coefficient difference, multiply the power station coefficient mean square error by Z to obtain the first product, multiply the power station coefficient mean square error by C to obtain the second product, and compare the absolute value of the power station coefficient difference with the first product and the second product respectively;

[0123] If the absolute value of the power station coefficient difference is less than or equal to the first product, no preliminary fault instruction is generated; this indicates that the power station coefficient is normal and the power station is in normal operation;

[0124] If the absolute value of the power station coefficient difference is greater than the first product and less than or equal to the second product, a preliminary first fault instruction and a preliminary second fault instruction are generated; indicating that the power station coefficient is abnormal, and thus indicating that the power station data is abnormal and needs further monitoring;

[0125] If the absolute value of the power station coefficient difference is greater than the second product, a preliminary first fault instruction and a preliminary third fault instruction are generated; this indicates that the power station coefficient is more abnormal than when the second fault instruction was generated, and further monitoring is required;

[0126] It should be noted that the preset constants Z and C are obtained by collecting power station data when multiple engines fail, power station coefficients when engines fail and environmental data are abnormal, and power station data when engines are operating normally during the historical power station operation process, and calculating the corresponding power station coefficients. The mean and mean square error of the power station coefficients corresponding to the normal operation of multiple power stations are calculated, the mean of the power station coefficients corresponding to the multiple engine failures is calculated, marked as the first failure mean, and the mean of the power station coefficients corresponding to the multiple engine failures and environmental data abnormalities is calculated, marked as the second failure mean; the first failure mean is divided by the mean square error to obtain the constant Z, and the second failure mean is divided by the mean square error to obtain the constant C;

[0127] If a preliminary first fault instruction is generated, the time point corresponding to the preliminary first fault instruction is marked as a suspected fault point, and a suspected fault point set is set. If the time points corresponding to multiple suspected fault points are continuous time points, the multiple continuous suspected fault points are added to the suspected fault point set in time point order. If the times corresponding to multiple suspected fault points are not continuous time points, the suspected fault points are not added to the suspected fault point set. The time point corresponding to the last suspected fault point in the suspected fault point set is subtracted from the time point corresponding to the first suspected fault point to obtain a time point difference, and the time point difference is marked as a duration. Exemplarily, the time points corresponding to the suspected fault points are the m+1th time point, the m+2th time point, and the m+3th time point, and these three time points are continuous time points. Therefore, the suspected fault points corresponding to these three time points are added to the suspected fault point set in time order, that is, the time point corresponding to the first suspected fault point is the m+1th time point, and the time point corresponding to the last suspected fault point is the m+3th time point. The time point difference obtained by subtracting the m+1th time point from the m+3th time point is 2, that is, the duration is 2.

[0128] comparing the duration to a preset duration threshold;

[0129] If the duration is greater than or equal to the duration threshold, a first fault instruction is generated; indicating that the variation range of the power station data is in an abnormal variation range for a long time, resulting in a fault in the power station operation state;

[0130] If the duration is less than the duration threshold, the first fault instruction is not generated; this indicates that the variation range of the power station data is within the normal variation range, and the operation status of the power station is normal;

[0131] If the first fault instruction is generated, the set of suspected fault points is cleared;

[0132] It should be noted that the duration threshold is obtained from the engine's technical specifications and user manual;

[0133] The method for determining whether to generate a second fault instruction includes:

[0134] If a preliminary second fault instruction is generated, the predicted environmental coefficient at the future time point is subtracted from the environmental coefficient corresponding to the time point before the time point to obtain an environmental coefficient difference. For example, the predicted environmental coefficient corresponding to the m+1th time point is subtracted from the environmental coefficient corresponding to the mth time point, and the predicted environmental coefficient corresponding to the m+2th time point is subtracted from the environmental coefficient corresponding to the m+1th time point;

[0135] Compare the absolute value of the environmental coefficient difference with a preset environmental difference threshold;

[0136] If the environmental coefficient difference is less than or equal to the environmental difference threshold, the second fault instruction is not generated; this indicates that the change in the environmental coefficient is normal at this time, and the environmental data is within the normal range of variation;

[0137] If the environmental coefficient difference is greater than the environmental difference threshold, a second fault instruction is generated, indicating that the change in the environmental coefficient is abnormal, and the environmental data is within an abnormal variation range, resulting in an abnormal operation state of the power station;

[0138] It should be noted that the environmental difference threshold is obtained by subtracting the environmental data corresponding to the first abnormal environmental data from the environmental data at the previous moment to obtain the difference when the environmental data is abnormal during the operation of the historical power station, and so on to obtain multiple differences, and the average of the multiple differences is used as the environmental difference threshold;

[0139] The method for determining whether to generate a third fault instruction includes:

[0140] If a preliminary third fault instruction is generated, the grid coefficient difference is obtained by subtracting the grid coefficient corresponding to the previous time point from the predicted future time point;

[0141] Comparing the absolute value of the grid coefficient difference with a preset grid difference threshold;

[0142] If the grid coefficient difference is less than or equal to the grid difference threshold, the third fault instruction is not generated; this indicates that the change in the grid coefficient is normal and the grid data is within the normal range of variation;

[0143] If the grid coefficient difference is greater than the grid difference threshold, a third fault instruction is generated, indicating that the change in the grid coefficient is abnormal, and the grid data is within an abnormal variation range, resulting in an abnormal operation state of the power station;

[0144] It should be noted that the grid difference threshold is obtained by subtracting the grid data corresponding to the first occurrence of grid data abnormality from the grid data at the previous moment to obtain the difference when the grid data abnormality occurs during the historical power station operation, and so on to obtain multiple differences, and the average of the multiple differences is used as the grid difference threshold;

[0145] Adjustment instructions: adjust the power plant data corresponding to the power plant coefficient at a future time point according to the fault instructions, and determine whether to generate an early warning instruction;

[0146] The method for adjusting the power plant data corresponding to the power plant coefficient at a future time point includes:

[0147] If the first fault instruction is generated, the rotation speed and power in the power station data corresponding to the future time point are adjusted;

[0148] The first product plus the mean of the power station coefficient is taken as the maximum power station coefficient, and the difference obtained by subtracting the maximum power station coefficient from the power station coefficient at the future time point is marked as the power station coefficient adjustment amount;

[0149] The power station coefficient adjustment amount is used as the input of the fuzzy controller, the speed adjustment amount and the power adjustment amount are used as the output of the fuzzy controller, and the range of the speed adjustment amount and the power adjustment amount is defined according to the speed range value and the power range value;

[0150] According to the relationship among the speed regulation, power regulation and power station coefficient regulation, the fuzzy control rule is designed, and the fuzzy control rule is expressed in the form of IF-THEN; the relationship among the speed regulation, power regulation and power station coefficient regulation is obtained according to the calculation formula of the power station coefficient;

[0151] The speed regulation variable is converted into a fuzzy set through fuzzification technology;

[0152] According to the fuzzified speed regulation value and fuzzy control rules, the fuzzy output is obtained;

[0153] The fuzzy output is converted into speed regulation quantity and power regulation quantity by using defuzzification technology;

[0154] The speed and power in the power station data corresponding to the future time point are adjusted according to the speed adjustment amount and the power adjustment amount;

[0155] The maximum power plant coefficient is the maximum value that the power plant coefficient can reach when the power plant is operating normally. Since the power plant data needs to be quickly restored to the normal range in a short time, the power plant coefficient in the future time is adjusted to the maximum power plant coefficient to ensure that the power plant quickly returns to normal status.

[0156] It should be noted that the speed range value and the power range value are obtained by the staff during the normal operation of the power station in history, collecting the speed and power of the engine in the power station multiple times, collecting multiple speeds and powers each time, calculating the corresponding speed average and power average for the multiple speeds and powers collected each time, sorting the multiple speed averages and power averages from large to small, and taking the first speed average and the last speed average as the maximum and minimum values ​​of the speed range value in positive order, and taking the first power average and the last power average as the maximum and minimum values ​​of the power range value;

[0157] The reason for adjusting only the rotation speed and power in the power plant data corresponding to the future time point is that the temperature and electromagnetic radiation in the power plant data cannot be automatically adjusted by the monitoring system, so only the rotation speed and power in the power plant data corresponding to the future time point are adjusted;

[0158] The method for determining whether to generate an early warning instruction includes:

[0159] If the second fault instruction or the third fault instruction is generated, an early warning instruction is generated; the power station is stopped to remind the staff to carry out maintenance on the power station in time to avoid the power station being in abnormal operation for a long time;

[0160] This embodiment acquires the historical characteristic coefficient time series of the power station coefficient, the power grid coefficient and the environmental coefficient by collecting and analyzing the historical power station characteristic data, and trains a plurality of accurate prediction models based on these sequences, so that the power station, the power grid and the environmental coefficient at the future time point can be predicted; these prediction models are used for monitoring data, and once an abnormal situation is detected, the monitoring system can generate fault instructions and early warning instructions in time, prompting the operation and maintenance personnel to make adjustments and maintenance, thereby improving the stability and reliability of the power station operation, reducing the potential risk of failure, and improving the monitoring and management level of the power station operation status.

[0161] Example 2

[0162] See also Figure 3 As shown, the part not described in detail in this embodiment is described in Example 1, and a method for monitoring the operation status of a power plant based on multivariate time series prediction is provided, including:

[0163] Collect m groups of historical power plant characteristic data, the historical power plant characteristic data including power plant data, power grid data and environmental data, where m is an integer greater than 1;

[0164] According to the historical power plant characteristic data, the corresponding historical characteristic coefficients are calculated, and the corresponding historical characteristic coefficient time series are constructed; the historical characteristic coefficients include power plant coefficients, power grid coefficients and environmental coefficients; the historical characteristic coefficient time series includes power plant coefficient time series, power grid coefficient time series and environmental coefficient time series;

[0165] Based on the power plant coefficient time series, a first coefficient prediction model for predicting the power plant coefficient at a future time point is trained; based on the power grid coefficient time series, a second coefficient prediction model for predicting the power grid coefficient at a future time point is trained; based on the environmental coefficient time series, a third coefficient prediction model for predicting the environmental coefficient at a future time point is trained;

[0166] Input the historical characteristic coefficient time series into the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model respectively, obtain the power station coefficient, the power grid coefficient and the environmental coefficient at the future time point, and mark them as monitoring data; determine whether to generate a fault instruction according to the monitoring data;

[0167] The power plant data corresponding to the power plant coefficient at a future time point is adjusted according to the fault instruction, and it is determined whether to generate an early warning instruction.

[0168] Further, the m groups of historical power station characteristic data correspond to m time points; the power station data include speed, power, temperature and electromagnetic radiation; the electromagnetic radiation includes electromagnetic radiation frequency and electromagnetic radiation intensity; the speed is the speed of the engine in the power station, the power is the output power of the engine in the power station, the temperature is the temperature of the engine in the power station, and the electromagnetic radiation is the electromagnetic radiation of the engine in the power station; the power grid data includes power grid voltage, power grid current and power grid frequency;

[0169] The environmental data includes first environmental data and second environmental data. The first environmental data includes a first environmental temperature and a first environmental humidity. The second environmental data includes a second environmental temperature and a second environmental humidity.

[0170] Furthermore, the calculation method of the power station coefficient includes:

[0171]

[0172] Where DZX i is the power station coefficient, ZS i is the speed, GL i is power, DW i is the temperature, DP i is the electromagnetic radiation frequency, DQ i is the electromagnetic radiation intensity, ω1, ω2, ω3 are preset weights, i is the i-th time point, i∈[1,m], m is the number of time points;

[0173] The calculation method of the power grid coefficient includes:

[0174]

[0175] Where DWX i is the power grid coefficient, DY i Grid voltage, DL i Grid current, PL i Grid frequency, ω4 and ω5 are preset weights;

[0176] The calculation method of the environmental factor includes:

[0177]

[0178] In the formula, HJX i is the environmental factor, is the first ambient temperature, is the first ambient humidity, is the second ambient temperature, is the second ambient humidity, ω6 and ω7 are preset weights, and A is a preset constant.

[0179] Furthermore, the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model are all LSTM models in the recurrent neural network model, and the training process is consistent;

[0180] The training process of the first coefficient prediction model includes:

[0181] Define the architecture of the coefficient prediction model, including input state, hidden state, number of output units, and hyperparameters; use the deep learning framework to create the coefficient prediction model network layer, which includes an input layer, multiple LSTM layers, and an output layer, where each LSTM layer consists of an input gate, a forget gate, an output gate, and a cell state; connect the created LSTM layers in sequence to form a complete coefficient prediction model, ensure that the input dimension matches the feature dimension of the data, and set the number of units in the output layer to accommodate the required prediction results;

[0182] The power plant coefficient time series is used as the input of the first coefficient prediction model. The first coefficient prediction model uses the power plant coefficient at a future time point as the output, and the actual power plant coefficient is used as the prediction target. The first coefficient prediction model is trained, and the prediction result is evaluated for model accuracy using the mean absolute percentage error MAPE. When the calculated MAPE is less than the preset MAPE, the training of the first coefficient prediction model is completed; wherein, the calculation formula of MAPE is: Where K is the Kth time point, α K is the actual power station coefficient corresponding to the Kth time point, μ K is the predicted power plant coefficient corresponding to the Kth time point, and R is the number of predicted power plant coefficients.

[0183] Further, the fault instruction includes a preliminary fault instruction, a first fault instruction, a second fault instruction and a third fault instruction; the preliminary fault instruction includes a preliminary first fault instruction, a preliminary second fault instruction and a preliminary third fault instruction;

[0184] The method for determining whether to generate a first fault instruction and a preliminary fault instruction includes:

[0185] According to m power station coefficients, calculate the corresponding power station coefficient mean and power station coefficient mean square error;

[0186] The calculation process of the mean power plant coefficient includes:

[0187]

[0188] In the formula, JZ is the mean value of the power station coefficient, and m is the number of power station coefficients;

[0189] The calculation process of the mean square error of the power station coefficient includes:

[0190]

[0191] Where, JFC is the mean square error of power plant coefficients, and m is the number of power plant coefficients;

[0192] Preset constants Z and C, C>Z; subtract the power station coefficient mean from the predicted future time point power station coefficient to obtain the power station coefficient difference, multiply the power station coefficient mean square error by Z to obtain the first product, multiply the power station coefficient mean square error by C to obtain the second product, and compare the absolute value of the power station coefficient difference with the first product and the second product respectively;

[0193] If the absolute value of the power station coefficient difference is less than or equal to the first product, no preliminary fault instruction is generated;

[0194] If the absolute value of the power station coefficient difference is greater than the first product and less than or equal to the second product, a preliminary first fault instruction and a preliminary second fault instruction are generated;

[0195] If the absolute value of the power station coefficient difference is greater than the second product, a preliminary first fault instruction and a preliminary third fault instruction are generated;

[0196] If a preliminary first fault instruction is generated, the time point corresponding to the preliminary first fault instruction is marked as a suspected fault point, and a set of suspected fault points is set. If the time points corresponding to multiple suspected fault points are continuous time points, the multiple continuous suspected fault points are added to the set of suspected fault points in the order of time points. If the times corresponding to multiple suspected fault points are not continuous time points, the suspected fault points are not added to the set of suspected fault points. The time point corresponding to the last suspected fault point in the set of suspected fault points is subtracted from the time point corresponding to the first suspected fault point to obtain the time point difference, and the time point difference is marked as the duration.

[0197] comparing the duration to a preset duration threshold;

[0198] If the duration is greater than or equal to the duration threshold, generating a first fault instruction;

[0199] If the duration is less than the duration threshold, no first fault instruction is generated;

[0200] If the first fault instruction is generated, the set of suspected fault points is cleared.

[0201] Further, the method for determining whether to generate the second fault instruction includes:

[0202] If a preliminary second fault instruction is generated, the predicted environmental coefficient at a future time point is subtracted from the environmental coefficient corresponding to the time point before the time point to obtain the difference in environmental coefficients;

[0203] Compare the absolute value of the environmental coefficient difference with a preset environmental difference threshold;

[0204] If the environmental coefficient difference is less than or equal to the environmental difference threshold, the second fault instruction is not generated;

[0205] If the environmental coefficient difference is greater than the environmental difference threshold, a second fault instruction is generated.

[0206] Further, the method for determining whether to generate the third fault instruction includes:

[0207] If a preliminary third fault instruction is generated, the grid coefficient difference is obtained by subtracting the grid coefficient corresponding to the previous time point from the predicted future time point;

[0208] Comparing the absolute value of the grid coefficient difference with a preset grid difference threshold;

[0209] If the grid coefficient difference is less than or equal to the grid difference threshold, the third fault instruction is not generated;

[0210] If the grid coefficient difference is greater than the grid difference threshold, a third fault instruction is generated.

[0211] Furthermore, the method for adjusting the power plant data corresponding to the power plant coefficient at a future time point includes:

[0212] If the first fault instruction is generated, the rotation speed and power in the power station data corresponding to the future time point are adjusted;

[0213] The first product plus the mean of the power station coefficient is taken as the maximum power station coefficient, and the difference obtained by subtracting the maximum power station coefficient from the power station coefficient at the future time point is marked as the power station coefficient adjustment amount;

[0214] The power station coefficient adjustment amount is used as the input of the fuzzy controller, the speed adjustment amount and the power adjustment amount are used as the output of the fuzzy controller, and the range of the speed adjustment amount and the power adjustment amount is defined according to the speed range value and the power range value;

[0215] According to the relationship among the speed regulation, power regulation and power station coefficient regulation, the fuzzy control rule is designed, and the fuzzy control rule is expressed in the form of IF-THEN; the relationship among the speed regulation, power regulation and power station coefficient regulation is obtained according to the calculation formula of the power station coefficient;

[0216] The speed regulation variable is converted into a fuzzy set through fuzzification technology;

[0217] According to the fuzzified speed regulation value and fuzzy control rules, the fuzzy output is obtained;

[0218] The fuzzy output is converted into speed regulation quantity and power regulation quantity by using defuzzification technology;

[0219] The speed and power in the power station data corresponding to the future time point are adjusted according to the speed adjustment amount and the power adjustment amount.

[0220] Further, the method for determining whether to generate an early warning instruction includes:

[0221] If the second fault instruction or the third fault instruction is generated, a warning instruction is generated.

[0222] Example 3

[0223] See also Figure 4 As shown, this embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any one of the power plant operation status monitoring methods based on multivariate time series prediction provided by the above methods is implemented.

[0224] Since the electronic device introduced in this embodiment is an electronic device used to implement the power plant operation status monitoring method based on multivariate time series prediction in the embodiment of the present application, based on the power plant operation status monitoring method based on multivariate time series prediction introduced in the embodiment of the present application, the technical personnel of the field can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as the technical personnel of the field implement the electronic device used by the power plant operation status monitoring method based on multivariate time series prediction in the embodiment of the present application, it belongs to the scope of protection of this application.

[0225] Example 4

[0226] This embodiment discloses a computer-readable storage medium, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the power plant operation status monitoring methods based on multivariate time series prediction provided by the above methods is implemented.

[0227] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0228] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0229] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0230] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0231] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0232] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0233] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0234] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power plant operation status monitoring method based on multivariate time series prediction, characterized in that: include: Collect m groups of historical power plant characteristic data, the historical power plant characteristic data including power plant data, power grid data and environmental data, where m is an integer greater than 1; According to the historical power plant characteristic data, the corresponding historical characteristic coefficients are calculated, and the corresponding historical characteristic coefficient time series is constructed; The historical characteristic coefficients include power plant coefficients, power grid coefficients and environmental coefficients; the historical characteristic coefficient time series include power plant coefficient time series, power grid coefficient time series and environmental coefficient time series; Based on the power plant coefficient time series, a first coefficient prediction model for predicting the power plant coefficient at a future time point is trained; based on the power grid coefficient time series, a second coefficient prediction model for predicting the power grid coefficient at a future time point is trained; based on the environmental coefficient time series, a third coefficient prediction model for predicting the environmental coefficient at a future time point is trained; Input the historical characteristic coefficient time series into the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model respectively, obtain the power station coefficient, the power grid coefficient and the environmental coefficient at the future time point, and mark them as monitoring data; determine whether to generate a fault instruction according to the monitoring data; The fault instructions include a preliminary fault instruction, a first fault instruction, a second fault instruction and a third fault instruction; the preliminary fault instructions include a preliminary first fault instruction, a preliminary second fault instruction and a preliminary third fault instruction; The method for determining whether to generate a first fault instruction and a preliminary fault instruction comprises: According to m power station coefficients, calculate the corresponding power station coefficient mean and power station coefficient mean square error; The calculation process of the mean power plant coefficient includes: In the formula, JZ is the mean value of the power station coefficient, and m is the number of power station coefficients; The calculation process of the mean square error of the power station coefficient includes: Where, JFC is the mean square error of power plant coefficients, and m is the number of power plant coefficients; Preset constants Z and C, C>Z; subtract the power station coefficient mean from the predicted future time point power station coefficient to obtain the power station coefficient difference, multiply the power station coefficient mean square error by Z to obtain the first product, multiply the power station coefficient mean square error by C to obtain the second product, and compare the absolute value of the power station coefficient difference with the first product and the second product respectively; If the absolute value of the power station coefficient difference is less than or equal to the first product, no preliminary fault instruction is generated; If the absolute value of the power station coefficient difference is greater than the first product and less than or equal to the second product, a preliminary first fault instruction and a preliminary second fault instruction are generated; If the absolute value of the power station coefficient difference is greater than the second product, a preliminary first fault instruction and a preliminary third fault instruction are generated; If a preliminary first fault instruction is generated, the time point corresponding to the preliminary first fault instruction is marked as a suspected fault point, and a set of suspected fault points is set. If the time points corresponding to multiple suspected fault points are continuous time points, the multiple continuous suspected fault points are added to the set of suspected fault points in the order of time points. If the times corresponding to multiple suspected fault points are not continuous time points, the suspected fault points are not added to the set of suspected fault points. The time point corresponding to the last suspected fault point in the set of suspected fault points is subtracted from the time point corresponding to the first suspected fault point to obtain the time point difference, and the time point difference is marked as the duration. comparing the duration to a preset duration threshold; If the duration is greater than or equal to the duration threshold, generating a first fault instruction; If the duration is less than the duration threshold, no first fault instruction is generated; If the first fault instruction is generated, the set of suspected fault points is cleared; The method for determining whether to generate a second fault instruction includes: If a preliminary second fault instruction is generated, the predicted environmental coefficient at the future time point is subtracted from the environmental coefficient corresponding to the time point before the time point to obtain the difference in environmental coefficients; Compare the absolute value of the environmental coefficient difference with a preset environmental difference threshold; If the environmental coefficient difference is less than or equal to the environmental difference threshold, the second fault instruction is not generated; If the environmental coefficient difference is greater than the environmental difference threshold, a second fault instruction is generated; The method for determining whether to generate a third fault instruction includes: If a preliminary third fault instruction is generated, the grid coefficient corresponding to the previous time point is subtracted from the predicted future time point grid coefficient to obtain the grid coefficient difference; Comparing the absolute value of the grid coefficient difference with a preset grid difference threshold; If the grid coefficient difference is less than or equal to the grid difference threshold, the third fault instruction is not generated; If the grid coefficient difference is greater than the grid difference threshold, a third fault instruction is generated; According to the fault instruction, the power plant data corresponding to the power plant coefficient at the future time point is adjusted, and it is determined whether to generate an early warning instruction; The method for determining whether to generate an early warning instruction includes: If the second fault instruction or the third fault instruction is generated, a warning instruction is generated.

2. The power plant operation status monitoring method based on multivariate time series prediction according to claim 1 is characterized in that: The m groups of historical power station characteristic data correspond to m time points; the power station data include rotation speed, power, temperature and electromagnetic radiation; the electromagnetic radiation includes electromagnetic radiation frequency and electromagnetic radiation intensity; the rotation speed is the rotation speed of the engine in the power station, the power is the output power of the engine in the power station, the temperature is the temperature of the engine in the power station, and the electromagnetic radiation is the electromagnetic radiation of the engine in the power station; the power grid data includes power grid voltage, power grid current and power grid frequency; The environmental data includes first environmental data and second environmental data, the first environmental data includes a first environmental temperature and a first environmental humidity, and the second environmental data includes a second environmental temperature and a second environmental humidity.

3. The power plant operation status monitoring method based on multivariate time series prediction according to claim 2 is characterized in that: The calculation method of the power station coefficient includes: In the formula, DZX i is the power station coefficient, ZS i is the rotation speed, GL i is power, DW i is the temperature, DP i is the electromagnetic radiation frequency, DQ i is the electromagnetic radiation intensity, ω1, ω2, ω3 are preset weights, i is the i-th time point, i∈[1,m], m is the number of time points; The calculation method of the power grid coefficient includes: Where DWX i is the power grid coefficient, DY i Grid voltage, DL i Grid current, PL i Grid frequency, ω4 and ω5 are preset weights; The calculation method of the environmental factor includes: HJXi=ω6×ln(WD1i+SD1i+A)+ω7×(WD2i+SD2i); Wherein, HJXi is the environmental coefficient, WD1i is the first environmental temperature, SD1i is the first environmental humidity, WD2i is the second environmental temperature, SD2i is the second environmental humidity, ω6 and ω7 are preset weights, and A is a preset constant.

4. The power plant operation status monitoring method based on multivariate time series prediction according to claim 3 is characterized in that: The first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model are all LSTM models in the recurrent neural network model, and the training processes are consistent; The training process of the first coefficient prediction model includes: Define the architecture of the coefficient prediction model, including input state, hidden state, number of output units, and hyperparameters; use the deep learning framework to create the coefficient prediction model network layer, which includes an input layer, multiple LSTM layers, and an output layer, where each LSTM layer consists of an input gate, a forget gate, an output gate, and a cell state; connect the created LSTM layers in sequence to form a complete coefficient prediction model, ensure that the input dimension matches the feature dimension of the data, and set the number of units in the output layer to accommodate the required prediction results; The power plant coefficient time series is used as the input of the first coefficient prediction model. The first coefficient prediction model takes the power plant coefficient at a future time point as the output, takes the actual power plant coefficient as the prediction target, trains the first coefficient prediction model, and uses the mean absolute percentage error MAPE to evaluate the model accuracy of the prediction result. When the calculated MAPE is less than the preset MAPE, the training of the first coefficient prediction model is completed; wherein, the calculation formula of MAPE is: Where K is the Kth time point, α K is the actual power station coefficient corresponding to the Kth time point, μ K is the predicted power plant coefficient corresponding to the Kth time point, and R is the number of predicted power plant coefficients.

5. The power plant operation status monitoring method based on multivariate time series prediction according to claim 4 is characterized in that: The method for adjusting the power plant data corresponding to the power plant coefficient at a future time point includes: If the first fault instruction is generated, the rotation speed and power in the power station data corresponding to the future time point are adjusted; The first product plus the mean of the power station coefficient is taken as the maximum power station coefficient, and the difference obtained by subtracting the maximum power station coefficient from the power station coefficient at the future time point is marked as the power station coefficient adjustment amount; The power station coefficient adjustment amount is used as the input of the fuzzy controller, the speed adjustment amount and the power adjustment amount are used as the output of the fuzzy controller, and the range of the speed adjustment amount and the power adjustment amount is defined according to the speed range value and the power range value; According to the relationship among the speed regulation, power regulation and power station coefficient regulation, the fuzzy control rule is designed, and the fuzzy control rule is expressed in the form of IF-THEN; the relationship among the speed regulation, power regulation and power station coefficient regulation is obtained according to the calculation formula of the power station coefficient; The speed regulation variable is converted into a fuzzy set through fuzzification technology; According to the fuzzified speed regulation value and fuzzy control rules, the fuzzy output is obtained; The fuzzy output is converted into speed regulation quantity and power regulation quantity by using defuzzification technology; The speed and power in the power station data corresponding to the future time point are adjusted according to the speed adjustment amount and the power adjustment amount.

6. A power plant operation status monitoring system based on multivariate time series prediction, implementing the power plant operation status monitoring method based on multivariate time series prediction according to any one of claims 1 to 5, characterized in that: include: A data collection module collects m groups of historical power plant characteristic data, the historical power plant characteristic data includes power plant data, power grid data and environmental data; The data processing module calculates the corresponding historical characteristic coefficients according to the historical power station characteristic data, and constructs the corresponding historical characteristic coefficient time series; The historical characteristic coefficients include power plant coefficients, power grid coefficients and environmental coefficients; the historical characteristic coefficient time series include power plant coefficient time series, power grid coefficient time series and environmental coefficient time series; The model training module trains a first coefficient prediction model for predicting the power plant coefficient at a future time point based on the power plant coefficient time series; trains a second coefficient prediction model for predicting the power grid coefficient at a future time point based on the power grid coefficient time series; and trains a third coefficient prediction model for predicting the environmental coefficient at a future time point based on the environmental coefficient time series; The state monitoring module inputs the historical characteristic coefficient time series into the first coefficient prediction model, the second coefficient prediction model and the third coefficient prediction model respectively, obtains the power station coefficient, the power grid coefficient and the environmental coefficient at the future time point, and marks them as monitoring data; determines whether to generate a fault instruction according to the monitoring data; The adjustment module adjusts the power plant data corresponding to the power plant coefficient at a future time point according to the fault instruction, and determines whether to generate an early warning instruction.

Images