A compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization

Abstract

The application discloses a kind of based on LSTM neural network and boundary constraint optimization's compressor performance degradation analysis and water washing recommended method, it is related to gas turbine generator set compressor performance degradation analysis technical field, including using multiple linear regression algorithm, the calculation equation of unit load-compressor efficiency is established, and the maximum / minimum curve of compressor efficiency before and after water washing and daily average efficiency curve are drawn.Through LSTM neural network prediction efficiency degradation trend, and with predicted data constructs boundary constraint optimization problem to calculate water washing remaining time, provides optimization suggestion.The method described in the application can quantitatively analyze and show the degradation of compressor efficiency and recommend water washing in advance, ensure that the compressor is always running in the best state, improve the economy and safety of operation, adapt to the compressor performance of different units, provide scientific and timely optimization guidance for operating personnel, and have wide application prospect.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of gas turbine generator set compressor performance degradation analysis, in particular to a compressor performance degradation analysis and water washing recommendation method based on an LSTM neural network and boundary constraint optimization. BACKGROUND

[0002] The purpose of the compressor performance degradation analysis of the gas turbine generator set is to analyze whether the compressor efficiency is too low through data, and to intuitively show the daily compressor efficiency degradation process in a chart, so that the operation personnel can judge whether the compressor efficiency needs to be improved through water washing.

[0003] At present, the compressors of most gas turbine generator sets are improved in efficiency through offline water washing, and the compressor is water washed before the unit is started every certain period of time.

[0004] At present, the compressor is water washed mainly through artificial experience judgment and a fixed cycle method, and the over-washing or under-washing situation often occurs, which causes certain negative effects on the safety and economy of the unit operation. SUMMARY

[0005] In view of the above problems, the application is proposed.

[0006] Therefore, the technical problem to be solved by the application is that the existing compressor water washing method has the over-washing or under-washing situation, which causes certain negative effects on the safety and economy of the unit operation.

[0007] To solve the above technical problems, the application provides the following technical scheme: a compressor performance degradation analysis and water washing recommendation method based on an LSTM neural network and boundary constraint optimization, which comprises collecting operation history data, fitting a load-efficiency upper and lower limit equation according to the data before and after water washing, outputting the compressor efficiency every minute and the mean value of the unit load, fitting a daily load-efficiency equation; drawing a compressor efficiency degradation curve based on the load-efficiency equation, training an LSTM neural network, and predicting the future compressor efficiency degradation state; outputting the cumulative efficiency gas cost and water washing cost based on the future compressor efficiency degradation state, constructing a boundary constraint optimization problem to output the water washing remaining time recommendation, and triggering the water washing warning according to the water washing remaining time and the compressor efficiency degradation state.

[0008] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization, the collection of operation history data comprises: collecting single-day effective operation history data of the compressor before water washing and single-day effective operation history data of the compressor after water washing from the SIS system, wherein the effective operation history data comprises compressor efficiency and gas turbine power generation load; outputting the average value of the compressor efficiency and the average value of the gas turbine power generation load every minute at an interval of 1 minute; and obtaining daily operation data of the compressor from the SIS system, wherein each piece of data comprises compressor efficiency and gas turbine power generation load; outputting the average value of the compressor efficiency and the average value of the gas turbine power generation load every minute at an interval of 1 minute.

[0009] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization, the fitting of the load-efficiency upper and lower limit equation according to the data before and after water washing comprises: fitting a quadratic equation of one variable of the compressor efficiency before water washing by using a quadratic polynomial fitting algorithm, wherein the average value of the compressor efficiency is taken as the dependent variable, and the average value of the gas turbine power generation load is taken as the independent variable; and fitting a quadratic equation of one variable of the compressor efficiency after water washing by using a quadratic polynomial fitting algorithm, wherein the average value of the compressor efficiency is taken as the dependent variable, and the average value of the gas turbine power generation load is taken as the independent variable.

[0010] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization, the fitting of the load-efficiency upper and lower limit equation according to the data before and after water washing comprises: fitting a quadratic equation of one variable of the compressor efficiency before water washing by using a quadratic polynomial fitting algorithm, wherein the average value of the compressor efficiency is taken as the dependent variable, and the average value of the gas turbine power generation load is taken as the independent variable; and fitting a quadratic equation of one variable of the compressor efficiency after water washing by using a quadratic polynomial fitting algorithm, wherein the average value of the compressor efficiency is taken as the dependent variable, and the average value of the gas turbine power generation load is taken as the independent variable.

[0011] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization, the drawing of the compressor efficiency degradation curve based on the load-efficiency equation comprises: taking the compressor efficiency as the y-axis and the gas turbine power generation load as the x-axis, and drawing a daily compressor efficiency degradation curve and an equation curve.

[0012] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization, the training of the LSTM neural network and the prediction of the future compressor efficiency degradation state comprise: using the LSTM neural network to perform model training; and performing time series prediction on real-time data of the power plant and outputting an array of time-load-efficiency.

[0013] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization, the cumulative efficiency gas cost and the water washing cost are output according to the future compressor efficiency degradation state, a boundary constraint optimization problem is constructed, a water washing remaining time recommendation is output, and a water washing early warning is triggered according to the water washing remaining time and the compressor efficiency degradation state.

[0014] Another object of the present application is to provide a compressor performance degradation analysis and water washing recommendation system based on an LSTM neural network and boundary constraint optimization, which can effectively avoid the influence of the gas turbine power generation load on the compressor efficiency by monitoring the degradation process of the compressor efficiency through a two-dimensional curve graph.

[0015] As a preferred scheme of the compressor performance degradation analysis and water washing recommendation system based on the LSTM neural network and boundary constraint optimization, the system comprises an initialization module, an algorithm fitting module, an LSTM model construction module and an LSTM model prediction module; the initialization module is used for collecting operation history data; the algorithm fitting module is used for fitting a load-efficiency upper and lower limit equation based on a one-dimensional quadratic polynomial fitting algorithm, outputting a per-minute compressor efficiency and a unit load mean fitting daily load-efficiency equation; the LSTM model construction module is used for drawing a daily compressor efficiency degradation curve graph, training an LSTM neural network and predicting a future compressor efficiency degradation state; and the LSTM model prediction module is used for outputting a water washing remaining time recommendation and triggering a water washing early warning according to the water washing remaining time and the compressor efficiency degradation state.

[0016] Still another object of the present application is to provide a compressor performance degradation analysis and water washing recommendation device based on an LSTM neural network and boundary constraint optimization, which comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the steps of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization.

[0017] Still another object of the present application is to provide a compressor performance degradation analysis and water washing recommendation storage medium based on an LSTM neural network and boundary constraint optimization, which stores a computer program, and the computer program is executed by a processor to realize the steps of the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization.

[0018] The compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application can overcome the problems of human subjectivity and operation lag of the traditional experience method through analysis of a large amount of historical data collected by SIS in a data-driven manner, can quantitatively analyze and display the degradation of the compressor efficiency and make water washing recommendations in advance, ensure that the compressor always operates in the best state, improve the economy and safety of operation, and adapt to the compressor performance of different units. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 A compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application embodiment 1 is provided.

[0021] Figure 2 The compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application embodiment 1 uses a quadratic polynomial fitting curve diagram for data before and after water washing of the compressor.

[0022] Figure 3 The compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application embodiment 1 calculates 1440 average compressor efficiency values per day.

[0023] Figure 4 The compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application embodiment 1 calculates 1440 average gas turbine power generation load values per day.

[0024] Figure 5 The compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application embodiment 1 draws a fitting compressor efficiency drop versus gas increment diagram.

[0025] Figure 6 The compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization provided by the application embodiment 1 draws a daily compressor efficiency degradation curve diagram.

[0026] Figure 7 A compressor water washing residual time recommendation diagram based on an LSTM neural network and boundary constraint optimization is provided for the compressor performance degradation analysis and water washing recommendation method of Embodiment 1 of the present application. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0028] Embodiment 1, with reference to Figures 1-7 For an embodiment of the present application, a compressor performance degradation analysis and water washing recommendation method based on an LSTM neural network and boundary constraint optimization is provided, comprising:

[0029] S1: Collecting running history data, fitting load-efficiency upper and lower limit equations according to data before and after water washing, outputting per-minute compressor efficiency, unit load mean fitting daily load-efficiency equation.

[0030] Further, sufficient number of single-day effective running history data before water washing and single-day effective running history data after water washing of the compressor are obtained from the SIS system, each piece of data including: compressor efficiency (%), gas turbine power generation load (MW), and the per-minute compressor efficiency average and the gas turbine power generation load average are calculated at 1-minute intervals.

[0031] It should be noted that the daily running data of the compressor are obtained from the SIS system, each piece of data including: compressor efficiency (%), gas turbine power generation load (MW), and the per-minute compressor efficiency average and the gas turbine power generation load average are calculated at 1-minute intervals.

[0032] It should also be noted that a one-dimensional quadratic polynomial fitting algorithm is used to fit a one-dimensional quadratic equation of the compressor efficiency before water washing with the compressor efficiency average as y and the gas turbine power generation load average as x. .

[0033] Further, a one-dimensional quadratic polynomial fitting algorithm is used to fit a one-dimensional quadratic equation of the compressor efficiency after water washing with the compressor efficiency average as y and the gas turbine power generation load average as x. .

[0034] It should be noted that the quadratic polynomial fitting algorithm is used to fit the quadratic equation of the compressor efficiency with the average compressor efficiency as y and the average gas turbine power generation load as x .

[0035] Table 1 is a data table of the compressor efficiency and the gas turbine power generation load before and after the compressor water washing according to the present application. The single-day data of the effective operation of the compressor before and after the water washing are found, the compressor efficiency and the gas turbine power generation load data are collected every 1 second, the average compressor efficiency of the 60 compressor efficiency data collected in the previous 1 minute is calculated, the average gas turbine power generation load of the 60 gas turbine power generation load data collected in the previous 1 minute is calculated, and the single-day each forms 1440 average compressor efficiency and average gas turbine power generation load before the water washing. The 1440 average compressor efficiency and average gas turbine power generation load formed in the single day before and after the water washing are judged for effectiveness. When the average gas turbine power generation load calculated in a certain minute is less than the set value Lm, the average compressor efficiency calculated in the minute is judged as invalid, otherwise it is valid. In the embodiment, Lm = 160 is set.

[0036] Table 1 is a data table of the compressor efficiency and the gas turbine power generation load before and after the compressor water washing according to the present application. The single-day data of the effective operation of the compressor before and after the water washing are found, the compressor efficiency and the gas turbine power generation load data are collected every 1 second, the average compressor efficiency of the 60 compressor efficiency data collected in the previous 1 minute is calculated, the average gas turbine power generation load of the 60 gas turbine power generation load data collected in the previous 1 minute is calculated, and the single-day each forms 1440 average compressor efficiency and average gas turbine power generation load before the water washing. The 1440 average compressor efficiency and average gas turbine power generation load formed in the single day before and after the water washing are judged for effectiveness. When the average gas turbine power generation load calculated in a certain minute is less than the set value Lm, the average compressor efficiency calculated in the minute is judged as invalid, otherwise it is valid. In the embodiment, Lm = 160 is set.

[0037]

[0038] Figure 2 The data before and after the water washing of the compressor according to the present application is fitted with a quadratic polynomial curve. After removing the invalid values from the 1440 average compressor efficiency and average gas turbine power generation load formed in the single day before and after the water washing, the quadratic equation of the compressor efficiency before the water washing with the average compressor efficiency as y and the average gas turbine power generation load as x is fitted by using the quadratic polynomial fitting algorithm , and the quadratic equation of the compressor efficiency after the water washing is expressed as:

[0039] ;

[0040] The quadratic equation of the compressor efficiency before the water washing in the embodiment is: ;

[0041] The quadratic equation of the compressor efficiency after the water washing is: .

[0042] S2: Draw the compressor efficiency degradation curve based on the load-efficiency equation, train the LSTM neural network, and predict the future compressor efficiency degradation state.

[0043] Further, the compressor efficiency (%) is taken as the y-axis, and the gas turbine power generation load (MW) is taken as the x-axis to draw a daily compressor efficiency degradation curve. The three equation curves in S1 are drawn in it.

[0044] It should be noted that the daily compressor efficiency degradation curve displays two efficiency curves before and after water washing of the compressor as the upper and lower lines of the interval, wherein the efficiency curve before water washing is the lower line, and the efficiency curve after water washing is the upper line, and then the daily form of the actual efficiency curve is drawn in the curve.

[0045] As Figure 3 and Figure 4 , the present application calculates 1440 compressor efficiency average values and gas turbine power generation load average values for each day.

[0046] To reflect the influence of the change of the efficiency of the No. 9 compressor on the gas consumption, as shown in Figure 5 , the gas consumption increases by 429.48 Nm 3 / h for each 1% decrease in compressor efficiency.

[0047] It should also be noted that the historical data of the unit operation from the water washing after the water washing of the compressor to the next water washing is screened, and time, unit load, and compressor efficiency are taken as inputs to import the LSTM training model. After completing the model training, the model is deployed on the platform, and real-time data of the unit operation is input. The prediction output data time length dp and the model calculation input data length dn are set. The platform records the time t0 of the last water washing, the current time tx, and the monitoring of the unit operation state to obtain the unit operation time tn after the last water washing. According to the field experience, the water washing interval is about 600 hours, dp is set to 600 h, and dn = 150 h. The LSTM model is operated as follows every 1 hour:

[0048] 1. When the time tn after the last water washing is less than the model calculation input data length dn, the model prediction calculation is not performed, and the prediction result is directly output, and the water washing prediction time Ts = dp-dn.

[0049] 2. When the time elapsed since the last water wash (tn) is greater than the length of the input data (dn) but less than dp, run the LSTM model, using the operating data within the most recent dn time elapsed as input, to predict the unit operating data for the next dp time elapsed. Slice the predicted output data by hour, obtaining dp arrays, each containing the unit load and compressor efficiency for the corresponding hour. Calculate the difference between each array and the previously obtained efficiency upper limit, and calculate the increase in operating gas caused by the decrease in compressor efficiency based on fitting experience, converting it into a continuous economic loss (Bn). Then, obtain the single cost (Bo) of the on-site water wash operation. Within the time interval [last water wash time, current time + dp], assuming a water wash operation is performed at a certain time (t), calculate (Bn + Bo) / t to calculate the total cost per unit time. Construct a boundary optimization problem to obtain the minimum total cost per unit time, obtain the time scale (t) corresponding to the minimum value, and output the predicted water wash time (Ts = t).

[0050] 3. If the time tn after the last wash is greater than the predicted output data length dp, no model prediction calculation is performed, and the prediction result is directly output, Ts=0.

[0051] The final front-end display shows the recommended wash duration Ts, which is also used for subsequent warning logic.

[0052] Furthermore, boundary optimization problems include:

[0053] ① (Last wash time) At present The compressor efficiency array output by the LSTM prediction model within the time range of [+prediction length dp]. , ,..., ], calculate the difference between it and the maximum compressor efficiency line, and convert it into a continuous economic loss array according to the efficiency-cost fitting formula. , ,..., ];

[0054] ② For solving This optimization problem, considering the continuous economic loss Bn, should exhibit a trend consistent with the characteristics of compressor performance degradation. As time increases, the rate of cumulative loss also increases, eventually tending towards a deteriorating maximum value, at which point the cumulative loss increment per unit time remains constant. Therefore, based on the analysis, the objective value of this optimization problem is... Generally, it will present a decline and then rise, there is a minimum. But because Bn is only theoretically deteriorate over time, in fact, it is not simply a monotonic function, while the deterioration of time span is long, there may also be data fluctuations, in the vicinity of the theoretical minimum may appear fluctuations in the optimization target, resulting in several approximate minimum. Therefore, the curve of the objective function is divided into three stages: the decline, fluctuations, rising. To simplify the optimization calculation process, reduce the consumption of computing resources, iterative analysis of the process from the decline to the fluctuations, fluctuations from the approximate range of the minimum value of the objective function is selected;

[0055] ③From start, calculate = ;

[0056] ④For simplicity, skip the array sequence k-1 data, calculate = ;

[0057] ⑤Judge and difference, when - > , it is determined that it is still in the decline, continue to iterate and skip k-1 data, calculate - ,..., - , indicates the iteration threshold;

[0058] ⑥If - < , suspect it into the fluctuations, further verification, calculate - > , if true, it is considered that this section is still in the decline, continue to iterate and calculate - , the iteration threshold is also reduced to , if there is a difference less than the threshold in the subsequent iteration, the threshold is reduced to , and so on;

[0059] ⑦Set iteration termination condition parameter z, when there is a segment - | , it is determined that the objective function has left the decline and entered the fluctuations, and the iteration is stopped;

[0060] ⑧After entering the fluctuations, calculate , ,.... The minimum value is selected from the array as the solution of the optimization problem.

[0061] S3: Based on the future compressor efficiency degradation state, the cumulative efficiency gas cost and water washing cost are output, the boundary constraint optimization problem is constructed, the water washing remaining time recommendation is output, and the water washing warning is triggered according to the water washing remaining time and the compressor efficiency degradation state.

[0062] Further, the water washing warning is triggered according to the recommended water washing time and the efficiency degradation value, and the triggering logic has two:

[0063] 1. When the recommended water washing remaining time is greater than 0 and any two of the degradation difference values calculated at 180MW, 200MW and 220MW are less than 0, the water washing warning is triggered.

[0064] That is, three degradation values are calculated first:

[0065] The degradation difference value at 180MW is represented as the efficiency at 180MW minus the lower limit of the efficiency at 180MW.

[0066] The degradation difference value at 200MW is represented as the efficiency at 200MW minus the lower limit of the efficiency at 200MW.

[0067] The degradation difference value at 220MW is represented as the efficiency at 220MW minus the lower limit of the efficiency at 220MW.

[0068] A. If there are three load points on the same day, any two of the degradation difference values are less than 0, the front-end water washing warning is triggered.

[0069] B. If there are only two of the three load points on the same day, any one of the degradation difference values is less than 0, the front-end water washing warning is triggered.

[0070] C. If there is only one of the three load points on the same day, the degradation difference value is less than 0, the front-end water washing warning is triggered.

[0071] 2. When the recommended water washing remaining time is less than 0 and any one of the degradation difference values calculated at 180MW, 200MW and 220MW is less than 0, the water washing warning is triggered.

[0072] Figure 6 、 Figure 7 The daily compressor efficiency degradation curve graph and the compressor water washing remaining time recommendation graph are drawn respectively.

[0073] In embodiment 2, one embodiment of the present application provides a compressor performance degradation analysis and water washing recommendation system based on an LSTM neural network and a boundary constraint optimization, which comprises an initialization module, an algorithm fitting module, an LSTM model construction module and an LSTM model prediction module.

[0074] The initialization module is used to collect operation history data, the algorithm fitting module is used to fit the load-efficiency upper and lower limit equation based on a unary quadratic polynomial fitting algorithm, the output per minute compressor efficiency and unit load average fit the daily load-efficiency equation, the LSTM model construction module is used to draw a daily compressor efficiency degradation curve, train an LSTM neural network, and predict future compressor efficiency degradation state, and the LSTM model prediction module is used to output water washing remaining time recommendation and trigger water washing warning according to the water washing remaining time and the compressor efficiency degradation state.

[0075] The embodiment also provides a computer device, including a memory and a processor, the memory stores a computer program, and the processor realizes the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization as proposed in the above embodiment when executing the computer program.

[0076] The embodiment also provides a computer readable storage medium, which stores a computer program, and the computer program realizes the compressor performance degradation analysis and water washing recommendation method based on the LSTM neural network and boundary constraint optimization as proposed in the above embodiment when executed by a processor.

[0077] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or parts of the present application that essentially contribute to the prior art or the technical solutions of the present application can be embodied in the form of software products. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device) to execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0078] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logical functions and can be specifically embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport the program for use by or in connection with the instruction execution system, apparatus or device or in conjunction with these instruction execution systems, apparatus or devices.

[0079] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer readable medium can be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example via an optical scanner, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.

[0080] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example in software or firmware, stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0081] It should be noted that the above examples are merely intended to illustrate the technical solutions of the present application and not to limit the same, and although the present application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all such modifications or replacements should be included in the scope of the claims of the present application.

Claims

1. A method for compressor performance degradation analysis and water washing recommendation based on LSTM neural network and boundary constraint optimization, characterized in that, include: Collect historical operating data, fit the load-efficiency upper and lower limit equations based on the data before and after water washing, and output the compressor efficiency per minute and the average unit load to fit the daily load-efficiency equation. The process of fitting the load-efficiency upper and lower limit equations based on data before and after water washing includes using a univariate quadratic polynomial fitting algorithm to fit a univariate quadratic equation for the compressor efficiency before water washing, with the average compressor efficiency as the dependent variable and the average gas turbine power generation load as the independent variable. Using a univariate quadratic polynomial fitting algorithm, a univariate quadratic equation is fitted with the average compressor efficiency as the dependent variable and the average gas turbine power generation load as the independent variable for the water-washed compressor efficiency. The process of fitting the load-efficiency upper and lower limit equation based on data before and after water washing includes using a univariate quadratic polynomial fitting algorithm to fit a univariate quadratic equation for the daily compressor efficiency with the average compressor efficiency as the dependent variable and the average gas turbine power generation load as the independent variable. Based on the load-efficiency equation, a compressor efficiency degradation curve is plotted, and an LSTM neural network is trained to predict the future compressor efficiency degradation state. Based on the cumulative efficiency, gas cost, and water washing cost output under the future compressor efficiency degradation state, a boundary constraint optimization problem is constructed to output a recommendation for the remaining water washing time. A water washing warning is triggered based on the remaining water washing time and the compressor efficiency degradation state. Based on the prediction results of the LSTM neural network, the compressor efficiency reduction is converted into the unit's gas loss. On-site water washing operation cost data is collected. A boundary optimization problem is constructed with the cumulative gas loss and water washing operation cost. By minimizing the total cost per unit time, the optimal remaining water washing time is output. Calculate the compressor efficiency degradation difference of the unit load on the day, and make a logical judgment based on the remaining water washing time to trigger a water washing warning.

2. The compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization as described in claim 1, characterized in that: The collected historical data includes, The system collects effective historical operating data of the compressor before and after water washing, including compressor efficiency and gas turbine power generation load, from the SIS system. Output the average compressor efficiency and the average gas turbine power generation load every minute, with a 1-minute interval; Daily operating data of the compressor is obtained from the SIS system, and each data point includes compressor efficiency and gas turbine power generation load. Output the average compressor efficiency and the average gas turbine power generation load every minute, with a 1-minute interval.

3. The compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization as described in claim 1 or 2, characterized in that: The method of plotting the compressor efficiency degradation curve based on the load-efficiency equation includes... Plot the daily compressor efficiency degradation curve with compressor efficiency as the y-axis and gas turbine power generation load as the x-axis, and draw the equation curve.

4. The compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization as described in claim 3, characterized in that: The trained LSTM neural network predicts future compressor efficiency degradation states, including... Model training is performed using an LSTM neural network; Perform time-series forecasting on real-time power plant data and output an array of time-load-efficiency.

5. A compressor performance degradation analysis and water washing recommendation system based on LSTM neural network and boundary constraint optimization, employing the compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization as described in any one of claims 1 to 4, characterized in that: It includes an initialization module, an algorithm fitting module, an LSTM model building module, and an LSTM model prediction module; The initialization module is used to collect historical operation data; The algorithm fitting module is used to fit the load-efficiency upper and lower limit equations based on the univariate quadratic polynomial fitting algorithm, and output the compressor efficiency per minute and the average unit load to fit the daily load-efficiency equation. The LSTM model building module is used to plot the daily compressor efficiency degradation curve, train the LSTM neural network, and predict the future compressor efficiency degradation state. The LSTM model prediction module is used to output a recommendation for the remaining water washing time, and to trigger a water washing warning based on the remaining water washing time and the compressor efficiency degradation status.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the compressor performance degradation analysis and water washing recommendation method based on LSTM neural network and boundary constraint optimization as described in any one of claims 1 to 4.

Images