A unit load section control system and method, electronic equipment and storage medium

Abstract

The present application belongs to the technical field of unit optimization control, and discloses a unit load section control system and method, an electronic device and a storage medium; the method comprises: collecting working condition analysis data of a unit in real time; processing the working condition analysis data to obtain processed data; analyzing the processed data to evaluate the unit operation condition; collecting unit characteristic parameters in real time; optimizing and adjusting the unit characteristic parameters according to the unit operation condition; predicting the state transition time according to the optimized and adjusted unit characteristic parameters; formulating an equipment start-stop optimization strategy according to the state transition time to intelligently control the equipment start-stop; the present application adopts a digital and data-driven mode for intelligent control, effectively solves the control problems in the unit load section, thereby improving the operation efficiency and performance of the unit, and further improving the operation safety and control robustness of the unit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unit optimization control, and more particularly, to a unit load section control system and method, an electronic device and a storage medium. BACKGROUND

[0002] With the growing demand for power systems and the trend of renewable energy access to the grid, the flexibility and adjustment capability of thermal power unit load are more important for the safe and stable operation of power systems. The dynamic characteristics of thermal power units under different load conditions (such as dry state, wet state, etc.) are quite different. During the dry-wet state conversion of thermal power units, the original AGC coordinated control system fails due to changes in the structure of the controlled object, and the start-stop of various devices causes strong disturbances. Therefore, during the dry-wet state conversion of thermal power units, there are large fluctuations in parameters such as steam temperature, steam pressure, and separator water level, and the conversion time is long, which makes it impossible to guarantee the safety of thermal power unit operation.

[0003] Chinese patent application No. CN112462604A discloses a supercritical unit dry-wet state modeling and coordinated control method based on a genetic algorithm. The method includes the following steps: using a step excitation signal to obtain object input and output data, and in the parameter identification of a continuous system, the initial state of the object is also used as a parameter identification item. Through iterative calculation of the least squares method, the model parameters are obtained, and the linear model of each load point of the unit coordinated control system under dry-wet state is established. For the rapid tracking task of large-scale variable load operation of the unit, with the help of H∞ control theory, a scheme based on H∞ loop shaping design controller is given, and through PID approximate model reduction, the large-scale variable load rapid tracking of large thermal power units is realized, and under the premise of retaining the main characteristics of the original controller, it is easier to implement in engineering.

[0004] The above technology realizes the automatic control of thermal power units through a genetic algorithm. However, if the initial population selection of the genetic algorithm is improper, and the operation parameters of crossover and mutation are not properly set, the final obtained result will fall into local optimum, and the global optimal solution cannot be found. In addition, the strong disturbance during the dry-wet state conversion of thermal power units is not considered in the above technology, which leads to poor parameter optimization effect and unstable control effect, thereby affecting the operation safety and robustness.

[0005] In view of this, the present application provides a unit load section control system, method, electronic device and storage medium to solve the above problems. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purposes, the present application provides the following technical scheme: a unit load section control method, comprising:

[0007] Real-time acquisition of working condition analysis data of the unit;

[0008] Processing the working condition analysis data to obtain processing data;

[0009] Analyzing the processing data to evaluate the unit operating condition;

[0010] Real-time collection of unit characteristic parameters;

[0011] Optimizing and adjusting the unit characteristic parameters according to the unit operating condition;

[0012] Predicting the state transition time according to the optimized and adjusted unit characteristic parameters;

[0013] Formulating a device start-stop optimization strategy according to the state transition time to intelligently control the device start-stop.

[0014] Further, the working condition analysis data includes power data and working condition characteristic data; the power data includes actual output power and rated power; the working condition characteristic data is data whose numerical fluctuation exceeds a corresponding preset fluctuation threshold during the unit dry-wet state transition process; the working condition analysis data is collected according to a preset collection frequency;

[0015] The processing data includes a load rate and a data change rate;

[0016] The load rate is obtained according to the ratio of the actual output power to the rated power;

[0017] The data change rate is the change rate of each data in the working condition characteristic data;

[0018] The data change rate is obtained by subtracting each data in the working condition characteristic data collected last time from each data in the real-time collected working condition characteristic data to obtain a data difference value of each data in the working condition characteristic data, and then obtaining the change rate of each data in the working condition characteristic data according to the ratio of the data difference value of each data in the working condition characteristic data to the corresponding data in the working condition characteristic data collected last time.

[0019] Further, the method for analyzing the processing data includes:

[0020] A preset load range value , ;

[0021] Comparing the load rate with the load range value ;

[0022] If , a preliminary conversion instruction is generated;

[0023] If or then no preliminary conversion instruction is generated;

[0024] a preset threshold set, the threshold set including a plurality of rate of change thresholds , , the rate of change thresholds corresponding to data in the operating condition characteristic data one by one;

[0025] each of the rates of change of data is divided by a corresponding rate of change threshold respectively to obtain a standardized rate of change of each data in the operating condition characteristic data, data with a standardized rate of change greater than 1 is marked as changed data, the number of the changed data is counted and marked as a change number ; each standardized rate of change is sequentially added to obtain a rate of change coefficient ;

[0026] if , a preliminary conversion instruction is generated, indicated as and

[0027] if , no preliminary conversion instruction is generated, indicated as or

[0028] Further, the method for evaluating the operating condition of the unit includes:

[0029] if two preliminary conversion instructions are generated, it is evaluated that the operating condition of the unit is in a dry-wet state conversion;

[0030] if only one preliminary conversion instruction is generated or no preliminary conversion instruction is generated, it is evaluated that the operating condition of the unit is not in a dry-wet state conversion.

[0031] Further, the unit characteristic parameter is a parameter that has a direct impact on the operating condition characteristic data when the unit characteristic parameter is self-regulated;

[0032] The step of optimizing the unit characteristic parameter includes:

[0033] Step 1: preset a parameter adjustment range, the parameter adjustment range including an adjustment range of each parameter in the unit characteristic parameter;

[0034] Step 2: construct a parameter set, an integer greater than 1, each parameter set being constructed by randomly selecting a numerical value from each adjustment range, each of the parameter sets being different; different digital labels are set for each different parameter set and marked as a set label, the range of the set label being ;​

[0035] Step 3: Use the improved clone selection algorithm to The best parameter set is selected from the parameter sets and marked as the best set, and the unit characteristic parameters are optimized and adjusted according to the best set.

[0036] Furthermore, the improved clone selection algorithm is used to The steps of selecting the best parameter set from the parameter sets include:

[0037] Step a: Randomly generate an initial antibody population of size q ; Preset iteration threshold; Among them, is the qth antibody, ;

[0038] Step b: Determine the fitness function;

[0039] Step c: Calculate the initial antibody population The fitness of each antibody in The antibody with the largest fitness and construct a memory population , number of cycles , ;

[0040] Step d: Memory population Each antibody in the cloning is cloned according to the calculated clone size to generate a clonal antibody population ,in is the qth cloned antibody in the clonal antibody population;

[0041] Step e: Cloning the antibody population Each antibody in the cloud uses the cloud adaptive mutation operator to perform cloud adaptive mutation to generate a mutant antibody group ;

[0042] Step f: Cloning the Antibody Population Perform antibody recombination to generate recombinant antibody populations ; and with the variant antibody group Merge to generate a pooled antibody population ;

[0043] Step g: Combined antibody population Use random replacement algorithm;

[0044] Step h: Calculate the dynamic boundary, generate the inverse solution corresponding to each antibody and add it to the merged antibody group middle;

[0045] Step i: If the number of cycles t is less than the iteration threshold, let , get a new memory population , loop step d to step h; if the number of iterations t is equal to the iteration threshold, the loop ends and the combined antibody group is obtained The set label corresponding to the antibody with the largest fitness is obtained according to the set label.

[0046] Furthermore, in step a, the initial antibody group The antibodies in the set correspond to the set labels one by one, and the range of the antibodies is the range of the set labels. ;

[0047] In step b, the fitness function is expressed as: ;

[0048] Where, is the fitness of the first antibody, is the matching degree of the first antibody, ;

[0049] Methods for obtaining the matching degree include:

[0050] According to the set label corresponding to the antibody, the corresponding parameter set is obtained; the parameter set, the unit characteristic parameter and one of the operating condition characteristic data are taken as a set of analysis data, and a total of Group analysis data, ;Will The group analysis data are input into the corresponding trained data prediction model respectively to predict the data in the corresponding working condition characteristic data, that is, the input and output of a data prediction model are the same data in the working condition characteristic data; the data prediction model has a total of indivual, , The data prediction model and the working condition characteristic data The data correspond one to one;

[0051] According to the operating condition characteristic data output by the data prediction model and the operating condition characteristic data collected last time, the data change rate is re-obtained and marked as the new data change rate; each change rate in the new data change rate is divided by the corresponding change rate threshold to obtain the new standardized change rate corresponding to each data in the operating condition characteristic data, each new standardized change rate is added in sequence to obtain the new change rate coefficient, and the new change rate coefficient is used as the matching degree.

[0052] Further, The training process of a data prediction model includes:

[0053] The data prediction model whose input and output are both steam temperature is labeled as steam temperature prediction model. The training process of the steam temperature prediction model includes:

[0054] Collect e sets of analysis data corresponding to steam temperatures in advance, where e is an integer greater than 1, and convert the analysis data and the corresponding steam temperatures into a corresponding set of feature vectors;

[0055] Each set of eigenvectors is used as input to a steam temperature prediction model. The steam temperature prediction model uses a set of predicted steam temperatures corresponding to each set of analysis data as output, and uses the actual steam temperature corresponding to each set of analysis data as a prediction target. The actual steam temperature is the pre-collected steam temperature corresponding to the analysis data. The prediction results are evaluated for model accuracy using the mean absolute percentage error (MAPE). When the calculated MAPE is less than the preset MAPE, the steam temperature prediction model training is completed. A steam temperature prediction model is generated that predicts the steam temperature based on the analysis data. The steam temperature prediction model is a deep belief network model.

[0056] the remaining The training process of the data prediction model is consistent with that of the steam temperature prediction model, and both are deep belief network models.

[0057] Furthermore, in step d, the cloned antibody group ,in The qth cloned antibody in the clonal antibody population; the calculation method of the clone size includes:

[0058] ;

[0059] ;

[0060] Where, is the clone size of the first antibody, Int is the rounding function, q is the size of the antibody group, is the affinity of the first antibody, min is the minimum function, exp is the natural exponential function, is the Euclidean distance between the I-th antibody and the J-th antibody, , ;

[0061] Furthermore, in step f, from the cloned antibody group Randomly select b antibodies as parent antibodies for antibody recombination to generate offspring antibodies , until the clonal antibody population When the number of unselected antibodies is less than b, antibody recombination is stopped and a recombinant antibody group is generated. ; Progeny antibodies The calculation methods include:

[0062] ;

[0063] Where, is the b-th parent antibody, is the bth scale factor, the scale factor is a randomly generated real number and not all b scale factors are 0;

[0064] In step g, d new antibodies are randomly generated and the antibody group is merged. The d antibodies with the smallest fitness are replaced by d new antibodies generated randomly. .

[0065] Furthermore, in step h, the dynamic boundary is , obtain the combined antibody group The set labels corresponding to the antibody with the smallest fitness and the set labels corresponding to the antibody with the largest fitness are compared. is a collection label with a smaller value, is a collection label with a larger value;

[0066] Methods for generating inverse solutions include:

[0067] ;

[0068] Where, is the reverse solution of the first antibody, for The random number in is the collection label of the first antibody;

[0069] If the generated reverse solution is greater than or less than the dynamic boundary, it is marked as a transcendental solution, and a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution;

[0070] In step i, the reverse solution of the antibody is also used as the antibody to calculate the combined antibody group The fitness of all antibodies in the pool is calculated and sorted from large to small, retaining the top q antibodies; select the combined antibody group The most adaptable antibodies, forming a new memory population .

[0071] Furthermore, the transition time is the duration of the dry-wet state conversion process of the unit;

[0072] The method for predicting transition time comprises:

[0073] Obtain the predicted operating condition data corresponding to the best set and mark it as the best operating condition data. The predicted operating condition data is the operating condition characteristic data output by the data prediction model.

[0074] Based on the optimal operating condition data and the last collected operating condition characteristic data, the optimal change rate is calculated. The optimal change rate includes the change rate of each data in the optimal operating condition data. The load rate, optimal change rate and optimized unit characteristic parameters are used as test data. The test data is input into the trained time prediction model to predict the corresponding transition time.

[0075] The training process of the time prediction model includes:

[0076] Pre-collect the transition times corresponding to f groups of test data, where f is an integer greater than 1, and convert the test data and the corresponding transition times into a corresponding set of feature vectors;

[0077] Each set of feature vectors is used as the input of a time prediction model. The time prediction model outputs a set of predicted transition times corresponding to each set of test data, and uses the actual transition time corresponding to each set of test data as the prediction target. The actual transition time is the pre-collected transition time corresponding to the test data. Minimizing the sum of the prediction errors of all test data is used as the training goal. The time prediction model is trained until the sum of the prediction errors reaches convergence, and training is stopped. The time prediction model is a deep neural network model.

[0078] Furthermore, the step of formulating the equipment start-stop optimization strategy includes:

[0079] Step A: Preset the start-stop sequence table. In the start-stop sequence table, the equipment is sorted according to the start-stop sequence of the equipment. The equipment is the equipment that needs to be started or stopped during the dry-wet state conversion process of the unit;

[0080] Step B: Preset a time interval and divide the transition time into H time points according to the time interval;

[0081] Step C: Construct N point label sets based on H time points;

[0082] Step D: Randomly select a set of point labels and mark it as the selected set, and calculate the interference degree of the selected set;

[0083] Step E: Return to step D until all N point label sets are marked as selected sets, then do not return to step D and proceed to step F;

[0084] Step F: Sort the interference levels of the N point label sets from small to large to generate a first sorting table, obtain the interference level at the top of the first sorting table and mark it as the minimum interference level, obtain the point label set corresponding to the minimum interference level and mark it as the optimal point label set;

[0085] Step G: According to the optimal point label set and the start-stop sequence table, set the corresponding start-stop time for each device.

[0086] Furthermore, in step C, a corresponding digital label is set for each time point in the transition time and marked as a point label. The point label range is ; According to the start-stop sequence table, count the number of devices M, ; Filter the point labels, filter out M point labels from the point label range each time, and build a point label set, , , .

[0087] Furthermore, in step D, the expression of interference degree is: ;

[0088] Where, is the degree of interference, For the The amount of interference to the i-th data in the working condition characteristic data when the equipment is started or stopped, ;

[0089] The method for obtaining the interference amount on the working condition characteristic data when each device is started and stopped includes:

[0090] Set a different digital label for each device and mark it as the device label. The range of the device label is ; Take a device tag and one of the optimized and adjusted working condition characteristic data as a set of evaluation data, and obtain Group assessment data, ;Will The group evaluation data are input into the trained data analysis model to predict the corresponding interference amount;

[0091] The training process of the data analysis model is consistent with that of the time prediction model, and both are deep neural network models.

[0092] Furthermore, in step G, the point labels in the optimal point label set are sorted from small to large to generate a second sorting table; the point labels in the second sorting table are matched one by one with the devices in the start-stop sequence table in positive order, and the corresponding time point is obtained according to the point label corresponding to each device. The time point corresponding to each device is the start-stop time corresponding to each device.

[0093] Further, the processed data is analyzed to identify the unit conversion status;

[0094] The unit conversion state includes dry-wet conversion and wet-dry conversion; two start-stop sequence tables are preset;

[0095] Methods for identifying the unit transfer status include:

[0096] The load rate obtained in real time is subtracted from the load rate obtained last time to obtain the load rate difference; the load rate obtained last time is the load rate obtained after processing the working condition analysis data collected last time;

[0097] The load factor difference and data change rate are used as research data, and the research data is input into the trained state recognition model. The corresponding recognition results are obtained based on the output recognition labels, and the unit conversion state is determined based on the recognition results; the corresponding start-stop sequence table is obtained based on the unit conversion state;

[0098] The training process of the state recognition model includes:

[0099] Pre-setting corresponding recognition results for g groups of research data, where g is an integer greater than 1, and the recognition results include dry-wet conversion and wet-dry conversion, with different digital labels set for both dry-wet conversion and wet-dry conversion; marking the digital labels of the recognition results as recognition labels, and converting the research data and the corresponding recognition labels into a corresponding set of feature vectors;

[0100] Each set of feature vectors is used as the input of the state recognition model. The state recognition model takes a set of predicted identification labels corresponding to each set of research data as output, and the actual identification labels corresponding to each set of research data as prediction targets. The actual identification labels are pre-set digital labels of the identification results corresponding to the research data. Minimizing the sum of the prediction errors of all research data is used as the training goal. The state recognition model is trained until the sum of the prediction errors reaches convergence and training is stopped. The state recognition model is a deep neural network model.

[0101] A unit load segment control system, used to implement the unit load segment control method, comprising:

[0102] Data acquisition module, used to collect real-time operating condition analysis data of the unit;

[0103] A data processing module is used to process the working condition analysis data and obtain the processed data;

[0104] The operating condition evaluation module is used to analyze the processed data and evaluate the unit operating conditions;

[0105] Parameter collection module, used to collect unit characteristic parameters in real time;

[0106] Parameter optimization module, used to optimize and adjust the unit characteristic parameters according to the unit operating conditions;

[0107] The time prediction module is used to predict the transition time based on the optimized and adjusted unit characteristic parameters;

[0108] The start-stop control module is used to formulate equipment start-stop optimization strategies based on transition time and intelligently control equipment start-stop.

[0109] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a method for controlling a load segment of a unit is implemented.

[0110] A computer-readable storage medium stores a computer program, and when the computer program is executed, the method for controlling a load segment of a unit is implemented.

[0111] The technical effects and advantages of the unit load segment control system, method, electronic equipment and storage medium of the present invention are as follows:

[0112] 1. Accurately judge the unit status by collecting and processing operating condition analysis data; use the improved clone selection algorithm to optimize and adjust the unit characteristic parameters, so as to find the global optimal solution of the unit characteristic parameters and achieve accurate adjustment of the unit characteristic parameters; at the same time, considering the strong disturbance caused by equipment start-up and shutdown, use deep learning technology to establish a time prediction model to predict the transition time, formulate equipment start-up and shutdown optimization strategies based on the transition time, and intelligently control the start-up and shutdown of equipment to resist the strong disturbance caused by equipment start-up and shutdown, and improve the optimization effect of the unit characteristic parameters; use digital and data-driven methods for intelligent control to effectively solve the control problems in the load section of the unit, thereby improving the operating efficiency and performance of the unit, and then improving the operating safety and control robustness of the unit.

[0113] 2. By calculating the load rate difference and combining it with the data change rate, deep learning technology is used to scientifically identify the unit's conversion status, automatically select a reasonable start-stop sequence table, and achieve targeted equipment start-stop optimization; it can better adapt to different conversion situations in actual operation, thereby achieving more intelligent and precise load segment control, further improving system stability and control accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] Figure 1 This is a schematic diagram of a load section control system for a unit according to embodiment 1 of the present invention;

[0115] Figure 2 This is a flow chart of the method for obtaining the optimal parameter set according to embodiment 1 of the present invention;

[0116] Figure 3 This is a flow chart of a method for formulating a device start-stop optimization strategy according to Example 1 of the present invention;

[0117] Figure 4 This is a schematic diagram of a load section control system for a unit according to embodiment 2 of the present invention;

[0118] Figure 5 This is a flow chart of a unit load section control method according to embodiment 3 of the present invention;

[0119] Figure 6 This is a schematic diagram of an electronic device according to embodiment 4 of the present invention;

[0120] Figure 7 This is a schematic diagram of the storage medium of Example 5 of the present invention. DETAILED DESCRIPTION

[0121] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0122] Example 1

[0123] See also Figure 1 As shown, the load segment control system of a unit described in this embodiment includes a data acquisition module, a data processing module, an operating condition evaluation module, a parameter collection module, a parameter optimization module, a time prediction module and a start-stop control module; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0124] The data acquisition module is used to collect the unit's operating condition analysis data in real time.

[0125] Working condition analysis data includes power data and working condition characteristic data;

[0126] Power data includes actual output power and rated power; the actual output power is obtained by the power sensor installed at the generating terminal of the unit; the rated power is obtained according to the equipment specification sheet or technical documents of the unit; the actual output power and rated power can reflect the load condition of the unit to determine whether the unit is performing dry-wet state conversion; the unit is a thermal power unit.

[0127] The operating condition characteristic data is data whose numerical fluctuation exceeds the corresponding preset fluctuation threshold during the dry-wet state conversion process of the unit. The operating condition characteristic data include steam temperature, steam pressure, separator water level, etc. The fluctuation threshold is obtained by technical personnel in this field when the unit is in the dry or wet state. Various data of the unit are measured multiple times, and the maximum value corresponding to each data is subtracted from the minimum value to obtain the fluctuation threshold of each data; the steam temperature is obtained by a thermocouple sensor installed above the steam cylinder in the boiler; the steam pressure is obtained by a pressure transmitter installed in the steam cylinder outlet pipeline in the boiler; the separator water level is obtained by a capacitive or ultrasonic water level meter installed inside the separator.

[0128] The operating condition analysis data is collected according to a preset collection frequency. The collection frequency is preset by technical personnel in this field based on the change rate of the operating condition analysis data and the conversion time between the dry and wet states of the unit, combined with actual experience.

[0129] The data processing module is used to process the working condition analysis data and obtain the processed data.

[0130] Processing data includes load rate and data change rate;

[0131] The load factor is obtained by the ratio of actual output power to rated power. The load factor expression is: Where, is the load rate, is the actual output power, is the rated power.

[0132] The data change rate is the change rate of each data in the working condition characteristic data;

[0133] The method for obtaining the data change rate is as follows: subtract the corresponding data in the last collected working condition characteristic data from each data in the real-time working condition characteristic data to obtain the data difference of each data in the working condition characteristic data; then, according to the ratio of the data difference of each data in the working condition characteristic data to the corresponding data in the last collected working condition characteristic data, obtain the change rate of each data in the working condition characteristic data; the expression of the data change rate is: Where, is the rate of change of the i-th data in the working condition characteristic data, is the i-th data in the real-time collected working condition characteristic data, is the i-th data in the last collected working condition characteristic data, , is the number of data in the working condition characteristic data, is an integer greater than 1.

[0134] The operating condition evaluation module is used to analyze the processed data and evaluate the operating condition of the unit.

[0135] Methods for analyzing processed data include:

[0136] Preset load range value , The load range value is preset in the following way: when the dry and wet operating conditions of the historical unit are switched, the technical personnel collect and calculate the load rate of the unit multiple times, and obtain multiple load rates each time; among the multiple load rates obtained at one time, the load rate with the largest value is marked as the maximum load, and the load rate with the smallest value is marked as the minimum load; the average of the multiple maximum loads obtained is taken as the maximum value of the load range. The mean value of the minimum load obtained multiple times is taken as the minimum value of the load range .

[0137] The load rate is compared with the load range value ;

[0138] If , a preliminary conversion instruction is generated, indicating that the load rate of the unit is in the load rate corresponding to the dry-wet state conversion of the unit, and the unit may be in dry-wet state conversion;

[0139] If or , no preliminary conversion instruction is generated, indicating that the load rate of the unit is not in the load rate corresponding to the dry-wet state conversion of the unit, and the unit is not in dry-wet state conversion.

[0140] A set of preset threshold values is set, including a change rate threshold value , , The change rate threshold value corresponds to the data in the operating condition characteristic data one by one; the threshold set is obtained by calculating the change rate of each data in the operating condition characteristic data multiple times by a person skilled in the art during the historical unit dry-wet state operating condition conversion, and taking the mean value of the multiple change rates corresponding to each data as the change rate threshold value of the corresponding data, and sequentially obtaining the change rate threshold value corresponding to each data in the operating condition characteristic data.

[0141] Each change rate in the data change rate is divided by the corresponding change rate threshold value , to obtain the standardized change rate of each data in the operating condition characteristic data, ; the data with a standardized change rate greater than 1 is marked as a change data, the number of change data is counted and marked as a change number ; each standardized change rate is sequentially added to obtain a change rate coefficient ;

[0142] If , a preliminary conversion instruction is generated, indicated as and; it is indicated that there are more data in the operating condition characteristic data with large change amplitudes, and the unit may be in dry-wet state conversion;

[0143] If , no preliminary conversion instruction is generated, indicated as or; it is indicated that there are fewer data in the operating condition characteristic data with large change amplitudes, and the unit is not in dry-wet state conversion.

[0144] The method for evaluating the operating condition of the unit comprises:

[0145] If two preliminary conversion instructions are generated, the unit operating condition is assessed to be in dry-wet state conversion;

[0146] If only one preliminary conversion instruction is generated or no preliminary conversion instruction is generated, it is assessed that the unit operating condition is not in the dry-wet state conversion.

[0147] It should be noted that the reason for first judging twice whether a preliminary conversion instruction is generated and then evaluating the unit operating condition is that if the unit operating condition is evaluated only based on the load rate, the load rate of the unit may change due to fluctuations in grid demand or changes in dispatch instructions, but the unit operating condition is not in the dry-wet state conversion, resulting in a misjudgment; similarly, if the unit operating condition is evaluated only based on the data change rate, the operating condition characteristic data may fluctuate significantly due to routine operations, such as adjusting fuel supply, boiler feed water, etc., or changes in external environmental factors, and the unit operating condition is not in the dry-wet state conversion, resulting in a misjudgment; therefore, it is necessary to combine the load rate with the data change rate to accurately evaluate the unit operating condition.

[0148] Parameter collection module, used to collect unit characteristic parameters in real time.

[0149] When the unit characteristic parameters are self-regulated, they will have a direct impact on the operating characteristic data. The unit characteristic parameters include spray opening (the opening size of the nozzle in the spray system), water supply control valve opening (the opening size of the control valve on the water supply pipe in the water supply system), water level control valve opening (the opening size of the control valve inside the boiler in the water supply system), etc.; the spray opening is obtained by the photoelectric or piezoelectric opening sensor installed on the nozzle; the water supply control valve opening is obtained by the contactless electromagnetic opening sensor integrated in the valve brake; the water level control valve opening is directly measured by the brake position sensor integrated in the valve brake, which measures the rotational position of the valve stem.

[0150] The parameter optimization module is used to optimize and adjust the unit characteristic parameters according to the unit operating conditions.

[0151] The steps for optimizing the unit characteristic parameters include:

[0152] Step 1: Preset the parameter adjustment range, which includes the adjustment range of each parameter in the unit characteristic parameters; the parameter adjustment range is obtained according to the equipment specification sheet or technical documents of the unit.

[0153] Step 2: Build A set of parameters, is an integer greater than 1. Each parameter set is constructed by randomly selecting a value from each adjustment range. Each parameter set is different; set a different numerical label for each different parameter set and mark it as a set label. The range of the set label is .

[0154] Step 3: Use the improved clone selection algorithm to The best parameter set is selected from the parameter sets and marked as the best set, and the unit characteristic parameters are optimized and adjusted according to the best set.

[0155] See also Figure 2 As shown, the improved clone selection algorithm is used from The steps of selecting the best parameter set from the parameter sets include:

[0156] Step a: Randomly generate an initial antibody population of size q ; Preset iteration threshold, the iteration threshold is pre-set according to the algorithm accuracy; Among them, is the qth antibody, .

[0157] Step b: Determine the fitness function.

[0158] Step c: Calculate the initial antibody population The fitness of each antibody in The antibody with the largest fitness and construct a memory population , number of cycles , .

[0159] Step d: Memory population Each antibody in the cloning is cloned according to the calculated clone size to generate a clonal antibody population ,in It is the qth cloned antibody in the clonal antibody population.

[0160] Step e: Cloning the antibody population Each antibody in the cloud uses the cloud adaptive mutation operator to perform cloud adaptive mutation to generate a mutant antibody group ; The cloud adaptive mutation operator is an existing technology and will not be described in detail here.

[0161] Step f: From the clonal antibody population Randomly select b antibodies as parent antibodies for antibody recombination to generate offspring antibodies , until the clonal antibody population When the number of unselected antibodies is less than b, antibody recombination is stopped and a recombinant antibody group is generated. ; In this embodiment, preferably b is 3; recombinant antibody group Multiple progeny antibodies .

[0162] Step g: Merging variant antibody groups and recombinant antibody groups , generating a pooled antibody population .

[0163] Step h: Randomly generate d new antibodies and merge the antibody groups The d antibodies with the smallest fitness are replaced by d new antibodies generated randomly. .

[0164] Step i: Calculate the dynamic boundary, generate the inverse solution corresponding to each antibody and add it to the merged antibody group middle.

[0165] Step j: If the number of cycles t is less than the iteration threshold, let , get a new memory population , loop step d to step i; if the number of iterations t is equal to the iteration threshold, the loop ends and the combined antibody group is obtained The set label corresponding to the antibody with the largest fitness is obtained according to the set label; among them, For computer language, the value of the iteration number t is added by 1 and then assigned to the iteration number t. The population is remembered in step d. The value of t plus 1 is the new memory population The value of t in .

[0166] In the above step a, the initial antibody group The antibodies in the set correspond to the set labels one by one, and the range of the antibodies is the range of the set labels. .

[0167] In the above step b, the expression of the fitness function is: ;

[0168] Where, is the fitness of the first antibody, is the matching degree of the first antibody, .

[0169] Methods for obtaining the matching degree include:

[0170] According to the set label corresponding to the antibody, the corresponding parameter set is obtained; the parameter set, the unit characteristic parameter and one of the operating condition characteristic data are taken as a set of analysis data, and a total of Group analysis data, ;Will The group analysis data are input into the corresponding trained data prediction model respectively to predict the data in the corresponding working condition characteristic data, that is, the input and output of a data prediction model are the same data in the working condition characteristic data; the data prediction model has a total of indivual, , The data prediction model and the working condition characteristic data The data correspond one to one.

[0171] The training process of a data prediction model includes:

[0172] The data prediction model whose input and output are both steam temperature is marked as steam temperature prediction model. The specific training process of the steam temperature prediction model includes:

[0173] The steam temperatures corresponding to e groups of analysis data are collected in advance, where e is an integer greater than 1, and the analysis data and the corresponding steam temperatures are converted into a corresponding set of feature vectors.

[0174] Each set of eigenvectors is used as the input of the steam temperature prediction model. The steam temperature prediction model uses a set of predicted steam temperatures corresponding to each set of analysis data as output, and the actual steam temperature corresponding to each set of analysis data as the prediction target. The actual steam temperature is the pre-collected steam temperature corresponding to the analysis data. The mean absolute percentage error (MAPE) is used to evaluate the model accuracy of the prediction results. When the calculated MAPE is less than the preset MAPE, the steam temperature prediction model training is completed. The calculation formula of MAPE is: ;in, is the actual steam temperature corresponding to the Lth group of analysis data, is the predicted steam temperature corresponding to the Lth group of analysis data, L is the group number of the characteristic vector corresponding to the analysis data, , R is the number of predicted steam temperatures; a steam temperature prediction model is generated to predict the steam temperature based on the analysis data; wherein the steam temperature prediction model is a deep belief network model, and the preset MAPE is pre-set by those skilled in the art according to the accuracy required by the steam temperature prediction model.

[0175] It should be noted that the steam temperature corresponding to the analysis data is determined by technicians in this field during the historical conversion process of the unit's dry-wet operating conditions. They collect e groups of different analysis data, conduct test experiments under the conditions of each group of analysis data, adjust the unit characteristic parameters in the analysis data according to the corresponding parameter set, and re-collect the steam temperature after the adjustment is completed, and set the corresponding steam temperature for the e groups of different analysis data.

[0176] the remaining The specific training process of the data prediction model is consistent with the training process of the steam temperature prediction model, and both are deep belief network models.

[0177] According to the working condition characteristic data output by the data prediction model and the last collected working condition characteristic data, the data change rate is reacquired and marked as a new data change rate; each change rate in the new data change rate is divided by the corresponding change rate threshold to obtain a new standardized change rate corresponding to each data in the working condition characteristic data, and each new standardized change rate is sequentially added to obtain a new change rate coefficient, which is taken as the matching degree.

[0178] In the step d, the calculation method of the clone size comprises:

[0179] ;

[0180] ;

[0181] In the formula, is the clone size of the Ith antibody, Int is the upward rounding function, q is the antibody group size, is the affinity of the Ith antibody, min is the minimum function, and exp is the natural exponential function, is the Euclidean distance between the Ith antibody and the Jth antibody, , .

[0182] It should be noted that the Euclidean distance is a prior art and will not be described in detail here; when calculating the Euclidean distance between the Ith antibody and the Jth antibody, normalization processing is performed, so that .

[0183] In the step f, the calculation method of the child antibody comprises:

[0184] ;

[0185] In the formula, is the bth parent antibody, is the bth scaling factor, and the scaling factor is a random real number and the b scaling factors are not all 0.

[0186] In the step i, the dynamic boundary is , the set label corresponding to the antibody with the minimum fitness and the set label corresponding to the antibody with the maximum fitness in the merged antibody group are obtained and compared, is the set label with a smaller value, is the set label with a larger value.

[0187] The generation method of the inverse solution comprises:

[0188] ;

[0189] Where, is the reverse solution of the first antibody, for The random number in is the collection label of the first antibody.

[0190] If the generated reverse solution is greater than or less than the dynamic boundary, it is marked as a transcendental solution, and a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution.

[0191] In the above step j, the reverse solution of the antibody is also used as the antibody to calculate the combined antibody group The fitness of all antibodies in the pool is calculated and sorted from large to small, retaining the top q antibodies; select the combined antibody group The most adaptable antibodies, forming a new memory population .

[0192] The time prediction module is used to predict the transition time based on the optimized and adjusted unit characteristic parameters.

[0193] The transition time is the duration of the unit's dry-wet state conversion process.

[0194] Methods for predicting transition time include:

[0195] Obtain the predicted operating condition data corresponding to the best set and mark it as the best operating condition data. The predicted operating condition data is the operating condition characteristic data output by the data prediction model. The predicted operating condition data corresponding to the best set is obtained when calculating the matching degree of the best set in step b above.

[0196] Based on the optimal operating condition data and the operating condition characteristic data collected last time, the optimal change rate is calculated. The optimal change rate includes the change rate of each data in the optimal operating condition data; the load rate, the optimal change rate and the optimized and adjusted unit characteristic parameters are used as test data, and the test data is input into the trained time prediction model to predict the corresponding transition time.

[0197] The specific training process of the time prediction model includes:

[0198] Pre-collect the transition times corresponding to f groups of test data, where f is an integer greater than 1, and convert the test data and the corresponding transition times into a corresponding set of feature vectors;

[0199] Each set of feature vectors is used as the input of the time prediction model. The time prediction model takes a set of predicted transition times corresponding to each set of test data as output, and the actual transition time corresponding to each set of test data as the prediction target. The actual transition time is the transition time corresponding to the test data collected in advance; the training goal is to minimize the sum of the prediction errors of all test data; wherein, the calculation formula of the prediction error is: ,in is the prediction error, is the group number of the feature vector corresponding to the test data, For the The predicted transition time corresponding to the test data set, For the The actual transition time corresponding to the test data of the group is obtained; the time prediction model is trained until the sum of the prediction errors reaches convergence and the training is stopped;

[0200] The above-mentioned time prediction model is specifically a deep neural network model; it includes an input layer, a hidden layer and an output layer; each hidden layer includes multiple neurons, each neuron is connected to the neurons in the next layer, and the connection contains weights, which determine the importance and influence of data transmission in the neural network; each neuron between the hidden layer and the output layer applies an activation function, which maps nonlinearity, allowing the network to learn more complex patterns and features.

[0201] It should be noted that the transition time corresponding to the test data is determined by technical personnel in this field. They collect f groups of different test data during the historical conversion process of dry-wet operating conditions of the unit, conduct test experiments under the conditions of each group of analysis data, collect the duration of the unit's dry-wet state conversion process multiple times under the conditions of one group of analysis data, and use the average of the multiple collected durations as the transition time of the corresponding analysis data. Corresponding transition times are set for the f groups of different test data in turn.

[0202] The start-stop control module is used to formulate equipment start-stop optimization strategies based on transition time and intelligently control equipment start-stop.

[0203] See also Figure 3 As shown in the figure, the steps to develop an equipment start-stop optimization strategy include:

[0204] Step A: Preset a start-stop sequence table. The equipment in the start-stop sequence table is sorted according to the start-stop sequence of the equipment. The equipment is the equipment that needs to be started or stopped during the dry-wet state conversion process of the unit; the start-stop sequence table is determined by technical personnel in this field by observing the start-stop sequence of the equipment during the historical dry-wet state conversion process of the unit. The equipment ranked first in the start-stop sequence table is the first equipment to be started or stopped.

[0205] Step B: Preset the time interval and divide the transition time into H time points according to the time interval; the time interval is pre-set by a technician in this field based on the historical equipment start and stop interval and the actual situation; for example, the transition time is 10 seconds and the time interval is 2 seconds, then 5 time points are divided, namely the 2nd second, 4th second, 6th second, 8th second and 10th second.

[0206] Step C: Construct N point label sets based on H time points.

[0207] Step D: Randomly select a set of point labels and mark it as the selected set, and calculate the interference degree of the selected set.

[0208] Step E: Return to step D until all N point label sets are marked as selected sets, then do not return to step D and proceed to step F.

[0209] Step F: Sort the interference levels of the N point label sets from small to large to generate a first sorting table, obtain the interference level at the top of the first sorting table, and mark it as the minimum interference level, obtain the point label set corresponding to the minimum interference level, and mark it as the optimal point label set.

[0210] Step G: According to the optimal point label set and the start-stop sequence table, set the corresponding start-stop time for each device.

[0211] In the above step C, set the corresponding digital label for each time point in the transition time and mark it as a point label. The point label range is ; According to the start-stop sequence table, count the number of devices M, ; Filter the point labels, filter out M point labels from the point label range each time, and build a point label set, , , .

[0212] In the above step D, the expression of interference degree is: ;

[0213] Where, is the degree of interference, For the The amount of interference to the i-th data in the working condition characteristic data when the equipment is started or stopped, .

[0214] The method for obtaining the interference amount on the working condition characteristic data when each device is started and stopped includes:

[0215] Set a different digital label for each device and mark it as the device label. The range of the device label is ; Take a device tag and one of the optimized and adjusted working condition characteristic data as a set of evaluation data, and obtain Group assessment data, ;Will The group evaluation data are input into the trained data analysis model to predict the corresponding interference amount;

[0216] The specific training process of the data analysis model is consistent with the training process of the time prediction model, and it is also a deep neural network model.

[0217] In the above step G, the point labels in the optimal point label set are sorted from small to large to generate a second sorting table; the point labels in the second sorting table are matched one by one with the devices in the start-stop sequence table in positive order, and the corresponding time point is obtained according to the point label corresponding to each device. The time point corresponding to each device is the start-stop time corresponding to each device.

[0218] It should be understood that the start and stop times of each device are different. The reason is that if multiple devices are started and stopped at the same time, the strong disturbances caused will be superimposed, thus forming a stronger comprehensive disturbance.

[0219] This embodiment collects and processes operating condition analysis data to accurately determine the unit status; uses an improved clone selection algorithm to optimize and adjust the unit characteristic parameters, making it easier to find the global optimal solution for the unit characteristic parameters and achieve accurate adjustment of the unit characteristic parameters; at the same time, taking into account the strong disturbance caused by equipment start-up and shutdown, deep learning technology is used to establish a time prediction model to predict the transition time, and an equipment start-up and shutdown optimization strategy is formulated based on the transition time. The equipment start-up and shutdown are intelligently controlled to resist the strong disturbance caused by equipment start-up and shutdown, thereby improving the optimization effect of the unit characteristic parameters; digital and data-driven intelligent control is used to effectively solve the control problems in the unit load section, thereby improving the unit's operating efficiency and performance, and further improving the unit's operating safety and control robustness.

[0220] Example 2

[0221] See also Figure 4 As shown, this embodiment further improves the design based on Example 1. The start-stop sequence table in Example 1 only includes one start-stop sequence. However, in actual situations, the dry-wet state conversion process of the unit includes two situations: dry-to-wet state conversion and wet-to-dry state conversion. The start-stop sequence of the equipment in these two situations is different. It is necessary to formulate an equipment start-stop optimization strategy based on the actual dry-wet state conversion situation of the unit. Therefore, this embodiment provides a unit load segment control system, which also includes a state identification module.

[0222] The state identification module is used to analyze the processed data and identify the unit conversion state.

[0223] The unit conversion status includes dry-wet conversion and wet-dry conversion; dry-wet conversion means that the unit is converted from dry state to wet state, and wet-dry conversion means that the unit is converted from wet state to dry state; two start-stop sequence tables are preset, one of which corresponds to dry-wet conversion, and the other corresponds to wet-dry conversion.

[0224] Methods for identifying the unit transfer status include:

[0225] The load rate obtained in real time is subtracted from the load rate obtained last time to obtain the load rate difference; the load rate obtained last time is the load rate obtained after processing the working condition analysis data collected last time;

[0226] It should be noted that since the load factor is higher when the unit is in dry operating conditions and lower when it is in wet operating conditions, the load factor should decrease when the unit is converted from dry to wet conditions, and increase when it is converted from wet to dry conditions. That is, the calculated load factor difference can be used as one of the identification factors of the unit conversion state.

[0227] The load rate difference and data change rate are used as research data, and the research data are input into the trained state recognition model. According to the output recognition label, the corresponding recognition result is obtained, and the unit conversion state is determined according to the recognition result; the corresponding start and stop sequence table is obtained according to the unit conversion state.

[0228] The specific training process of the state recognition model includes:

[0229] Corresponding identification results are set in advance for g groups of research data, where g is an integer greater than 1, and the identification results include dry-wet conversion and wet-dry conversion. Different digital labels are set for the dry-wet conversion and the wet-dry conversion. For example, the digital label for the dry-wet conversion is set to 0, and the digital label for the wet-dry conversion is set to 1. The identification results corresponding to the research data are collected by those skilled in the art during the historical dry-wet operating condition conversion process of the unit, by collecting g groups of different research data and the corresponding unit conversion states, and setting corresponding identification results for the g groups of different research data in sequence;

[0230] Mark the digital labels of the recognition results as identification labels, and convert the research data and the corresponding identification labels into a corresponding set of feature vectors;

[0231] Each set of feature vectors is used as the input of the state recognition model. The state recognition model takes a set of predicted recognition labels corresponding to each set of research data as output, and the actual recognition labels corresponding to each set of research data as the prediction target. The actual recognition labels are the pre-set digital labels of the recognition results corresponding to the research data. The training goal is to minimize the sum of the prediction errors of all research data. The calculation formula of the prediction error is: ,in is the prediction error, is the group number of the eigenvector corresponding to the research data, For the The predicted identification labels corresponding to the group research data, For the The actual identification labels corresponding to the research data are set; the state recognition model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the state recognition model is specifically a deep neural network model;

[0232] This embodiment calculates the load rate difference and combines it with the data change rate, using deep learning technology to scientifically identify the unit conversion status, automatically select a reasonable start and stop sequence table, and achieve targeted equipment start and stop optimization; it can better adapt to different conversion situations in actual operation, thereby achieving more intelligent and precise load segment control, and further improving the stability and control accuracy of the system.

[0233] Example 3

[0234] See also Figure 5 As shown, for the parts not described in detail in this embodiment, please refer to the description of embodiment 1 and embodiment 2. A method for controlling the load segment of a unit is provided, and the method includes:

[0235] Real-time collection of unit operating condition analysis data;

[0236] Process the working condition analysis data and obtain processed data;

[0237] Analyze the processed data and evaluate the unit operating conditions;

[0238] Collect unit characteristic parameters in real time;

[0239] Optimize and adjust the unit characteristic parameters according to the unit operating conditions;

[0240] Predict the transition time based on the optimized and adjusted unit characteristic parameters;

[0241] According to the transition time, formulate equipment start-stop optimization strategy and intelligently control equipment start-stop.

[0242] Furthermore, the operating condition analysis data includes power data and operating condition characteristic data; the power data includes actual output power and rated power; the operating condition characteristic data is data whose value fluctuation exceeds a corresponding preset fluctuation threshold during the dry-wet state conversion process of the unit; the operating condition analysis data is collected according to a preset collection frequency;

[0243] The processing data includes load rate and data change rate;

[0244] The load factor is obtained by calculating the ratio of the actual output power to the rated power.

[0245] The data change rate is the change rate of each data in the working condition characteristic data;

[0246] The method for obtaining the data change rate is: subtract the corresponding data in the operating condition characteristic data collected last time from each data in the operating condition characteristic data collected in real time, obtain the data difference of each data in the operating condition characteristic data, and then obtain the change rate of each data in the operating condition characteristic data based on the ratio of the data difference of each data in the operating condition characteristic data to the corresponding data in the operating condition characteristic data collected last time.

[0247] Furthermore, the method for analyzing the processed data includes:

[0248] Preset load range value , ;

[0249] The load rate and load range values Make a comparison;

[0250] like , then generate preliminary conversion instructions;

[0251] like or , no preliminary conversion instruction is generated;

[0252] Preset threshold set, the threshold set includes rate of change threshold , , , the change rate threshold corresponds one-to-one with the data in the working condition characteristic data;

[0253] Each change rate in the data change rate Divide by the corresponding change rate threshold , obtain the standardized change rate of each data in the working condition characteristic data, ; Mark the data with a standardized change rate greater than 1 as changed data, count the number of changed data, and mark it as the number of changes ; Add each standardized rate of change in sequence to obtain the rate of change coefficient ;

[0254] like , then generate preliminary conversion instructions, Expressed as and;

[0255] like , no preliminary conversion instructions are generated, Expressed as or.

[0256] Furthermore, the method for evaluating the unit operating condition includes:

[0257] If two preliminary conversion instructions are generated, the unit operating condition is assessed to be in dry-wet state conversion;

[0258] If only one preliminary conversion instruction is generated or no preliminary conversion instruction is generated, it is assessed that the unit operating condition is not in the dry-wet state conversion.

[0259] Furthermore, the unit characteristic parameters are parameters that directly affect the operating characteristic data when the unit is self-regulated;

[0260] The steps for optimizing the unit characteristic parameters include:

[0261] Step 1: Preset the parameter adjustment range, which includes the adjustment range of each parameter in the unit characteristic parameters;

[0262] Step 2: Build A set of parameters, is an integer greater than 1. Each parameter set is constructed by randomly selecting a value from each adjustment range. Each parameter set is different; set a different numerical label for each different parameter set and mark it as a set label. The range of the set label is ;

[0263] Step 3: Use the improved clone selection algorithm to The best parameter set is selected from the parameter sets and marked as the best set, and the unit characteristic parameters are optimized and adjusted according to the best set.

[0264] Furthermore, the improved clone selection algorithm is used to The steps of selecting the best parameter set from the parameter sets include:

[0265] Step a: Randomly generate an initial antibody population of size q ; Preset iteration threshold; Among them, is the qth antibody, ;

[0266] Step b: Determine the fitness function;

[0267] Step c: Calculate the initial antibody population The fitness of each antibody in The antibody with the largest fitness and construct a memory population , number of cycles , ;

[0268] Step d: Memory population Each antibody in the cloning is cloned according to the calculated clone size to generate a clonal antibody population ,in is the qth cloned antibody in the clonal antibody population;

[0269] Step e: Cloning the antibody population Each antibody in the cloud uses the cloud adaptive mutation operator to perform cloud adaptive mutation to generate a mutant antibody group ;

[0270] Step f: Cloning the Antibody Population Perform antibody recombination to generate recombinant antibody populations ; and with the variant antibody group Merge to generate a pooled antibody population ;

[0271] Step g: Combined antibody population Use random replacement algorithm;

[0272] Step h: Calculate the dynamic boundary, generate the inverse solution corresponding to each antibody and add it to the merged antibody group middle;

[0273] Step i: If the number of cycles t is less than the iteration threshold, let , get a new memory population , loop step d to step h; if the number of iterations t is equal to the iteration threshold, the loop ends and the combined antibody group is obtained The set label corresponding to the antibody with the largest fitness is obtained according to the set label.

[0274] Furthermore, in step a, the initial antibody group The antibodies in the set correspond to the set labels one by one, and the range of the antibodies is the range of the set labels. ;

[0275] In step b, the fitness function is expressed as: ;

[0276] Where, is the fitness of the first antibody, is the matching degree of the first antibody, ;

[0277] Methods for obtaining the matching degree include:

[0278] According to the set label corresponding to the antibody, the corresponding parameter set is obtained; the parameter set, the unit characteristic parameter and one of the operating condition characteristic data are taken as a set of analysis data, and a total of Group analysis data, ;Will The group analysis data are input into the corresponding trained data prediction model respectively to predict the data in the corresponding working condition characteristic data, that is, the input and output of a data prediction model are the same data in the working condition characteristic data; the data prediction model has a total of indivual, , The data prediction model and the working condition characteristic data The data correspond one to one;

[0279] According to the operating condition characteristic data output by the data prediction model and the operating condition characteristic data collected last time, the data change rate is re-obtained and marked as the new data change rate; each change rate in the new data change rate is divided by the corresponding change rate threshold to obtain the new standardized change rate corresponding to each data in the operating condition characteristic data, each new standardized change rate is added in sequence to obtain the new change rate coefficient, and the new change rate coefficient is used as the matching degree.

[0280] Furthermore, The training process of a data prediction model includes:

[0281] The data prediction model whose input and output are both steam temperature is labeled as steam temperature prediction model. The training process of the steam temperature prediction model includes:

[0282] Collect e sets of analysis data corresponding to steam temperatures in advance, where e is an integer greater than 1, and convert the analysis data and the corresponding steam temperatures into a corresponding set of feature vectors;

[0283] Each set of eigenvectors is used as input to a steam temperature prediction model. The steam temperature prediction model uses a set of predicted steam temperatures corresponding to each set of analysis data as output, and uses the actual steam temperature corresponding to each set of analysis data as a prediction target. The actual steam temperature is the pre-collected steam temperature corresponding to the analysis data. The prediction results are evaluated for model accuracy using the mean absolute percentage error (MAPE). When the calculated MAPE is less than the preset MAPE, the steam temperature prediction model training is completed. A steam temperature prediction model is generated that predicts the steam temperature based on the analysis data. The steam temperature prediction model is a deep belief network model.

[0284] the remaining The training process of the data prediction model is consistent with that of the steam temperature prediction model, and both are deep belief network models.

[0285] Furthermore, in step d, the cloned antibody group ,in The qth cloned antibody in the clonal antibody population; the calculation method of the clone size includes:

[0286] ;

[0287] ;

[0288] Where, is the clone size of the first antibody, Int is the rounding function, q is the size of the antibody group, is the affinity of the first antibody, min is the minimum function, exp is the natural exponential function, is the Euclidean distance between the I-th antibody and the J-th antibody, , ;

[0289] Furthermore, in step f, from the cloned antibody group Randomly select b antibodies as parent antibodies for antibody recombination to generate offspring antibodies , until the clonal antibody population When the number of unselected antibodies is less than b, antibody recombination is stopped and a recombinant antibody group is generated. ; Progeny antibodies The calculation methods include:

[0290] ;

[0291] Where, is the b-th parent antibody, is the bth scale factor, the scale factor is a randomly generated real number and not all b scale factors are 0;

[0292] In step g, d new antibodies are randomly generated and the antibody group is merged. The d antibodies with the smallest fitness are replaced by d new antibodies generated randomly. .

[0293] Furthermore, in step h, the dynamic boundary is , obtain the combined antibody group The set labels corresponding to the antibody with the smallest fitness and the set labels corresponding to the antibody with the largest fitness are compared. is a collection label with a smaller value, is a collection label with a larger value;

[0294] Methods for generating inverse solutions include:

[0295] ;

[0296] Where, is the reverse solution of the first antibody, for The random number in is the collection label of the first antibody;

[0297] If the generated reverse solution is greater than or less than the dynamic boundary, it is marked as a transcendental solution, and a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution;

[0298] In step i, the reverse solution of the antibody is also used as the antibody to calculate the combined antibody group The fitness of all antibodies in the pool is calculated and sorted from large to small, retaining the top q antibodies; select the combined antibody group The most adaptable antibodies, forming a new memory population .

[0299] Furthermore, the transition time is the duration of the dry-wet state conversion process of the unit;

[0300] The method for predicting transition time comprises:

[0301] Obtain the predicted operating condition data corresponding to the best set and mark it as the best operating condition data. The predicted operating condition data is the operating condition characteristic data output by the data prediction model.

[0302] Based on the optimal operating condition data and the last collected operating condition characteristic data, the optimal change rate is calculated. The optimal change rate includes the change rate of each data in the optimal operating condition data. The load rate, optimal change rate and optimized unit characteristic parameters are used as test data. The test data is input into the trained time prediction model to predict the corresponding transition time.

[0303] The training process of the time prediction model includes:

[0304] Pre-collect the transition times corresponding to f groups of test data, where f is an integer greater than 1, and convert the test data and the corresponding transition times into a corresponding set of feature vectors;

[0305] Each set of feature vectors is used as the input of a time prediction model. The time prediction model outputs a set of predicted transition times corresponding to each set of test data, and uses the actual transition time corresponding to each set of test data as the prediction target. The actual transition time is the pre-collected transition time corresponding to the test data. Minimizing the sum of the prediction errors of all test data is used as the training goal. The time prediction model is trained until the sum of the prediction errors reaches convergence, and training is stopped. The time prediction model is a deep neural network model.

[0306] Furthermore, the step of formulating the equipment start-stop optimization strategy includes:

[0307] Step A: Preset the start-stop sequence table. In the start-stop sequence table, the equipment is sorted according to the start-stop sequence of the equipment. The equipment is the equipment that needs to be started or stopped during the dry-wet state conversion process of the unit;

[0308] Step B: Preset a time interval and divide the transition time into H time points according to the time interval;

[0309] Step C: Construct N point label sets based on H time points;

[0310] Step D: Randomly select a set of point labels and mark it as the selected set, and calculate the interference degree of the selected set;

[0311] Step E: Return to step D until all N point label sets are marked as selected sets, then do not return to step D and proceed to step F;

[0312] Step F: Sort the interference levels of the N point label sets from small to large to generate a first sorting table, obtain the interference level at the top of the first sorting table and mark it as the minimum interference level, obtain the point label set corresponding to the minimum interference level and mark it as the optimal point label set;

[0313] Step G: According to the optimal point label set and the start-stop sequence table, set the corresponding start-stop time for each device.

[0314] Furthermore, in step C, a corresponding digital label is set for each time point in the transition time and marked as a point label. The range of the point label is ; According to the start-stop sequence table, count the number of devices M, ; Filter the point labels, filter out M point labels from the point label range each time, and build a point label set, , , .

[0315] Furthermore, in step D, the expression of interference degree is: ;

[0316] Where, is the degree of interference, For the The amount of interference to the i-th data in the working condition characteristic data when the equipment is started or stopped, ;

[0317] The method for obtaining the interference amount on the working condition characteristic data when each device is started and stopped includes:

[0318] Set a different digital label for each device and mark it as the device label. The range of the device label is ; Take a device tag and one of the optimized and adjusted working condition characteristic data as a set of evaluation data, and obtain Group assessment data, ;Will The group evaluation data are input into the trained data analysis model to predict the corresponding interference amount;

[0319] The training process of the data analysis model is consistent with that of the time prediction model, and both are deep neural network models.

[0320] Furthermore, in step G, the point labels in the optimal point label set are sorted from small to large to generate a second sorting table; the point labels in the second sorting table are matched one by one with the devices in the start-stop sequence table in positive order, and the corresponding time point is obtained according to the point label corresponding to each device. The time point corresponding to each device is the start-stop time corresponding to each device.

[0321] Further, the processed data is analyzed to identify the unit conversion status;

[0322] The unit conversion state includes dry-wet conversion and wet-dry conversion; two start-stop sequence tables are preset;

[0323] Methods for identifying the unit transfer status include:

[0324] The load rate obtained in real time is subtracted from the load rate obtained last time to obtain the load rate difference; the load rate obtained last time is the load rate obtained after processing the working condition analysis data collected last time;

[0325] The load factor difference and data change rate are used as research data, and the research data is input into the trained state recognition model. The corresponding recognition results are obtained based on the output recognition labels, and the unit conversion state is determined based on the recognition results; the corresponding start-stop sequence table is obtained based on the unit conversion state;

[0326] The training process of the state recognition model includes:

[0327] Pre-setting corresponding recognition results for g groups of research data, where g is an integer greater than 1, and the recognition results include dry-wet conversion and wet-dry conversion, with different digital labels set for both dry-wet conversion and wet-dry conversion; marking the digital labels of the recognition results as recognition labels, and converting the research data and the corresponding recognition labels into a corresponding set of feature vectors;

[0328] Each set of feature vectors is used as the input of the state recognition model. The state recognition model takes a set of predicted identification labels corresponding to each set of research data as output, and the actual identification labels corresponding to each set of research data as prediction targets. The actual identification labels are pre-set digital labels of the identification results corresponding to the research data. Minimizing the sum of the prediction errors of all research data is used as the training goal. The state recognition model is trained until the sum of the prediction errors reaches convergence and training is stopped. The state recognition model is a deep neural network model.

[0329] Example 4

[0330] See also Figure 6 As shown, the present application further provides an electronic device 500. The electronic device 500 may include one or more processors and one or more memories. The memories may store computer-readable code, which, when executed by the one or more processors, may execute the above-described method for controlling a load segment of a unit.

[0331] The method or system according to the embodiment of the present application can also be used Figure 6 The electronic device architecture shown in FIG. Figure 6 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store a unit load segment control method provided in this application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 6 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 6 One or more components of an electronic device are shown.

[0332] Example 5

[0333] See also Figure 7 As shown, one embodiment of the present application discloses a storage medium 600. Computer-readable instructions are stored on the storage medium 600. When the computer-readable instructions are executed by a processor, a unit load segment control method according to an embodiment of the present application described with reference to the above figures can be executed. The storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc.

[0334] Furthermore, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions capable of being executed by a processor to perform instructions corresponding to the steps of the method provided herein, such as a unit load segment control method. When executed by a central processing unit (CPU), the computer program performs the functions defined in the method of the present application.

[0335] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

[0336] 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 scope of protection of the present invention.

Claims

1. A method for controlling a unit load segment, characterized in that: include: Real-time collection of unit operating condition analysis data; Process the working condition analysis data and obtain processed data; Analyze the processed data and evaluate the unit operating conditions; Collect unit characteristic parameters in real time; Optimize and adjust the unit characteristic parameters according to the unit operating conditions; Predict the transition time based on the optimized and adjusted unit characteristic parameters; According to the transition time, formulate equipment start-stop optimization strategy and intelligently control equipment start-stop; The operating condition analysis data includes power data and operating condition characteristic data; the power data includes actual output power and rated power; the operating condition characteristic data is data whose value fluctuation exceeds the corresponding preset fluctuation threshold during the dry-wet state conversion process of the unit; The working condition analysis data is collected according to the preset collection frequency; The processing data includes load rate and data change rate; The load factor is obtained by calculating the ratio of the actual output power to the rated power. The data change rate is the change rate of each data in the working condition characteristic data; The data change rate is obtained by: subtracting the corresponding data in the last collected working condition characteristic data from each data in the real-time working condition characteristic data to obtain the data difference of each data in the working condition characteristic data; and then obtaining the change rate of each data in the working condition characteristic data based on the ratio of the data difference of each data in the working condition characteristic data to the corresponding data in the last collected working condition characteristic data; The method for analyzing the processed data comprises: Preset load range value , ; The load rate and load range values Make a comparison; like , then generate preliminary conversion instructions; like or , no preliminary conversion instruction is generated; Preset threshold set, the threshold set includes rate of change threshold , , , the change rate threshold corresponds one-to-one with the data in the working condition characteristic data; Each change rate in the data change rate Divide by the corresponding change rate threshold , obtain the standardized change rate of each data in the working condition characteristic data, ; Mark the data with a standardized change rate greater than 1 as changed data, count the number of changed data, and mark it as the number of changes ; Add each standardized rate of change in sequence to obtain the rate of change coefficient ; like , then generate preliminary conversion instructions, Expressed as and; like , no preliminary conversion instructions are generated, Expressed as or.

2. A method for controlling a unit load section according to claim 1, characterized in that: The method for evaluating the unit operating condition includes: If two preliminary conversion instructions are generated, the unit operating condition is assessed to be in dry-wet state conversion; If only one preliminary conversion instruction is generated or no preliminary conversion instruction is generated, it is assessed that the unit operating condition is not in the dry-wet state conversion.

3. A method for controlling a unit load section according to claim 2, characterized in that: The unit characteristic parameters are parameters that will have a direct impact on the operating characteristic data when the unit is self-regulated; The steps for optimizing the unit characteristic parameters include: Step 1: Preset the parameter adjustment range, which includes the adjustment range of each parameter in the unit characteristic parameters; Step 2: Build A set of parameters, is an integer greater than 1. Each parameter set is constructed by randomly selecting a value from each adjustment range. Each parameter set is different; set a different numerical label for each different parameter set and mark it as a set label. The range of the set label is ; Step 3: Use the improved clone selection algorithm to The best parameter set is selected from the parameter sets and marked as the best set, and the unit characteristic parameters are optimized and adjusted according to the best set.

4. A method for controlling a unit load section according to claim 3, characterized in that: The improved clone selection algorithm is used to The steps of selecting the best parameter set from the parameter sets include: Step a: Randomly generate an initial antibody population of size q ; Preset iteration threshold; Among them, is the qth antibody, ; Step b: Determine the fitness function; Step c: Calculate the initial antibody population The fitness of each antibody in The antibody with the largest fitness and construct a memory population , number of cycles , ; Step d: Memory population Each antibody in the cloning is cloned according to the calculated clone size to generate a clonal antibody population ; Step e: Cloning the antibody population Each antibody in the cloud uses the cloud adaptive mutation operator to perform cloud adaptive mutation to generate a mutant antibody group ; Step f: Cloning the Antibody Population Perform antibody recombination to generate recombinant antibody populations ; and with the variant antibody group Merge to generate a pooled antibody population ; Step g: Combined antibody population Use random replacement algorithm; Step h: Calculate the dynamic boundary, generate the inverse solution corresponding to each antibody and add it to the merged antibody group middle; Step i: If the number of cycles t is less than the iteration threshold, let , get a new memory population , loop step d to step h; if the number of iterations t is equal to the iteration threshold, the loop ends and the combined antibody group is obtained The set label corresponding to the antibody with the largest fitness is obtained according to the set label.

5. A method for controlling a unit load section according to claim 4, characterized in that: In step a, the initial antibody group The antibodies in the set correspond to the set labels one by one, and the range of the antibodies is the range of the set labels. ; In step b, the fitness function is expressed as: ; Where, is the fitness of the first antibody, is the matching degree of the first antibody, ; Methods for obtaining the matching degree include: According to the set label corresponding to the antibody, the corresponding parameter set is obtained; the parameter set, the unit characteristic parameter and one of the operating condition characteristic data are taken as a set of analysis data, and a total of Group analysis data, ;Will The group analysis data are input into the corresponding trained data prediction model respectively to predict the data in the corresponding working condition characteristic data, that is, the input and output of a data prediction model are the same data in the working condition characteristic data; the data prediction model has a total of indivual, , The data prediction model and the working condition characteristic data The data correspond one to one; According to the operating condition characteristic data output by the data prediction model and the operating condition characteristic data collected last time, the data change rate is re-obtained and marked as the new data change rate; each change rate in the new data change rate is divided by the corresponding change rate threshold to obtain the new standardized change rate corresponding to each data in the operating condition characteristic data, each new standardized change rate is added in sequence to obtain the new change rate coefficient, and the new change rate coefficient is used as the matching degree.

6. A method for controlling a unit load section according to claim 5, characterized in that: The training process of a data prediction model includes: The data prediction model whose input and output are both steam temperature is labeled as steam temperature prediction model. The training process of the steam temperature prediction model includes: Collect e sets of analysis data corresponding to steam temperatures in advance, where e is an integer greater than 1, and convert the analysis data and the corresponding steam temperatures into a corresponding set of feature vectors; Each set of eigenvectors is used as input to a steam temperature prediction model. The steam temperature prediction model uses a set of predicted steam temperatures corresponding to each set of analysis data as output, and uses the actual steam temperature corresponding to each set of analysis data as a prediction target. The actual steam temperature is the pre-collected steam temperature corresponding to the analysis data. The prediction results are evaluated for model accuracy using the mean absolute percentage error (MAPE). When the calculated MAPE is less than the preset MAPE, the steam temperature prediction model training is completed. A steam temperature prediction model is generated that predicts the steam temperature based on the analysis data. The steam temperature prediction model is a deep belief network model. the remaining The training process of the data prediction model is consistent with that of the steam temperature prediction model, and both are deep belief network models.

7. A method for controlling a unit load section according to claim 6, characterized in that: In the step d, the cloned antibody group ,in The qth cloned antibody in the clonal antibody population; the calculation method of the clone size includes: ； ； Where, is the clone size of the first antibody, Int is the rounding function, q is the size of the antibody group, is the affinity of the first antibody, min is the minimum function, exp is the natural exponential function, is the Euclidean distance between the I-th antibody and the J-th antibody, , .

8. A method for controlling a unit load section according to claim 7, characterized in that: In the step f, from the cloned antibody group Randomly select b antibodies as parent antibodies for antibody recombination to generate offspring antibodies , until the clonal antibody population When the number of unselected antibodies is less than b, antibody recombination is stopped and a recombinant antibody group is generated. ; Progeny antibodies The calculation methods include: ； Where, is the bth parent antibody, is the bth scale factor, the scale factor is a randomly generated real number and not all b scale factors are 0; In step g, d new antibodies are randomly generated and the antibody group is merged. The d antibodies with the smallest fitness are replaced by d new antibodies generated randomly. .

9. A method for controlling a unit load section according to claim 8, characterized in that: In step h, the dynamic boundary is , obtain the combined antibody group The set labels corresponding to the antibody with the smallest fitness and the set labels corresponding to the antibody with the largest fitness are compared. is a collection label with a smaller value, is a collection label with a larger value; Methods for generating inverse solutions include: ； Where, is the reverse solution of the first antibody, for The random number in is the collection label of the first antibody; If the generated reverse solution is greater than or less than the dynamic boundary, it is marked as a transcendental solution, and a random number function is used to randomly generate a value within the dynamic boundary, and the randomly generated value is assigned to the transcendental solution; In step i, the reverse solution of the antibody is also used as the antibody to calculate the combined antibody group The fitness of all antibodies in the pool is calculated and sorted from large to small, retaining the top q antibodies; select the combined antibody group The most adaptable antibodies, forming a new memory population .

10. A method for controlling a unit load section according to claim 9, characterized in that: The transition time is the duration of the dry-wet state conversion process of the unit; The method for predicting transition time comprises: Obtain the predicted operating condition data corresponding to the best set and mark it as the best operating condition data. The predicted operating condition data is the operating condition characteristic data output by the data prediction model. Based on the optimal operating condition data and the last collected operating condition characteristic data, the optimal change rate is calculated. The optimal change rate includes the change rate of each data in the optimal operating condition data. The load rate, optimal change rate and optimized unit characteristic parameters are used as test data. The test data is input into the trained time prediction model to predict the corresponding transition time. The training process of the time prediction model includes: Pre-collect the transition times corresponding to f groups of test data, where f is an integer greater than 1, and convert the test data and the corresponding transition times into a corresponding set of feature vectors; Each set of feature vectors is used as the input of a time prediction model. The time prediction model outputs a set of predicted transition times corresponding to each set of test data, and uses the actual transition time corresponding to each set of test data as the prediction target. The actual transition time is the pre-collected transition time corresponding to the test data. Minimizing the sum of the prediction errors of all test data is used as the training goal. The time prediction model is trained until the sum of the prediction errors reaches convergence, and training is stopped. The time prediction model is a deep neural network model.

11. A method for controlling a unit load section according to claim 10, characterized in that: The steps of formulating the equipment start-stop optimization strategy include: Step A: Preset the start-stop sequence table. In the start-stop sequence table, the equipment is sorted according to the start-stop sequence of the equipment. The equipment is the equipment that needs to be started or stopped during the dry-wet state conversion process of the unit; Step B: Preset a time interval and divide the transition time into H time points according to the time interval; Step C: Construct N point label sets based on H time points; Step D: Randomly select a set of point labels and mark it as the selected set, and calculate the interference degree of the selected set; Step E: Return to step D until all N point label sets are marked as selected sets, then do not return to step D and proceed to step F; Step F: Sort the interference levels of the N point label sets from small to large to generate a first sorting table, obtain the interference level at the top of the first sorting table and mark it as the minimum interference level, obtain the point label set corresponding to the minimum interference level and mark it as the optimal point label set; Step G: According to the optimal point label set and the start-stop sequence table, set the corresponding start-stop time for each device.

12. A method for controlling a unit load section according to claim 11, characterized in that: In step C, a corresponding digital label is set for each time point in the transition time and marked as a point label. The point label range is ; According to the start-stop sequence table, count the number of devices M, ; Filter the point labels, filter out M point labels from the point label range each time, and build a point label set, , , .

13. A method for controlling a unit load section according to claim 12, characterized in that: In step D, the expression of interference degree is: ; Where, is the degree of interference, For the The amount of interference to the i-th data in the working condition characteristic data when the equipment is started or stopped, ; The method for obtaining the interference amount on the working condition characteristic data when each device is started and stopped includes: Set a different digital label for each device and mark it as the device label. The range of the device label is ; Take a device tag and one of the optimized and adjusted working condition characteristic data as a set of evaluation data, and obtain Group assessment data, ;Will The group evaluation data are input into the trained data analysis model to predict the corresponding interference amount; The training process of the data analysis model is consistent with that of the time prediction model, and both are deep neural network models.

14. A method for controlling a unit load section according to claim 13, characterized in that: In step G, the point labels in the optimal point label set are sorted from small to large to generate a second sorting table; the point labels in the second sorting table are matched one by one with the devices in the start-stop sequence table in positive order, and the corresponding time point is obtained according to the point label corresponding to each device. The time point corresponding to each device is the start-stop time corresponding to each device.

15. A method for controlling a unit load section according to claim 14, characterized in that: Analyze the processed data and identify the unit conversion status; The unit conversion state includes dry-wet conversion and wet-dry conversion; Two start and stop sequence tables are preset; Methods for identifying the unit transfer status include: The load rate obtained in real time is subtracted from the load rate obtained last time to obtain the load rate difference; the load rate obtained last time is the load rate obtained after processing the working condition analysis data collected last time; The load factor difference and data change rate are used as research data, and the research data is input into the trained state recognition model. The corresponding recognition results are obtained based on the output recognition labels, and the unit conversion state is determined based on the recognition results; the corresponding start-stop sequence table is obtained based on the unit conversion state; The training process of the state recognition model includes: Pre-setting corresponding recognition results for g groups of research data, where g is an integer greater than 1, and the recognition results include dry-wet conversion and wet-dry conversion, with different digital labels set for both dry-wet conversion and wet-dry conversion; marking the digital labels of the recognition results as recognition labels, and converting the research data and the corresponding recognition labels into a corresponding set of feature vectors; Each set of feature vectors is used as the input of the state recognition model. The state recognition model takes a set of predicted identification labels corresponding to each set of research data as output, and the actual identification labels corresponding to each set of research data as prediction targets. The actual identification labels are pre-set digital labels of the identification results corresponding to the research data. Minimizing the sum of the prediction errors of all research data is used as the training goal. The state recognition model is trained until the sum of the prediction errors reaches convergence and training is stopped. The state recognition model is a deep neural network model.

16. A unit load segment control system, implementing a unit load segment control method according to any one of claims 1 to 15, characterized in that: include: A data acquisition module is used to collect the unit's operating condition analysis data in real time; the operating condition analysis data includes power data and operating condition characteristic data; the power data includes actual output power and rated power; the operating condition characteristic data is data whose value fluctuation exceeds the corresponding preset fluctuation threshold during the unit's dry-wet state conversion process; The working condition analysis data is collected according to the preset collection frequency; A data processing module is used to process the operating condition analysis data and obtain processed data; the processed data includes a load rate and a data change rate; the load rate is obtained by: obtaining the load rate based on the ratio of the actual output power to the rated power; the data change rate is the change rate of each data in the operating condition characteristic data; The data change rate is obtained by: subtracting the corresponding data in the last collected working condition characteristic data from each data in the real-time working condition characteristic data to obtain the data difference of each data in the working condition characteristic data; and then obtaining the change rate of each data in the working condition characteristic data based on the ratio of the data difference of each data in the working condition characteristic data to the corresponding data in the last collected working condition characteristic data; The operating condition evaluation module is used to analyze the processed data and evaluate the operating condition of the unit; The method for analyzing the processed data includes: presetting the load range value , ; Load rate and load range values For comparison; if , then generate preliminary conversion instructions; if or , then no preliminary conversion instruction is generated; a preset threshold set, the threshold set includes rate of change threshold , , , the change rate threshold corresponds to the data in the working condition characteristic data one by one; each change rate in the data change rate Divide by the corresponding change rate threshold , obtain the standardized change rate of each data in the working condition characteristic data, ; Mark the data with a standardized change rate greater than 1 as changed data, count the number of changed data, and mark it as the number of changes ; Add each standardized rate of change in sequence to obtain the rate of change coefficient ;like , then generate preliminary conversion instructions, Expressed as and; if , no preliminary conversion instructions are generated, Expressed as or; Parameter collection module, used to collect unit characteristic parameters in real time; Parameter optimization module, used to optimize and adjust the unit characteristic parameters according to the unit operating conditions; The time prediction module is used to predict the transition time based on the optimized and adjusted unit characteristic parameters; The start-stop control module is used to formulate equipment start-stop optimization strategies based on transition time and intelligently control equipment start-stop.

17. An electronic device comprising 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, the unit load segment control method described in any one of claims 1-15 is implemented.

18. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed, the unit load segment control method according to any one of claims 1 to 15 is implemented.

Images