Wide-load intelligent dispatching management method and system for power plant

Abstract

The invention relates to the technical field of power load dispatching, in particular to a wide-load intelligent dispatching management method and system for a power plant, and the method comprises the steps: integrating a voltage signal with the operation parameters of each coal-fired unit, and forming an input data set; analyzing each coal-fired unit in the input data set by adopting a time-delay model, recording the time-delay time of each coal-fired unit, and obtaining a time-delay mapping relation deduced by the model at any moment; determining effective scheduling sub-intervals in different load intervals by utilizing a time delay mapping relation and combining adjustment stability under different load intervals; drawing a multi-delay change curve by taking the timestamps corresponding to the effective scheduling sub-intervals as horizontal axes and the load values and the delay time as longitudinal axes, and determining the preparation delay time of each effective scheduling sub-interval; and taking data corresponding to the effective scheduling sub-interval and the preparation delay time as input, and solving an optimal operation parameter at each moment. And the power dispatching efficiency and accuracy are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power load scheduling, in particular to a power plant wide load intelligent scheduling management method and system. BACKGROUND

[0002] Currently, the coal-fired unit coordinated control system mainly adopts the traditional load feedforward + PID control strategy. However, under the background of new power system construction, the coal-fired unit generally has problems such as unstable combustion due to variable coal quality, large fluctuation of main parameters in complex working conditions, safe operation of heating surface, and poor grid performance under wide load. The existing control technology has deficiencies in precise modeling, working condition adaptation, overall coordination ability, etc., and it is difficult to realize the comprehensive improvement of the coordinated control ability of the unit under complex working conditions. At the same time, due to the large inertia and nonlinear dynamic characteristics of the coal-fired unit itself, as well as the performance limitations of the conventional PID plus feedforward control strategy algorithm in traditional DCS control, the parameter operation stability and rapidity are not high.

[0003] For example, Chinese patent publication CN119994923A discloses a resource scheduling method, device, electronic equipment and storage medium. The method includes: obtaining adjustable load resource data of a power system within a preset time period, inputting the adjustable load resource data into a preset generation side low-carbon scheduling model to obtain n groups of resource scheduling modes, determining n groups of generation feature indexes corresponding to the n groups of resource scheduling modes, inputting the n groups of resource scheduling modes into a preset load side low-carbon scheduling model to obtain n groups of load side power consumption data, determining n groups of load side power consumption feature indexes corresponding to the n groups of load side power consumption data, determining n coordination degree values based on the n groups of generation feature indexes and the n groups of load side power consumption feature indexes, determining a maximum coordination degree value in the n coordination degree values, determining a resource scheduling mode corresponding to the maximum coordination degree value, and obtaining a target resource scheduling mode.

[0004] For example, Chinese patent publication CN119742795A discloses a power equipment load scheduling method and system based on big data analysis. A multi-feature time series machine learning model is used to predict the future power load in the selected scheduling area, and to determine whether to generate a power load supply shortage prompt. In the case of power load supply shortage, the power supply parameters of the standby power source are combined to determine whether there is a usable standby power source, and the power load of the standby power source is scheduled according to the scheduling duration.

[0005] Existing technologies quantify resource scheduling methods through power supply costs and complete power scheduling allocation by measuring equipment power supply preparation time. These methods tend to describe the power cost of equipment scheduling and the time cost of initiating equipment scheduling, but they ignore the time lag of power equipment and its impact. This makes it difficult to identify response lags in the power scheduling process, which can easily lead to instability and rigidity in regulation, resulting in redundancy and wasted time in power scheduling. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a power plant wide-load intelligent dispatch management method, including: S1, when the dispatch command is executed, continuously collecting the voltage signal on the high-voltage side, integrating the voltage signal with the operating parameters of each coal-fired unit to form an input dataset.

[0007] S2 uses a time delay model to analyze each coal-fired unit in the input dataset, records the time delay of each coal-fired unit, and obtains the time delay mapping relationship derived by the model at any time.

[0008] S3 utilizes time-delay mapping relationships and combines them with regulation stability under different load ranges to identify effective scheduling sub-ranges in different load ranges.

[0009] S4. Plot a multi-delay variation curve with the timestamp corresponding to the effective scheduling sub-interval as the horizontal axis and the load value and time delay as the vertical axis to determine the preparation delay time of each effective scheduling sub-interval.

[0010] S5 uses the data corresponding to the effective scheduling sub-interval and the preparation delay time as input to solve for the optimal operating parameters at each time step.

[0011] A power plant wide-load intelligent dispatch management system includes: a data acquisition module, used to continuously acquire voltage signals on the high-voltage side when dispatch instructions are executed, and integrate the voltage signals with the operating parameters of each coal-fired unit to form an input dataset.

[0012] The time delay analysis module is used to analyze each coal-fired unit in the input dataset using a time delay model, record the time delay time of each coal-fired unit, and obtain the time delay mapping relationship derived by the model at any time.

[0013] The interval judgment module is used to identify the effective scheduling sub-intervals in different load intervals by utilizing time delay mapping relationships and combining the adjustment stability under different load intervals.

[0014] The delay identification module is used to plot a multi-delay variation curve with the timestamp corresponding to the effective scheduling sub-interval as the horizontal axis and the load value and time delay time as the vertical axis, and to determine the preparation delay time of each effective scheduling sub-interval.

[0015] The parameter output module is used to solve for the optimal operating parameters at each time step by taking the data corresponding to the effective scheduling sub-interval and the preparation delay time as input.

[0016] The beneficial effects of this invention are as follows: First, by continuously collecting high-voltage side voltage signals and integrating the operating parameters of each coal-fired unit, this invention adopts a load interval approach, records the time delay time according to the load interval, generates a multi-dimensional state vector and fits the time delay mapping relationship, avoiding the problem of time delay mapping distortion caused by parameter confusion, and sets up quantitative data for the current coal-fired unit's response to dispatching instructions, which is convenient for subsequent analysis of parameter analysis under the influence of multiple time delays.

[0017] Second, this invention divides the load range into fine-grained load points, and determines the data that meets the normal parameter constraint conditions in the current load range in the form of parameter constraints and load fluctuations, so as to avoid the impact of invalid load points on the scheduling process, ensure that the load change of equipment under different load ranges can be determined, and ensure the safe and stable operation of the unit in a wide load range.

[0018] Third, this invention calculates the preparation delay time for dispatch instructions, determines key nodes such as the instruction issuance point, output achievement point, and average time delay when each dispatch instruction meets its target, and judges whether the instruction window meets the merging conditions to achieve the association and merging of multiple instruction issuances, enabling instruction association and interval time delay to be processed collaboratively; finally, it solves the optimal operating parameters for multiple objectives, so that the effective dispatch sub-intervals under time delay analysis can be matched with the corresponding parameters, thereby improving the efficiency and accuracy of power dispatch. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] Figure 1 This is a flowchart illustrating a power plant wide-load intelligent dispatching and management method.

[0021] Figure 2 This is a flowchart illustrating step S2 of a power plant wide-load intelligent dispatch management method.

[0022] Figure 3 This is a flowchart illustrating step S3 of a power plant wide-load intelligent dispatch management method.

[0023] Figure 4 This is a flowchart illustrating step S4 of a power plant wide-load intelligent dispatch management method.

[0024] Figure 5 This is a system framework diagram of a power plant wide-load intelligent dispatch and management system. Detailed Implementation

[0025] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0026] See Figure 1 A power plant wide-load intelligent dispatch management method includes: S1, when the dispatch command is executed, continuously collecting the voltage signal on the high-voltage side, integrating the voltage signal with the operating parameters of each coal-fired unit to form an input dataset.

[0027] S2 uses a time delay model to analyze each coal-fired unit in the input dataset, records the time delay of each coal-fired unit, and obtains the time delay mapping relationship derived by the model at any time.

[0028] S3 utilizes time-delay mapping relationships and combines them with regulation stability under different load ranges to identify effective scheduling sub-ranges in different load ranges.

[0029] S4. Plot a multi-delay variation curve with the timestamp corresponding to the effective scheduling sub-interval as the horizontal axis and the load value and time delay as the vertical axis to determine the preparation delay time of each effective scheduling sub-interval.

[0030] S5 uses the data corresponding to the effective scheduling sub-interval and the preparation delay time as input to solve for the optimal operating parameters at each time step.

[0031] In step S1, voltage signals from the high-voltage side will be continuously collected and bound to the operating parameters of each coal-fired unit. The operating parameters may include boiler temperature, pressure, flow rate, coal input, air-coal ratio, and induced draft fan speed, which represent the operating data of the coal-fired unit, as well as the dispatch instructions and load data set by the current power grid under dispatch, so as to monitor the working status of each coal-fired unit under the power grid dispatch.

[0032] When the voltage signal on the high-voltage side is known, the effective value of the high-voltage side line voltage and the effective value of the high-voltage side line current can be measured through the voltage signal. Then, its active power, reactive power, and the ratio of active power to the rated active power of the unit (percentage load) can be obtained directly. Thus, the actual load of the coal-fired unit under the grid dispatch can be known, and the binding of these load values ​​with the operating parameters of each coal-fired unit can be calculated.

[0033] The implementation of step S1 includes: when any scheduling instruction is obtained, determining the scheduling instruction for each coal-fired unit during operation based on the operating parameters of each coal-fired unit; determining the active power under the current operation based on the voltage signal under the execution of the scheduling instruction, recording the load range of each coal-fired unit based on the active power, and synchronizing the value range of the load range to the input dataset.

[0034] During the execution of wide-load dispatch, the load range at the time of each dispatch command will be recorded in real time to determine the working status of each coal-fired unit. The load range represents the range of percentage load under the corresponding command, and explains the current output active power of the coal-fired unit.

[0035] The output data will be divided according to the value range of its load interval, forming multiple sets of input datasets, such as priority scheduling load interval with a load percentage below 35%, regular load interval with a load percentage of 45%-100%, and transition load interval with a load percentage of about 40%±5%. By quantifying the processing method under different load interval scheduling in the form of values ​​of each load interval, the waiting time and processing efficiency required for each scheduling instruction can be determined.

[0036] In one embodiment of the present invention, such as Figure 2 As shown, the implementation of step S2 includes: S21, recording the time delay of each coal-fired unit after the execution of the dispatch instruction, based on the load interval corresponding to the input dataset.

[0037] The currently recorded time delay is the time difference between the moment when the input parameters begin to change after the scheduling command is executed and the moment when the actual output begins to change significantly. This time difference is used to characterize the time delay characteristics corresponding to the current input parameters. As for the set time delay model, it is used to analyze the gain under the current scheduling command execution, thereby distinguishing the time delay mapping relationship under different load intervals.

[0038] When using a time-delay model, the MPC control algorithm is generally employed, and its model... ;in, This represents the output value of the MPC control algorithm, indicating the gain of the i-th load interval after the execution of the current scheduling instruction; i represents the index of different load intervals, such as i=1, 2, 3 corresponding to the priority scheduling load interval, the normal load interval, and the transition load interval, respectively. This represents the gain coefficient for the i-th load interval, indicating the proportion of steady-state output change caused by input variation; This represents the time delay of the i-th load interval. When recording the time delay, the input parameters must change by at least 2%, and the actual output must change by at least 5%, so as to determine the working status of each coal-fired unit under the current dispatch command. This represents the inertial time constant of the i-th load interval, reflecting the speed of the system response; This represents the Laplace operator. Based on the description of this time-delay model, the time delay and relative parameters under each scheduling instruction are recorded, thereby quantifying the control status of each scheduling instruction on the coal-fired unit.

[0039] S22, when the data in the input dataset changes, a multidimensional state vector is generated using each changed input data as a variable.

[0040] When generating a multidimensional state vector, it is necessary to determine the parameters in the input dataset that mainly affect the execution effect of scheduling instructions, and combine these parameters into a multidimensional state vector to obtain the mapping relationship under different load intervals.

[0041] The implementation of step S22 includes: setting the initial state based on the initial state of the input dataset and constructing an initial state vector.

[0042] For each initial state vector, determine the parameter category corresponding to the initial state vector; use the parameter combination under each parameter category as a set of reference state vectors.

[0043] The multidimensional state vector is obtained by taking the dimension and arrangement order of each baseline state vector.

[0044] When constructing a multidimensional state vector, it is first necessary to determine its parameter categories in order to filter the parameters that respond to the scheduling instructions. For example, the parameter categories will be set based on the category of the current scheduling instructions, such as grid-side instructions, unit-side fuel instructions, and unit-side steam instructions. Each instruction corresponds to multiple parameters, such as electrical parameters for grid-side instructions, combustion parameters for unit-side fuel instructions, and equipment parameters for unit-side steam instructions.

[0045] At this time, the grid-side command can correspond to the effective value of the high-voltage line voltage, the effective value of the high-voltage line current, the target load, and the grid frequency deviation. These parameters will directly affect the load tracking accuracy and the frequency regulation response effect during dispatch.

[0046] The unit-side fuel command can correspond to the coal input quantity, the same-layer coal powder concentration deviation, the same-layer coal powder flow rate deviation, and the actual air-coal ratio. These parameters can cause output deviation, thereby affecting the overall output load accuracy. Since there is a related time lag in combustion, it can affect the execution of power dispatch.

[0047] The steam commands on the unit side can correspond to the main steam pressure, main steam temperature, turbine valve opening, and metal tube wall temperature. These parameters represent the actual operating conditions of the coal-fired unit and can indirectly affect the actual output under power dispatch. They are the main focus of monitoring the normal operation of each piece of equipment. The introduction of time delay is to determine the dispatching situation under normal operation of each coal-fired unit and then to calculate the impact of time delay on each command.

[0048] When setting the baseline state vector, the implementation method also includes: using the real-time operating parameters of the coal-fired unit as operating state components.

[0049] The operating parameters controlled by the scheduling instructions are used as the instruction scheduling components.

[0050] The running state component and the instruction scheduling component are concatenated to form the base state vector of the output.

[0051] At this point, the real-time state and the target state are distinguished. For example, the parameters of the operating state component may include parameters such as the actual coal input, the actual air-coal ratio, the actual main steam pressure, and the actual valve opening. The command and scheduling component is the target value under the operating state component, which explains the relationship between the target value and the real-time value of the current multi-dimensional state vector in different load ranges.

[0052] As for the dimensions and arrangement order of each baseline state vector, the dimensions represent parameter categories and will be sorted according to the order of electricity → fuel → equipment to explain the sequential process under different parameter changes.

[0053] It should be noted that when setting up a multidimensional state vector, the gain of the multidimensional state vector will only be calculated if any parameter in the multidimensional state vector changes by more than 2% and the actual output changes by at least 5%. Otherwise, the fitting coefficients are only calculated using the corresponding parameters to complete the time-delay mapping relationship between the input dataset and the load range.

[0054] S23. Use multidimensional state vectors to fit the time delay and load interval at any time. Determine the time delay mapping relationship between the input dataset and the load interval by fitting the load interval in each load interval.

[0055] During fitting, the parameters corresponding to the multidimensional state vector during generation will be used as inputs, along with the multidimensional state vector and the load interval. The data within each load interval will be processed using a quadratic polynomial fitting method.

[0056] For example, its fitting method is expressed as: ;in, It represents the time delay under any parameter variation in the real-time load percentage and multidimensional state vector; This represents the real-time load percentage, and its value corresponds to the range of each load interval. Parameters representing multidimensional state vectors, such as coal content and valve opening, represent parameters that can be adjusted and affect the operation of coal-fired units after dispatch instructions are issued; , , and This represents the fitting coefficient for the i-th load interval. This fitting coefficient is obtained by the least squares method to illustrate the changes that occur in each load interval when the input dataset changes.

[0057] The final output time-delay mapping relationship will record its gain value, fitting coefficient, load range, etc., to explain the relative gain under each instruction execution.

[0058] In one embodiment of the present invention, in step S3, the load points under different load ranges are determined by using the time delay mapping relationship, and each data point is judged to determine whether it conforms to the description of stable regulation, so as to prevent the coal-fired unit operating parameters from changing too much when the dispatching command is executed, resulting in unstable overall output and reduced dispatching efficiency.

[0059] like Figure 3 As shown, the implementation of step S3 includes: S31, dividing the load interval into multiple load points according to the actual load percentage value within the load interval; each divided load point represents a certain sampling time, such as the value of a set of input datasets within the load interval at a sampling interval of 1s. At this time, based on the load value of each coal-fired unit at each scheduling, the values ​​of multiple load points within the current load interval will be determined. When all constraints or indicators are satisfied, the load point is marked as valid, thereby distinguishing the sub-intervals of each load interval that can effectively respond to the instructions under different scheduling instructions, thus describing the load scheduling situation under multiple scenarios.

[0060] S32 retrieves the range and fluctuation amplitude of the time delay corresponding to the load point, and determines the controllability of the time delay corresponding to each load point.

[0061] The core of time delay controllability is that the time delay time is within a reasonable range and the fluctuation amplitude is small. Its controllability can be determined by extracting the grid's mandatory requirements on the unit's response speed and converting them into the upper limit of the time delay. Then, by combining the physical characteristics of multiple devices (the time delay time required in the process of combustion → steam → output), the lower limit of the time delay can be determined. Finally, by combining its historical operating parameters, the fluctuation of the upper and lower limits of the time delay can be determined, thereby determining the time delay controllability of each load point.

[0062] Therefore, the range of time delay values ​​will be derived from the time delay mapping relationship, and the fluctuation amplitude of the time delay will be quantified according to the standard deviation of the same load point in 10 consecutive calculations. For example, the time delay value range in the priority scheduling load interval is [25s, 45s], with a fluctuation amplitude ≤ 5s; the time delay value range in the normal load interval is [15s, 30s], with a fluctuation amplitude ≤ 2s. The normal section has stable operating conditions, and its time delay fluctuation needs to be even smaller; the value in the transition load interval is between the priority scheduling and normal conditions, with a time delay value range of [20s, 35s] and a fluctuation amplitude ≤ 3s. This is used to judge the time delay controllability of the load point under multiple load intervals.

[0063] The values ​​given above are for illustrative purposes only. If the scheduling operation data is to be used as the main setting, the upper and lower limits of the time delay can be set in the form of the average value ± 2 times the standard deviation according to the confidence interval of the time delay. The standard deviation can be calculated according to the historical data of the same load point to quantify the fluctuation range, thereby explaining the response to the scheduling command under the current situation.

[0064] S33: After the time delay controllability is determined, query the constraint conditions corresponding to the load point, and use the data value of each load point at the corresponding time point to determine the adjustment stability under the execution of the scheduling command.

[0065] When verifying regulation stability, the constraints are mostly voltage signal stability constraints to prevent large voltage fluctuations from causing a decline in power quality. For example, first determine if the operating parameters of the coal-fired unit, such as main steam pressure and pulverized coal concentration, fall within the confidence interval under normal operation. Then, examine the load deviation and voltage fluctuation corresponding to the actual output, using these two as the primary constraints. Once both the load deviation and voltage fluctuation are within the normal range for their respective load intervals, it can be determined that the voltage fluctuation is relatively small under the execution of dispatch commands. Finally, the load points that meet the constraints are output.

[0066] The current load point represents the smallest dispatch unit load value that is divided from the original load range according to the sensitivity of the operating conditions. For example, with a step size of 0.5%, each 0.5% is a load point. Then, the absolute deviation between the actual processing of the same unit at the same time and the target load of the dispatch instruction is viewed to define the load deviation. At the same time, the voltage fluctuation at the corresponding time is retrieved to determine the regulation stability of the dispatch instruction execution.

[0067] At this point, each load point will retrieve data from the multidimensional state vector at the same timestamp. One is the actual output under the motion state component, and the other is the target load under the command scheduling component. This will give the absolute deviation between the actual output and the target load. This deviation represents the relative deviation of the data corresponding to the load value at a certain minimum scheduling unit.

[0068] It should be noted that each load point represents the load value statistically recorded at a corresponding time. Each load point directly corresponds to specific data at a given time. As for the deviation corresponding to the load point, this deviation indicates the difference between the target load of the scheduling instruction and the actual load at the corresponding time, which is the main content for verifying the stability of the load point.

[0069] For load deviation at load points, different judgment thresholds will be set based on different load intervals, and the current load deviation will be required to be less than the judgment threshold. For example, the priority scheduling / normal / transition load intervals will be set to 3%, 2% and 2.5% respectively, requiring the load deviation at the corresponding time point to be as small as possible.

[0070] Meanwhile, the voltage fluctuation at the corresponding time will be determined according to the power grid regulations. Regardless of the load range, the voltage fluctuation amplitude shall not exceed ±2% of the rated voltage. Specifically, for each load range, the voltage fluctuation can be set to ±1.8%, ±1.5%, and ±1.6% of the rated voltage in the order of priority dispatch / regular / transitional load range. In addition to these values, the voltage fluctuation can also be set using the confidence interval of historical data to verify the specific situation handled at the corresponding time of the load point.

[0071] S34, load points that pass both time delay controllability and regulation stability judgments are considered valid load points. Continuously valid load points are merged to form an output valid scheduling sub-interval.

[0072] Only load points that pass both the stability and time delay controllability tests are considered valid load points to explain the effective execution of scheduling instructions at the corresponding time. The corresponding load points are then grouped into a sub-interval. Essentially, this sub-interval is a set of load points selected from the original load interval that have controllable time delay, stable regulation, and meet the constraints, with the aim of ensuring that scheduling instructions can be executed accurately. The corresponding data can be regarded as a reliable peak-shaving and frequency regulation data pool, which facilitates optimization and control during subsequent scheduling instruction execution.

[0073] In one embodiment of the present invention, this step analyzes multi-time-delay change curves to identify load abrupt change points before and after scheduling, predicts the risk of time-delay superposition, and determines the preparation delay time under the corresponding scheduling command, so as to determine the response of each unit under time-delay response.

[0074] The resulting multi-time-delay variation curve has a horizontal axis representing a timestamp, a left vertical axis representing the load value (which can be expressed as actual output or load percentage), and a right vertical axis representing the time delay, with the scale indicating the time delay range of the sub-interval.

[0075] like Figure 4 As shown, the implementation of step S4 includes: S41, based on the multi-time-delay change curve, obtaining the instruction issuance point, output target achievement point, and time delay average of the effective scheduling sub-interval, respectively, as a list of key nodes of the multi-time-delay change curve; the instruction issuance point represents the time point when the scheduling instruction is issued. If there are multiple consecutive instructions within the same effective scheduling sub-interval, the time point of each scheduling instruction needs to be recorded; the output target achievement point requires the time point when the actual load coincides with the target load or the time point when it enters ±1%, to determine the time when the actual output reaches the target load.

[0076] As for the time delay mean, it represents the average value under the execution of a single scheduling instruction. If there are multiple scheduling instructions in the same effective scheduling sub-interval, then the average value of these instructions needs to be taken.

[0077] S42, based on the input list of key nodes, determine the instruction window when each scheduling instruction is issued, and according to the sending status of each instruction window, determine whether the scheduling instructions before the current instruction window meet the merging conditions.

[0078] After obtaining the list of key nodes, the time window for each scheduling instruction is divided from the time period of the currently effective scheduling sub-interval. It is then determined whether multiple instruction windows can be merged to avoid recalculating the preparation delay time.

[0079] The dimensions of the merging conditions will be based on the mean deviation of the time delay (obtained by the coefficient of variation of the time delay time, and quantified by the ratio of the standard deviation to the mean), the instruction window timing (representing the time interval between two instruction windows), and the load change rate, thereby determining the merging situation under the execution of multiple scheduling instructions.

[0080] Therefore, the implementation of step S42 includes: for any instruction window, determining the merging conditions of the current instruction window; the dimensions of the merging conditions will be based on the mean deviation of the time delay, the timing of the instruction window, and the load change rate.

[0081] When all merging conditions are met, the scheduling instructions preceding the current instruction window are considered to have met the merging conditions.

[0082] The merging condition for the mean deviation of time delay is considered to be met when the coefficient of variation between consecutive instruction windows is ≤5%.

[0083] For command window timing, the time interval between the command window of the previous scheduling command and the current command window must be less than a preset time interval. In other words, it is determined that the command window corresponding to the previous scheduling command and the current command window are either continuous or overlapping. The preset time interval can be set based on the average time interval between consecutive command windows, or a confidence interval with a 95% confidence level can be selected and set in the form of its average value ± 1.96 standard deviations. The upper limit of the confidence interval is set as the preset time interval to cover the response requirements of more than 95% of the units, thereby indicating the continuity between command windows.

[0084] As for the conditions for merging load change rates, it requires that after merging the load values ​​of two instruction windows, the load target change rate of multiple instructions is at least less than 2%, that is, the actual scheduling targets of multiple scheduling instructions are close, so that data under the same description can be merged.

[0085] S43, if the merging conditions are met, set the preparation delay time for the corresponding effective scheduling sub-interval based on the merged instruction issuance point and output target point.

[0086] After the merging conditions are met, the time points of multiple scheduling instructions are merged, and the earliest instruction issuance point and the latest output target achievement point are re-acquired. The difference between these two time points indicates the preparation delay time required for actual scheduling completion after the current scheduling instruction is issued. This time will be bound to the corresponding load interval to determine the required time length for the corresponding load interval under scheduling adjustment.

[0087] S44. If the merging condition is not met, determine whether each scheduling instruction has been executed and extract the instruction window length corresponding to the current scheduling instruction, and output the preparation delay time based on its window length.

[0088] If the merging conditions are not met, it indicates that the interval between each scheduling instruction is too long. It is necessary to determine whether each scheduling instruction has been completed. This can be done by checking the load value. Check whether the load value is within 1% of the target load at consecutive time points. If the values ​​at consecutive time points meet the condition, it means that the scheduling instruction has been completed. In this case, the output preparation delay time can be obtained by subtracting the instruction issuance point from the output target point.

[0089] If the load value does not approach the target load, it means that the current instruction has not been completed or the current scheduling instruction has not been executed to the required standard. It is necessary to mark the corresponding data and continuously obtain its load value in order to obtain the time point when the scheduling instruction was completed.

[0090] In one embodiment of the present invention, in step S5, in addition to the parameters representing the interval and time nature, the effective scheduling sub-interval and the preparation delay time essentially correspond to a set of parameters under certain constraints. For example, in step S3, the effective scheduling sub-interval restricts multiple sets of data constraints, output deviations, etc., and the preparation delay time represents the time constraints when the adjustment command is started and when the target is reached. Under the parameter set constrained by these two parameters, the optimal operating parameters at the corresponding time are solved, and then the output data combination can be obtained.

[0091] Therefore, the implementation of step S5 includes: S51, setting objective functions for each effective scheduling sub-interval with the objectives of minimizing coal consumption, minimizing output deviation, and minimizing parameter fluctuation.

[0092] When setting the objective function, priority should be given to minimizing coal consumption, minimizing output deviation, and minimizing parameter fluctuations. Minimizing coal consumption means the ratio of coal consumption rate to the maximum allowable coal consumption rate in each effective dispatch sub-interval. Similarly, minimizing output deviation means the ratio of output deviation to the maximum allowable output deviation, which can be used to ensure responsiveness to dispatch instructions. Minimizing parameter fluctuations means the ratio of main steam pressure fluctuations, CO concentration fluctuations, and voltage fluctuations to the maximum allowable fluctuations. The average of these three fluctuation ratios will be calculated to illustrate the parameter fluctuation situation and avoid equipment fatigue under power dispatch.

[0093] The objective function is set by weighting the values ​​of minimizing coal consumption, minimizing output deviation, and minimizing parameter fluctuations. The weights can be set to 0.4, 0.3, and 0.3 respectively. Priority is given to ensuring the coal consumption of coal-fired units, and their response and fluctuation are checked simultaneously. The weights can also be quantified by the proportion of each parameter to the total eigenvalues ​​through eigenvalue decomposition, thus determining the priority of each parameter under power dispatch.

[0094] S52, based on the objective function, performs particle swarm optimization to determine the optimal operating parameters corresponding to the effective scheduling sub-interval.

[0095] When solving particle swarm optimization, the following steps can be followed: 1. Initialize 100 parameter particles, each parameter particle corresponds to a set of data. Assume that the air-coal ratio and coal consumption are used as schematic values, with an air-coal ratio of 1.4-1.5 and a coal consumption of 80-90 t / h.

[0096] 2. Calculate the objective function value for each particle; 3. Iterate 50 times, moving the particle towards the minimum objective function value; 4. Calculate the parameters corresponding to the particle with the minimum F; The final output parameters are the optimal operating parameters.

[0097] When the effective scheduling sub-interval belongs to different load intervals, the calculation method will be adjusted. For example, if it is in the priority scheduling load interval, the above calculation process will be used to obtain the optimal operating parameters under fast scheduling.

[0098] If the load is within the normal range, the objective function will obtain multiple sets of parameters through multiple iterations after the outlier solution is completed. Then, based on the solution values ​​of these parameters for minimizing coal consumption, the operating parameters that minimize coal consumption as much as possible will be selected to obtain the optimal operating parameters for the normal mode.

[0099] If the load is in the transitional load range, the objective function will be additionally filtered for load change rate during particle swarm initialization, selecting particles in the transitional condition, such as particles with load change rate ≤ 5% / min, and finally iterating to find the optimal operating parameters for the transitional load range.

[0100] like Figure 5 As shown, the present invention also provides a power plant wide-load intelligent dispatch management system, including: a data acquisition module, a time delay analysis module, an interval judgment module, a delay identification module, and a parameter output module; wherein, the output end of the data acquisition module is connected to the time delay analysis module, the output end of the time delay analysis module is connected to the interval judgment module, the output end of the interval judgment module is connected to the delay identification module, and the output end of the delay identification module is connected to the parameter output module.

[0101] The data acquisition module is used to continuously acquire voltage signals from the high-voltage side when scheduling instructions are executed, and integrate the voltage signals with the operating parameters of each coal-fired unit to form an input dataset.

[0102] The time delay analysis module is used to analyze each coal-fired unit in the input dataset using a time delay model, record the time delay time of each coal-fired unit, and obtain the time delay mapping relationship derived by the model at any time.

[0103] The interval judgment module is used to identify the effective scheduling sub-intervals in different load intervals by utilizing time delay mapping relationships and combining the adjustment stability under different load intervals.

[0104] The delay identification module is used to plot a multi-delay variation curve with the timestamp corresponding to the effective scheduling sub-interval as the horizontal axis and the load value and time delay time as the vertical axis, and to determine the preparation delay time of each effective scheduling sub-interval.

[0105] The parameter output module is used to solve for the optimal operating parameters at each time step by taking the data corresponding to the effective scheduling sub-interval and the preparation delay time as input.

[0106] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for intelligent dispatching and management of power plants under wide load conditions, characterized in that, include: S1 continuously collects voltage signals from the high-voltage side when the dispatching command is executed, and integrates the voltage signals with the operating parameters of each coal-fired unit to form an input dataset; S2 uses a time delay model to analyze each coal-fired unit in the input dataset, records the time delay time of each coal-fired unit, and obtains the time delay mapping relationship derived by the model at any time. S3, by utilizing the time delay mapping relationship and combining the regulation stability under different load intervals, the effective scheduling sub-intervals in different load intervals are identified; S4. Plot a multi-delay variation curve with the timestamp corresponding to the effective scheduling sub-interval as the horizontal axis and the load value and time delay time as the vertical axis to determine the preparation delay time of each effective scheduling sub-interval. S5 uses the data corresponding to the effective scheduling sub-interval and the preparation delay time as input to solve for the optimal operating parameters at each time step.

2. The intelligent dispatching and management method for power plants with wide load range according to claim 1, characterized in that, The implementation methods for step S1 include: When any scheduling instruction is obtained, the scheduling instruction for each coal-fired unit is determined based on the operating parameters of each coal-fired unit. The active power under the current operation is determined by the voltage signal under the execution of the scheduling instruction, and the load range of each coal-fired unit is recorded by the active power. The value range of the load range is synchronized to the input dataset.

3. The intelligent dispatching and management method for power plants with wide load range according to claim 1, characterized in that, Step S2 can be implemented in the following ways: S21, based on the load interval corresponding to the input dataset, record the time delay of each coal-fired unit after the execution of the dispatch instruction; S22, When the data in the input dataset changes, a multidimensional state vector is generated using each changed input data as a variable; S23. Use multidimensional state vectors to fit the time delay and load interval at any time. Determine the time delay mapping relationship between the input dataset and the load interval by fitting the load interval in each load interval.

4. The intelligent dispatching and management method for power plants with wide load range according to claim 3, characterized in that, The implementation methods of step S22 include: An initial state vector is constructed based on the initial state of the input dataset. For each initial state vector, determine the parameter category corresponding to the initial state vector; use the parameter combination under each parameter category as a set of reference state vectors. The multidimensional state vector is obtained by taking the dimension and arrangement order of each baseline state vector.

5. The intelligent dispatching and management method for power plants with wide load range according to claim 4, characterized in that, When setting the baseline state vector, the implementation methods also include: The real-time operating parameters of the coal-fired power unit are used as operating status components; The operating parameters controlled by the scheduling instructions are used as the instruction scheduling components; The running state component and the instruction scheduling component are concatenated to form the base state vector of the output.

6. The intelligent dispatching and management method for power plants with wide load range according to claim 1, characterized in that, Step S3 can be implemented in the following ways: S31, divide the load interval into multiple load points according to the actual load percentage within the load interval; S32, retrieve the value range and fluctuation range of the time delay corresponding to the load point, and determine the controllability of the time delay corresponding to each load point; S33, after the time delay controllability is determined, query the constraint conditions corresponding to the load point, and use the data value of each load point at the corresponding time point to determine the adjustment stability under the execution of the scheduling command; S34, load points that pass both time delay controllability and regulation stability judgments are considered valid load points. Continuously valid load points are merged to form an output valid scheduling sub-interval.

7. The intelligent dispatching and management method for power plants with wide load range according to claim 1, characterized in that, Step S4 can be implemented in the following ways: S41. Based on the multi-time-delay variation curve, obtain the instruction issuance point, output target point and time delay average of the effective scheduling sub-interval, and use them as the key node list of the multi-time-delay variation curve. S42, based on the input list of key nodes, determine the instruction window when each scheduling instruction is issued, and determine whether the scheduling instructions before the current instruction window meet the merging conditions according to the sending status of each instruction window; S43, if the merging conditions are met, set the preparation delay time for the corresponding effective scheduling sub-interval based on the merged instruction issuance point and output target achievement point; S44. If the merging condition is not met, determine whether each scheduling instruction has been executed and extract the instruction window length corresponding to the current scheduling instruction, and output the preparation delay time based on its window length.

8. The intelligent dispatching and management method for power plants with wide load range according to claim 7, characterized in that, The implementation methods of step S42 include: For any instruction window, determine the merging conditions for the current instruction window; the dimensions of the merging conditions will be based on the mean deviation of the time delay, the timing of the instruction window, and the load change rate. When all merging conditions are met, the scheduling instructions preceding the current instruction window are considered to have met the merging conditions.

9. The intelligent dispatching and management method for power plants with wide load range according to claim 1, characterized in that, Step S5 can be implemented in the following ways: S51, with the objectives of minimizing coal consumption, minimizing output deviation and minimizing parameter fluctuation, set objective functions for each effective scheduling sub-interval; S52, based on the objective function, performs particle swarm optimization to determine the optimal operating parameters corresponding to the effective scheduling sub-interval.

10. A power plant wide-load intelligent dispatching and management system, characterized in that, include: The data acquisition module is used to continuously acquire voltage signals from the high-voltage side when dispatch instructions are executed, and integrate the voltage signals with the operating parameters of each coal-fired unit to form an input dataset; The time delay analysis module is used to analyze each coal-fired unit in the input dataset using a time delay model, record the time delay time of each coal-fired unit, and obtain the time delay mapping relationship derived by the model at any time. The interval judgment module is used to identify the effective scheduling sub-intervals in different load intervals by utilizing time delay mapping relationships and combining the adjustment stability under different load intervals. The delay identification module is used to plot a multi-delay variation curve with the timestamp corresponding to the effective scheduling sub-interval as the horizontal axis and the load value and time delay time as the vertical axis, and to determine the preparation delay time of each effective scheduling sub-interval. The parameter output module is used to solve for the optimal operating parameters at each time step by taking the data corresponding to the effective scheduling sub-interval and the preparation delay time as input.

Images